Tech & AI
530 issues · 530 keepers · 37 tier-5 · 493 tier-4
9 tier-5 · 76 tier-4
No company anchors The Information's coverage like OpenAI, and the running argument across these pieces is that the most-watched private company in tech is also the most financially precarious. The reporters break OpenAI's actual numbers from shareholder disclosures - $4.3B first-half sales against $2.5B of cash burn, a path to $25B+ annualized revenue, a projected $115B burn through 2029 - and trace the structural bets meant to close the gap: a pivot from premium subscriptions to an $8 ad-supported ChatGPT, a serial restructuring of the Microsoft relationship (revenue share cut toward 8%, Microsoft's 27% stake, the AGG clause), and a $100B-plus mega-round courting SoftBank, Nvidia, and Amazon. Read together they form the single best longitudinal record of whether the AI era's flagship can grow into its valuation before the money runs out.
Stephanie Palazzolo reports on a coding technique OpenAI uses to improve its ChatGPT agents, framed around how increasingly capable agents threaten standalone startups. A substantive AI-engineering/strategy piece, though the email is a thin teaser with the detail behind the paywall.
An analysis of why ChatGPT can't access Slack, framing a broader enterprise data war in which incumbents restrict API and data access to AI startups. It matters because data-access control is becoming a key competitive moat as AI agents try to reach into enterprise systems.
An exclusive scoop that Peter Thiel's Founders Fund and Dragoneer have each committed over $1 billion to the second $30 billion installment of OpenAI's record $40 billion raise, with SV Angel adding $200 million. It maps out who is backstopping the largest private financing in tech and signals continued insider conviction in OpenAI.
Exclusive that OpenAI's upcoming GPT-5 is markedly improved at coding per an early user, a meaningful pre-release signal in the developer-tools/coding-agent race against Anthropic's Claude. The email is the paywalled teaser, but the scoop itself is a high-interest data point on GPT-5's positioning.
The Information's exclusive scoop reporting OpenAI's annualized revenue reaching $12 billion as ChatGPT weekly active users pass 700 million. A meaningful single-data-point scoop on the most-watched AI company's revenue trajectory and user scale; the standalone exclusive carries more reference value than the digest restatement. Body is paywalled but the metric itself is the story.
The flagship exclusive detailing OpenAI's technical struggles building GPT-5 -- dwindling high-quality training data, diminishing pre-training returns, and o3 gains that degraded when shipped as a chatbot -- as evidence the industry's exponential-progress era is slowing. Landmark reporting widely cited as a marker that scaling laws are bending.
Explains how OpenAI's 'universal verifiers' could extend reinforcement-learning gains beyond coding and math into harder-to-grade domains, a key technique for improving model performance as pre-training plateaus. A substantive explainer of a frontier-model training method.
Stephanie Palazzolo's Applied AI column unpacks GPT-5's shift to an all-in-one system that auto-routes how much compute to spend (removing model choice from users/developers) and reveals from OpenAI's model cards that its new open-weight models cost under $500K and ~$5M to train, undercutting DeepSeek. It frames the open question of whether training-efficiency gains will offset labs' ever-rising compute appetite, and adds a segment on AI hiring-tool discrimination liability (the Workday/HiredScore case). Substantive analysis with original cost data and a clear thesis on training economics.
Stephanie Palazzolo argues GPT-5 is a good model that nonetheless undershot the hype, marking a 'trough of disillusionment' moment for the AI cycle. A substantive single-author analysis of where expectations vs. reality have diverged, though the body is paywalled in the email.
SoftBank's Vision Fund 2 sits on $22B in investment losses, but a single bet-OpenAI-could erase them and vindicate Masayoshi Son personally (he holds a 17.25% fund stake). A sharp Dealmaker analysis of how concentrated SoftBank's fortunes now are on one AI position.
Ken Brown's Briefing uses crypto exchange Bullish's 84% IPO pop (~$10B value on $80M of prior-year profit) to argue crypto has become a $400B-plus, possibly house-of-cards slice of a frothy stock market-ironically embracing the traditional finance it claims to disrupt-plus a note on OpenAI surviving Musk's motion to dismiss. A pointed, well-reasoned skeptic's take on the crypto-IPO wave.
An analysis ('The Power of Defaults') of how the major model makers are offering discounted pricing, case studies and perks to become the default model inside third-party apps, recognizing that default placement drives durable usage. A substantive look at distribution strategy and competitive dynamics in the model market.
Martin Peers' nightly column builds an argument from Bret Taylor and Winston Weinberg's TITV remarks that AI agent firms will end up with healthy SaaS-like margins while foundation-model makers (OpenAI, Anthropic) resemble low-margin infrastructure providers, mirroring how cable/telecom apps out-earned the pipe-builders. A genuinely useful analytical framing on where AI-stack value accrues, above the usual nightly roundup.
Applied AI essay rebutting the 'intelligence too cheap to meter' narrative from Altman and Pichai: state-of-the-art model prices haven't fallen, and agents consume far more tokens (forcing Cursor and Replit to raise prices), so AI buyers are scrambling to pass costs on. Counterpoint from Sierra's Bret Taylor argues token cost is irrelevant when agents undercut human labor 100x. A sharp, framework-level take on AI unit economics that reframes the cost debate.
Exclusive revealing that OpenAI, while trying to unseat Google in search, relies on Google search results (scraped from the web) to power ChatGPT's answers on current events. Exposes a deep strategic dependence at the heart of the search-vs-chatbot battle and the paradox of rivals being mutually reliant. A landmark insight into AI competitive dynamics with lasting reference value.
An exclusive that OpenAI's employee tender could grow from $6B to up to $8B at a $500B valuation, which would rival the entire volume of startup-employee share sales in all of last year. A meaningful datapoint on OpenAI's soaring valuation and the scale of insider liquidity in private AI.
Amir Efrati's column reads OpenAI's hire of a health-products VP as a clear signal it will build consumer and physician health products, leaning on ChatGPT's strength at diagnosis and lab analysis and likely chasing rivals like OpenEvidence and AI scribes (Abridge, Nuance). A substantive analysis of how a major lab is positioning for a huge vertical market.
A 'Big Read' analysis arguing that Elon Musk's lawsuit against OpenAI over its for-profit conversion has more legal merit than dismissive observers assume, as the Musk-Altman feud escalates toward a courtroom showdown. Matters because the case bears on whether OpenAI can complete its restructuring ahead of a possible IPO.
Meta's AI leaders have discussed embedding third-party models from Google and OpenAI into Meta AI and other app features, a striking admission of weakness for a company that has championed its own Llama models. It signals that even a top-tier lab may treat foundation models as commoditized infrastructure to buy rather than build when its own offerings lag.
Daily 'Information AM' digest leading with the landmark Google remedy ruling-no Chrome/Android divestiture, exclusive-distribution deals barred, mandatory search-data sharing, Apple's $20B/yr payments likely preserved-plus Anthropic's $13B raise at $170B valuation, OpenAI's $1.1B Statsig acquisition (Vijaye Raji to CTO of Applications), and Nvidia's Solver buy. Dense, high-signal roundup of a major news day.
The original exclusive (Aaron Holmes): Microsoft will pay for Anthropic AI to power Office 365 apps, a notable partial shift away from its OpenAI dependence amid restructuring tensions. A consequential Big Tech / AI-vendor scoop on the fraying Microsoft-OpenAI exclusivity; email body paywalled but the story is the source of the next day's headlines.
Amir Efrati's analytical Briefing decoding why Microsoft and OpenAI issued a vague 49-word non-binding MOU this week-reading it as triple-purpose PR: Microsoft reassuring investors amid Oracle's $300B OpenAI deal and the Anthropic-in-Copilot hedge, OpenAI signaling to the California AG (agreeing the nonprofit gets a $100B+ stake), and both calming prospective investors as OpenAI raises to cover $115B in cash burn. Sharp motive analysis plus a curated stories-of-the-week map; strong analytical column.
Exclusive detailing OpenAI's projection to shareholders that it will cut the revenue share paid to commercial partners (chiefly Microsoft) from ~20% today to ~8% by 2030, retaining over $50B it needs to offset record compute costs-despite Microsoft's contractual 20% entitlement through 2030. A concrete, numbers-backed window into the OpenAI-Microsoft economic renegotiation that underpins the restructuring. Substantive scoop.
Exclusive scoop on how Anthropic and OpenAI are training AI 'co-workers' by setting them loose inside clones of enterprise apps to learn real workflows. The technique is central to the enterprise-agent race and to whether AI can credibly replace junior knowledge-worker roles. Body is paywalled here but the lede establishes a meaningful methodological angle on agent development.
A Dealmaker breakdown of OpenAI's approximate post-restructuring cap table, showing the stakes of major backers once the for-profit conversion completes. A high-reference-value ownership chart that clarifies who controls OpenAI ahead of its restructuring and potential IPO.
Human domain experts are struggling to teach OpenAI's and Anthropic's frontier models new things because the models increasingly outpace their tutors-a structural challenge for the RLHF/expert-data approach to model improvement. An important AI-Agenda piece flagging a looming ceiling in post-training data, with lasting relevance to how labs keep models improving.
A skeptical analysis arguing Oracle's stock surge on its massive new AI cloud contract (largely OpenAI) rests on shakier ground than investors assume. A useful contrarian True Value column dissecting the economics and concentration risk behind Oracle's AI-driven re-rating.
An analysis comparing OpenAI's and Anthropic's newly disclosed data on how their chatbots are actually used. Substantive AI-Agenda reporting on usage patterns that matter for understanding consumer vs. enterprise adoption of the two leading models.
OpenAI is stepping up poaching of Apple hardware talent and tapping Apple's supply-chain partners as it builds its own consumer AI devices. A well-sourced scoop on OpenAI's hardware ambitions and the emerging talent war with Apple.
Reported piece on the booming business of AI red-teaming and security firms that probe and break OpenAI and Anthropic models. Captures an emerging market category in AI safety and adversarial testing as model deployment scales.
The flagship exclusive breaking OpenAI's first-half 2025 financials from shareholder disclosures: ~$4.3B revenue (16% more than all of 2024), $2.5B cash burn, with detail on R&D, ChatGPT operating costs, and non-cash stock comp. Rare hard financial visibility into the most-watched private AI company, high lasting reference value for tracking OpenAI's economics.
OpenAI offered roughly $500M for a startup chiefly to acquire its videogame data, plus context on Sora's IP pivot. Illuminates how valuable specialized training data has become and the lengths OpenAI will go to secure it.
Peers unpacks OpenAI's 'Everything Everywhere All at Once' day: the AMD chip deal plus the DevDay reveal that apps (Zillow, Spotify, Expedia, Figma, Canva) now run inside ChatGPT, arguing Altman may have finally cracked the U.S. 'super app' that eluded everyone. Thoughtful analysis of ChatGPT as a potential primary portal for digital activity, plus the Verizon CEO shake-up.
Aaron Holmes's Applied AI column examines the recurring panic that OpenAI demos (DocuGPT, Agent Builder) will kill incumbent SaaS, triggering selloffs in DocuSign, HubSpot, Salesforce and Zapier's competitive moat anxiety. The grounded thesis is that the threat hasn't materialized because AI still can't reliably reproduce the customized workflows locking customers into apps like Salesforce, even as labs insist it's only a matter of time. Useful, balanced analysis of AI-vs-SaaS disruption, plus a clear OpenAI-vs-Google agent-builder comparison.
Wayt's sharp strategic essay on Amazon's dilemma after Walmart became the first major retailer on ChatGPT Instant Checkout: as the dominant e-commerce site with a huge ad business, partnering with OpenAI risks cannibalization, yet refusing could be 'like refusing to let Google index your site in 2000.' Strong analysis of platform strategy under AI shopping disruption (plus a colorful Paxos $300T fat-finger aside).
Applied AI dispatch with two substantive threads: Salesforce leaning on its Regrello acquisition to push Agentforce into supply-chain 'donkey work' (with Dell as flagship), and a detailed case study of how Southeast Asia's Grab picks models task-by-task-GPT-4o/4.1 fine-tuned on 140k voice samples for driver assistants, Claude Sonnet for merchant empathy, Gemini for safety. A useful real-world enterprise-adoption explainer showing models are not commodities.
Editor Jessica Lessin's pick framing the Dreamforce story as a strategic retreat: Benioff now concedes AI capability is 'outstripping customer adoption' and Salesforce is opening Agentforce/CRM data to ChatGPT, an acknowledgement of OpenAI/Anthropic's growing leverage over enterprise buyers and that LLMs aren't interchangeable commodities. A curated, thesis-driven analysis of an incumbent scrambling to defend its turf.
Briefing unpacking OpenAI's completed restructuring into a near-normal company controlled by its nonprofit board, turning Microsoft's $13B stake into $135B while a Musk lawsuit (trial next March) keeps a cloud over it. Reflects on OpenAI as a singular corporate creation with $1.4T in compute commitments, and the lingering question of who ultimately controls the nonprofit.
Reports on a newly public Ilya Sutskever deposition revealing Anthropic expressed 'excitement' about merging with OpenAI in the chaotic days after the 2023 Altman firing-a deal that could have put Dario Amodei atop OpenAI. A rare documented window into how the AI labs' leaders maneuvered during the boardroom crisis, with lasting reference value on AI-industry history.
An Editor's Pick wrapping The Information's scoop on Anthropic's internal projections: up to $70B revenue by 2028, cash-flow-positive as soon as 2027 (vs OpenAI's 2030), with >80% of revenue from API/enterprise and Claude Code nearing $1B ARR. The numbers are a genuinely useful reference on Anthropic's enterprise-first, capital-efficient strategy, even though delivered as a teaser.
The full Applied AI analysis behind the agent free-for-all: 16 incumbents from Microsoft to IBM to OpenAI now sell near-identical general-purpose agents, leaving buyers bewildered, while the real winners are the model providers (OpenAI, Anthropic) that power most agents and squeeze incumbents' gross margins, even as they launch competing agents of their own. A useful structural read on enterprise AI's value-capture shift, backed by a tracking table and buyer quotes.
Exclusive on an internal Sam Altman memo declaring a 'code red' to marshal resources toward improving ChatGPT as Google and other rivals close in, forcing OpenAI to delay initiatives like advertising. It is a concrete signal of competitive pressure reshaping OpenAI's roadmap and priorities at a pivotal moment in the AI race.
Applied AI scoops that Anthropic, OpenAI, Google, Microsoft and others are forming the Linux Foundation-organized Agentic AI Foundation to standardize open-source agent tooling, initially focused on MCP, OpenAI's Agents.md, and Block's Goose. The piece frames it as analogous to interbank payment standards and grounds it in real CIO concerns about MCP security (prompt injection, patching). Substantive analysis of the emerging agent-interoperability stack and why standards matter for enterprise adoption.
Single-article scoop that OpenAI, Anthropic and Google are courting specialized-data deals - talking to Revvity, Xero, and biotech/healthcare/fintech firms to license genomics and other proprietary datasets for model training. Signals the next frontier of the data-acquisition race as public web data is exhausted. Substantive original reporting on a structural AI-industry trend.
Applied AI explainer on UCLA Anderson building OpenAI/Anthropic-powered tools to automate teaching-assistant work - essay feedback, recommendation-letter drafting via StackAI - as universities face enrollment and funding pressure. Faculty remain unconvinced AI can replace human TAs, and cost savings are unproven. A concrete enterprise/education AI-adoption case study with named tools and honest caveats.
Applied AI scoop on Microsoft tackling weak 365 Copilot adoption by subsidizing customer training (tens to hundreds of thousands of dollars) when large enterprises buy seats, illustrated via the City of Raleigh's ~200-seat deal. Frames a broader enterprise-AI truth: the bottleneck is often user skill, not the tool, with Microsoft needing to prove Copilot beats cheaper ChatGPT/Gemini, and OpenAI facing the same usage gap. Useful, original reporting on AI go-to-market and adoption economics.
Applied AI analysis contrasts Amazon's aggressive 'Buy for Me' agentic-shopping approach (scraping/listing third-party products without permission) against OpenAI/Shopify's slower, opt-in API-based checkout, arguing Amazon's speed could let it ceded ground to none while alienating retailers. Useful strategic framing of two diverging agentic-commerce playbooks, plus a sharp 'Dell says people don't care about AI PCs' aside.
Applied AI deep-dive on AI-security firm Tenzai's research showing websites vibe-coded by OpenAI, Anthropic, Cursor, Replit and Devin all carry exploitable flaws (e.g. negative-quantity checkout credits, leaked buyer data), with no tool a clear winner. Substantive, evidence-based caution on vibecoding at scale, plus a colorful note on Linus Torvalds praising Google's Antigravity tool.
Martin Peers analyzes OpenAI's same-day launch of ChatGPT ads and an $8/month ChatGPT Go tier as narrative-control moves ahead of a ~$100B raise, benchmarking the ad ramp against Netflix's slow ad build and projecting ~$2B in 'new products' revenue this year rising past $10B by 2027. Sharp framing of why ads were inevitable given ~900M mostly-free users and how the tiered ad/no-ad model mirrors Google and Amazon. A substantive analytical take, plus a Thinking Machines exodus sidebar.
Exclusive revealing the mechanics of OpenAI's first ChatGPT ad push: dozens of advertisers offered sub-$1M trial commitments, view-based (not click-based) pricing, an early-February launch, and self-service tooling still in development. The pricing and structure details matter because they signal how the most-watched new ad inventory in tech-reaching ~900M weekly users-will actually be sold. Substantive single-story scoop on a major Big Tech monetization shift.
Applied AI essay on OpenAI CFO Sarah Friar's 'value sharing' thesis-that OpenAI wants profit-sharing stakes or IP licenses in customer outcomes (drug discovery, energy, finance) rather than just selling model access or ads. It frames a genuinely new monetization model that rivals Anthropic, Google DeepMind and Isomorphic are also chasing, and ties in Anthropic's 23%-over-forecast inference costs and the AI-pin hardware race. Worth reading as a signal of where frontier-lab business models may head.
Exclusive on OpenAI's enterprise offensive-Altman wooing CEOs like Disney's Bob Iger over a lavish dinner, pitching OpenAI as a one-stop AI shop (ChatGPT, Codex, workflow models) to win business from Anthropic. Substantive scoop on the central enterprise-AI rivalry, though the email is a single-article teaser.
Aaron Holmes' Applied AI essay establishes a market structure for AI app subscriptions: Microsoft's 15M paying 365 Copilot seats put it No. 2 behind ChatGPT's 35M, ahead of Google Gemini Enterprise's ~8M and Anthropic. Analyzes Microsoft's room to grow into 400M Office subscribers and the Azure-vs-Copilot capacity tradeoff Hood flagged. A useful framing of who's actually monetizing enterprise AI.
OpenAI is hiring hundreds of forward-deployed engineers to build custom AI apps and agents for big enterprises, a move to counter Anthropic and pave the way to monetize business customers ahead of an IPO. A clean exclusive documenting the consulting-led enterprise land grab and the FDE staffing trend.
Microsoft CCO Judson Althoff emailed sales staff talking points to compete against OpenAI's new Frontier agent platform, which automates office tasks across apps including Microsoft's own. The scoop is concrete evidence of the Microsoft-OpenAI rivalry hardening into a direct enterprise-agent turf war.
A Briefing whose lead dissects the financial engineering behind Masayoshi Son's OpenAI bet: SoftBank repaid the Vision Fund 2 loan and released Son's ~$1B personal guarantee, while he retains a 17.25% profit-share that could pay him billions personally if OpenAI trades up post-IPO. A genuinely original analysis of an under-examined governance/incentive structure, above typical Briefing fare.
An exclusive scoop that OpenAI is in advanced talks to hire OpenClaw founder Peter Steinberger and other maintainers of the viral open-source personal-agent software, with Meta also competing for him. It matters as a marker of the talent war over consumer AI agents and a possible acqui-hire of the hottest open-source agent project.
Applied AI analysis contrasting Anthropic's 'replace the worker, keep the software' positioning with OpenAI's explicit pitch to investors that its agents will replace Salesforce/Workday/Adobe, and documents SaaS firms (HubSpot, ServiceNow) plotting 'toll-gate' defenses against AI agents tapping their data. Worth reading for its concrete framing of the SaaS-vs-AI disruption battle, including a CrowdStrike-replacement case study.
Exclusive reporting that Amazon's up-to-$50B stake in OpenAI would be structured as $15B upfront plus $35B contingent on OpenAI reaching AGI or going public - terms shaped by Microsoft's existing Azure exclusivity-until-AGI clause. Matters because it reveals how 'AGI' is becoming a contractual financial trigger and shows the Microsoft partnership constraining new cloud-investor deals.
Reports that the Anthropic-Pentagon standoff (Friday deadline to grant 'unfettered access' or be cut off) has industry-wide implications, with OpenAI and Google facing the same surveillance and autonomous-weapons questions on their own military deals. Altman told staff he hoped to broker an industry-wide safeguard deal. Strong analysis showing the conflict extends well beyond one company.
AM digest leading with the Pentagon's extraordinary move to designate Anthropic a 'supply chain risk'-a sanction usually reserved for foreign adversaries-barring any contractor doing business with the US military from commercial activity with Anthropic, with unclear scope for cloud partners Amazon and Google. Companion items: Altman defending OpenAI's Pentagon deal, Amazon's $50B OpenAI investment, and Nvidia's new Groq-tech chip. The landmark sanction story carries lasting significance despite the roundup wrapper.
OpenAI has picked Cooley and Wachtell Lipton to prepare for an IPO that could come this year, one of the first concrete steps toward a listing (valued at $730B in the ongoing round, pre the $110B investment). Such an IPO would rank among the largest ever. A meaningful scoop signaling the IPO machinery is now in motion.
OpenAI hit $25B annualized revenue (up 17% from $21.4B at year-end), while Anthropic recently topped $19B, up nearly 3x from year-end and 36% in just two weeks. The data point quantifies how fast Anthropic is closing the revenue gap with its archrival. A concrete, reference-grade financial benchmark for the two leading labs.
A reported feature on Anthropic's simultaneous fights-its principled Pentagon standoff over surveillance and autonomous weapons that drew a 'supply chain risk' designation, and Dario Amodei's fiery internal memo disparaging OpenAI and Sam Altman. It captures how the AI-lab rivalries have become personal grudge matches tangled with national-security politics.
Exclusive that OpenAI will fold Sora video generation into ChatGPT, a strategy reversal after the standalone TikTok-style Sora app failed to gain traction, which could lift ChatGPT usage but sharply raise compute costs. A meaningful product-direction scoop on how OpenAI is consolidating consumer AI into its core chatbot.
An Applied AI deep dive arguing Palantir is better positioned than OpenAI or Anthropic to sell enterprise AI agents, because it already offers end-to-end data integration, the coveted forward-deployed-engineer model, and an 'AI action layer' atop legacy apps-evidenced by 109% U.S. commercial growth and customer cases like Centrus and World View. It also explains why OpenAI and Anthropic are courting PE firms (Blackstone, TPG, Bain) to learn enterprise selling. A framework-rich, comparative analysis with lasting reference value on the enterprise-AI battle.
Martin Peers' column dissects OpenAI's strategic 'refocus' on business customers and away from side quests, contrasting its management chaos with the deeper turmoil at Musk's Tesla, SpaceX and xAI, and argues long-view investors may forgive the U-turns. A sharp second riff floats whether Dustin Moskovitz should take Asana private amid the SaaS-apocalypse selloff, with the math worked out. Thoughtful opinion analysis that connects strategy to the IPO question.
As OpenAI prepares to widen ChatGPT ad sales next month, early advertisers report the process was 'low tech' and they've gotten little data proving ads drove measurable business outcomes. OpenAI is moving to streamline ad buying via ad-tech partnerships or its own ad management system. Matters as an early read on whether ChatGPT advertising can become a real revenue pillar for OpenAI's monetization push.
Peers uses OpenAI's abrupt Sora shutdown (which prompted Disney to abandon its $1B investment and deal) and the earlier PayPal Instant Checkout retreat to argue OpenAI's willingness to 'turn on a dime' risks its enterprise partnerships, just as it and Meta race into AI-driven shopping and the looming battle for retailer ad dollars. A coherent strategic thesis tying together partnership risk, commerce, and the Meta/OpenAI/Google ad rivalry.
Martin Peers parses OpenAI's $122B in commitments (much of it staged future cash from Amazon and SoftBank, not wired now), then pivots to a sharp retrospective on the IPO class of 2021 (Allbirds, Rent the Runway, BuzzFeed, UiPath, Bumble) to warn small investors how badly hyped listings can age. The class-of-2021 essay is a useful, evidence-rich caution as mega-IPOs loom.
Cory Weinberg digs into softening signals beneath OpenAI's $122B raise: SoftBank (a quarter of its NAV tied to OpenAI) down 17% this year, and secondary-market data from Caplight showing $1B of OpenAI sell orders versus only $200M of buy orders in 2026 so far, a sharp reversal from late 2025. A genuinely analytical read on private-market sentiment ahead of the IPO, beyond the headline number.
Exclusive reporting that Sam Altman wants OpenAI public as soon as Q4 despite a committed $600B five-year spend and expectations of burning $200B+ before turning cash-positive, while CFO Sarah Friar privately doubts the company will be ready this year. The CEO/CFO split exposes the financial tension underneath OpenAI's ambitions and is a meaningful signal for IPO timing.
Peers dissects the Anthropic-vs-OpenAI revenue race-Anthropic's $30B annualized rate implies it passed OpenAI's ~$24B, yet OpenAI carries an $852B valuation versus Anthropic's $380B-while flagging that the two count revenue differently and that 'annualized' is a shaky metric. A sharp Tesla section argues the stock stays expensive despite a 23% drop, propped only by Musk's robotaxi/robotics promises. Useful comparative framing of valuations and the annualized-revenue caveat.
Previously undisclosed OpenAI projections show a deliberate business-model gamble: the $8 ad-supported ChatGPT Go is forecast to surge ~36x to 112 million subscribers while ChatGPT Plus craters 80% to ~9 million, betting ad revenue on a larger free-ish base beats premium subscriptions. A landmark scoop quantifying OpenAI's pivot from subscription to advertising and the structural shift in its consumer economics.
Martin Peers argues Wall Street is belatedly waking up to OpenAI's troubles - missed revenue projections, a user base stuck at 920 million below the 1 billion goal, chaotic strategy pivots (shopping, Sora), Altman-Friar IPO tension, and Anthropic overtaking its run rate - while OpenAI bets its future on an uncertain consumer ad business. A pointed contrarian thesis on the company's deteriorating fundamentals amid renewed AI bullishness.
Previously undisclosed terms of OpenAI's renegotiated Microsoft deal cut its revenue-share obligation from a potential $135B (20% of revenue) to a fraction of that, saving OpenAI ~$97B through 2030. The scoop quantifies how dramatically OpenAI improved its economics versus its early backer as it pushes toward profitability and an IPO.
Anthropic is in advanced talks to acquire developer-tools startup Stainless for at least $300M; Stainless builds SDK/API-access software used by Anthropic, OpenAI and Google to help developers and AI agents reach models faster. The deal signals that infrastructure for agent-driven model access (Claude Code, OpenClaw) is becoming strategically valuable enough to vertically integrate.
Peers argues Microsoft-down ~16% YTD and the worst big-tech performer, with TCI dumping its stake over AI-disruption fears-is ripe for an activist push, possibly to spin off gaming/LinkedIn/Bing, drawing a parallel to ValueAct's 2013 stake that preceded the Nadella transformation. He counters that some negativity is overdone given ~40% Azure growth and OpenAI gains. Thoughtful strategic argument with concrete historical precedent and financial framing.
Courtroom reporting on Sam Altman's high-stakes testimony in Musk's lawsuit, where he argued Musk abandoned and tried to undermine the group they co-founded, while facing repeated cross-examination attacks on his honesty over allegations he 'stole' the OpenAI charity. The single-article deep coverage of the trial's pivotal moment. Substantive on-the-ground reporting of a defining Silicon Valley legal fight.
Data-driven analysis from The Information's Generative AI Database showing 34 leading AI startups now generate nearly $80 billion in annualized revenue ($6.6B/month), up 112% in six months-with Anthropic and OpenAI capturing 89% of it, a widening duopoly. The proprietary dataset gives a rare quantified picture of AI revenue concentration. High reference value for tracking the market's structure.
Martin Peers argues Google I/O's most significant move is folding AI agents and Gemini capabilities directly into Search, so the real competitive frame is now ChatGPT vs Google Search (3B users) rather than ChatGPT vs Gemini (900M). He flags the cost question-more compute and unclear ad integration-and that this makes Google hard to beat once all products are AI-overhauled. A sharp strategic read with a useful competitive reframing, though delivered as a nightly column.
A landmark competitive-reversal analysis: Anthropic's revenue grew 5x in five months to a ~$45B run rate, leapfrogging OpenAI's ~$33B, with Anthropic projecting a Q2 operating profit (5% margin) while OpenAI ran a negative 122% operating margin. The Briefing rigorously unpacks why-coding vs. chatbot monetization, OpenAI's free-user subsidy, the Microsoft 20% revenue share-and argues OpenAI's rush to IPO ahead of a financially stronger Anthropic is prudent. High reference value for tracking the two firms' divergence.
Profile of OpenAI CRO Denise Dresser (ex-Slack CEO) and how she cleaned up the company's clunky enterprise sales-untangling a confused Databricks co-selling partnership to land Hertz and KPMG. A look at OpenAI's maturing go-to-market machine as it tries to close the enterprise gap with Anthropic, with enterprise now ~40% of its revenue.
AM digest led by Anthropic's confidential draft S-1 filing (Q4 target), following its $65B raise at a $900B valuation and a reported $47B annualized revenue run rate-the first time it has overtaken OpenAI's valuation. Also carries Florida's lawsuit against OpenAI and Sam Altman over ChatGPT safety harms, Google's $80B equity raise for AI spend, and OpenAI poaching a Salesforce partnerships exec.
Single-article analysis estimating that OpenAI and Anthropic employees and early investors have sold a combined $14 billion in private shares over five years via tender offers. The thesis: these secondary sales have already relieved the employee-liquidity pressure that typically pushes startups toward IPOs, complicating the read on why both are now filing. A useful original data point for understanding AI-lab IPO dynamics. ---
8 tier-5 · 63 tier-4
This is the beat where the AI hype meets the procurement department, and it is where The Information's Applied AI column does its sharpest work. The throughline is a reckoning: will autonomous agents eat the incumbent SaaS stack (Salesforce, ServiceNow, Workday, Microsoft), or will the incumbents' workflow and data moats absorb the threat? The reporting refuses easy answers - it documents Benioff's stumbles with Agentforce, the unpredictable token bills that blow through annual budgets, the 'guardian' apps born from agents misbehaving, insurers refusing to cover AI damages, and the slow pivot from per-seat to consumption and outcome-based pricing. The framework pieces (the unit-economics of flat agent pricing, illusory software profits, the agent-management control plane) are the reusable mental models; the company-by-company dispatches show them playing out in real budgets.
Aaron Holmes details AWS's free 'Transform' agent that auto-rewrites Windows server code to Linux, neutralizing Microsoft's longtime cloud lock-in advantage (Thomson Reuters cut cloud costs ~30% and a year-long migration to two months). Includes analysis of OpenAI's long road to a Microsoft/Google productivity-suite rival. A full-bodied Applied AI essay with concrete enterprise case studies and strategic framing.
A 'True Value' analysis arguing Figma's hot IPO may not be the slam dunk investors expect, scrutinizing its growth outlook beneath the buzz. Substantive financial analysis of a closely-watched design-software listing.
Box CEO Aaron Levie argues that AI will entrench incumbent enterprise apps (Salesforce, Oracle, Workday) rather than disrupt them, because systems-of-record vendors are best positioned to expose powerful agents that ChatGPT-style orchestrators query. The issue also recounts the cautionary Replit vibe-coding incident where an AI agent deleted a user's production database. A substantive Applied AI essay on the agent-vs-incumbent debate.
An Applied AI column reporting AWS quietly failed to publicly disclose a hacker exploit in its Q Developer coding assistant (a VS Code extension that could be made to delete customer data), arguing the silence was a bigger misstep than the bug itself. Paired with a substantive ServiceNow case study on AI cutting ~5% of opex and the data-integration barriers blocking agent ROI; strong enterprise-AI analysis.
An exclusive arguing that as AI cuts software switching costs (rewriting code and migrating data off legacy systems), Microsoft's rivals are prying away longtime customers-reframing AI as a force that erodes enterprise-software lock-in and incumbency advantage. A substantive strategic thesis with broad implications for the enterprise SaaS landscape; pairs with the Databricks migration explainer. Body paywalled; the lock-in-erosion argument is the takeaway.
An exclusive that enterprise-AI maker Cohere told investors it expects to hit a $200M annualized revenue pace by year-end as it raises a new round. A concrete, original financial data point on a notable second-tier AI model maker, useful for tracking the enterprise-LLM competitive field. Body paywalled; the revenue figure is the scoop.
News report on Figma's blockbuster debut: shares soared 250% to push market value above $69B, making it the hottest in a string of recent tech IPOs. A significant market event marking the reopening of the tech IPO window.
Aaron Holmes' Applied AI essay explains why flat per-seat pricing breaks for AI agents: rising compute per task with flat token prices is squeezing margins, pushing Replit, Cursor and Anthropic toward effort/consumption-based pricing despite customer 'sticker shock.' A useful framework piece on AI software unit economics, backed by a Chargebee survey showing ~75% now use consumption pricing.
Kevin McLaughlin's Applied AI reveals Snowflake's undisclosed AI revenue (~$51M ARR mid-year, on track for a $100M target), with Cortex AISQL the biggest driver, as evidence enterprises keep choosing pricier brand-name models (OpenAI, Anthropic, Google) over cheaper alternatives. Substantive enterprise-software analysis with an exclusive data point and a clear thesis about why CIOs pay up despite thinner margins.
An AI Agenda exclusive on why a16z and Bain backed a three-year-old AI customer-service startup positioning itself as a 'Zendesk killer' in an increasingly crowded field. A useful VC-thesis and competitive-dynamics read on the AI customer-support category.
After years of falling prices, the cost of buying AI is rising again, benefiting Microsoft and other vendors as agentic workloads consume more compute. A useful enterprise-software/AI-economics analysis on the reversal of the deflationary pricing trend.
Anita Ramaswamy's True Value column dissects how software companies massage the net revenue retention (NRR) metric to look healthier in the AI era, obscuring real churn and growth quality. A useful financial-literacy explainer for evaluating SaaS firms, with lasting reference value on how to read disclosures critically. Substantive analysis beyond a news beat.
Efrati marshals The Information's Generative AI Database to rebut the viral MIT '95% of AI projects fail' study, showing 18 AI-native startups now generate $18.5B annualized revenue (88% OpenAI/Anthropic) plus a 'shadow AI economy' the study ignored, and citing concrete enterprise ROI (Intuit, Walmart, T-Mobile). A reference-grade, data-driven framework for the AI-ROI debate that reframes a widely-cited skeptical study.
Pinecone, the best-funded vector-database startup, is exploring a sale as model makers and AWS bake vector search into their own stacks, eroding the standalone value of dedicated vector DB tech. A pointed signal that infrastructure layers built around early RAG architectures are being commoditized by the platforms.
Data-point analysis: after an M&A slump, tech bankers are riding a deal boom as established software companies-afraid of being made irrelevant by AI-snap up AI startups. Frames the 'existential AI crisis' for incumbent SaaS as a driver of consolidation; useful read on the deal cycle, though the captured body is mostly paywall stub.
Applied AI column arguing that Marc Benioff's defensive, victory-lap earnings-call rhetoric (and a weak forecast plus half of Agentforce users not yet paying) inadvertently fed the narrative that legacy SaaS faces an existential AI crisis, even though Salesforce's data lock-in makes it hard to displace. A substantive, original-analysis read on the enterprise-software-vs-AI debate.
Applied AI column (Aaron Holmes) on Census Bureau data showing corporate AI use falling since June-most sharply at large firms (15% to 11%)-as ROI patience runs out, read alongside Salesforce's earnings call and MIT's enterprise-GenAI paper. Counterpoints from Databricks ($4B run-rate, +50%) and notes on OpenAI's consulting push and 'Critterz' AI-film bet. A well-sourced, data-grounded explainer on the AI-ROI reality check.
An Applied AI column on the enterprise-search fight: Atlassian CEO Mike Cannon-Brookes claims its Rovo product out-uses Glean thanks to Atlassian's long-built knowledge graph and breadth of apps, while downplaying the API rate limits Atlassian imposed on data-pulling rivals. A second item details a startup (Zulip) unable to get Microsoft's Teams data-export tool to work, illustrating the 'AI data wars' and lock-in as the contested moat in enterprise software.
Exclusive feature on how Salesforce's reality fell short of CEO Marc Benioff's confident claims that building production AI agents would be easy-the team found getting Agentforce agents to work reliably for customers was far harder and costlier (OpenAI/Anthropic model expenses) than promised. A grounded case study in the gap between AI-agent marketing and engineering reality. Body paywalled but the angle has lasting value.
Editor-in-chief Jessica Lessin's pick spotlighting Kevin McLaughlin's reported story that Salesforce's Agentforce is far harder to deploy than CEO Marc Benioff publicly claimed-a technical team that flagged the complexity was reportedly disbanded, and fewer than 5% of customers pay for the product. A vivid case study of founder ego vs. AI market reality and enterprise buyers' skepticism about paying for AI.
Amir Efrati's Applied AI argues enterprise incumbents (Workday, Salesforce, ServiceNow, SAP, Google) are converging on identical AI product portfolios-Workday's $1.1B Sana acquisition being the latest-creating a confusing, commoditized 'who wears AI best' market that CIOs must wade through. A useful framework on enterprise-software competitive dynamics, plus a notable aside that some AI detectors (Pangram) now achieve sub-0.005% false-positive rates.
Applied AI explainer on a Snowflake-led consortium of a dozen-plus enterprise vendors (including Salesforce, ThoughtSpot, BlackRock) committing to a standardized format so conversational AI can query data across siloed apps at once. Frames the data-access interoperability fight - and MCP's role - that is currently throttling enterprise AI search, paired with Hemant Taneja's take on why most AI deployments fail.
An exclusive reporting that developer-security startup Snyk weighed and rejected a private-equity buyout proposal after its IPO plans stalled. A useful original scoop on the late-stage-startup exit/IPO logjam in cybersecurity, byline-heavy signaling depth.
A Sylvia Varnham O'Regan piece on whether Microsoft will resist President Trump's demand that it fire a specific executive, set against a broader pattern of tech companies capitulating to White House pressure. Matters as reporting on the escalating politicization of corporate personnel decisions and the tech-Washington power dynamic. Substantive policy/Big-Tech analysis, though the email is a headline teaser.
Salesforce, Microsoft and a new venture from AppNexus founder Brian O'Kelley are pitching AI agents to plan ad campaigns and measure performance, challenging the Google/Amazon ad-tech duopoly. The thesis is that agents which analyze vast historical campaign data could let marketers compare media effectiveness across platforms, a genuine 'land grab' moment. Matters because it reframes the digital-ad incumbency fight around agentic AI rather than inventory.
Strong Applied AI essay puncturing the '2025 is the year of agents' hype with on-the-ground evidence: Karpathy's decade-out timeline and 'it's slop' critique, plus CEOs (AODocs, Gumroad) confirming agents still need heavy human supervision-Gumroad's Lavingia notes he replaced $400K coders with one $400K agent-manager rather than truly eliminating the role. Also debunks the GPT-5 'solved Erdos problems' claim (it surfaced existing solutions). A durable, well-sourced reference on the real state of enterprise AI agents.
Applied AI analysis of Australia's antitrust suit over Microsoft bundling Copilot into Office and raising consumer prices 30% while obscuring the cheaper classic option-illustrating the regulatory peril of Microsoft's bundle-and-conquer strategy as it weighs the same move for enterprise. Second half covers Caylent's acquisition of Trek10 and how AWS consulting firms are using AI agents to automate managed services and drive enterprise cloud adoption.
Applied AI analysis showing Microsoft kept margins flat despite tens of millions of free Copilot Chat users by owning its data centers and boosting GPT-5 token throughput 30% on Azure-contrasting with AI app startups whose margins got squeezed. Second half tracks the rise of forward-deployed engineers ('AI black belts') at Snowflake, ServiceNow, Salesforce and others as a difference-maker in enterprise AI deals.
A substantive Applied AI piece on the enterprise-AI paradox: AI is steady in budgets yet disappoints on high-accuracy automation-a cybersecurity exec calls Microsoft Security Copilot 'lighting money on fire,' Bosch shelved a hallucinating support chatbot, while narrower tools (SAP ticket-routing at 95%) deliver. Frames an emerging 'AI haves and have-nots' divide based on who gets white-glove vendor support, alongside Snowflake Intelligence's GA (12,000 agents, SAP data deal).
An Applied AI column on Microsoft's Ignite launch of Agent 365, an orchestration layer to track and govern the thousands of AI agents (including rivals') a company runs, with Nadella's 'cable TV of all these AI agents' framing of GitHub/Office as the management hub. Argues that even if Microsoft doesn't sell the most agents, owning the control plane preserves subscription revenue. Pairs it with a telling reversal: Klarna's once-bullish CEO now 'nervous' about AI data-center spend and ROI.
Applied AI explainer on how Rubrik and rivals (Sierra, Decagon, Cursor, Replit) are building 'undo' and rollback tools to reverse damage from rogue AI agents that delete databases or corrupt records, addressing CIOs' top concern that managing agent risk is harder than building agents. A second segment details how PromptArmor tricked 'Claude for Excel' into exfiltrating data via a prompt-injection attack. Practical, durable enterprise-AI-risk reporting.
A skeptical close-read of Salesforce's AI numbers: Agentforce's $540M ARR implies only ~$135M of actual quarterly revenue across 9,500 paid deals, and Peers cautions that ARR is a flattering metric and the 114% combined AI growth is actually a slowdown. The column situates Salesforce as the bellwether for whether AI scrambles or threatens enterprise-software incumbents. A useful analytical corrective to the headline ARR figures.
Applied AI surfaces Microsoft EVP Rajesh Jha's striking pitch that AI agents will need their own seat-based Office 365 subscriptions, identities, mailboxes and computers-a way out of the paradox that automating jobs shrinks the human seat base. A second segment argues AWS can keep its cloud lead via trust and incumbency even if its Nova models and Trainium chips lag in the agent race. Two substantive, original-thinking enterprise-AI analyses worth reading.
Briefing dissects ServiceNow's 11.5% stock drop on its rumored $7B Armis acquisition - its fourth large deal this year - laying out the bear thesis that AI may cannibalize SaaS and that serial M&A signals organic-growth fears, then rebutting it with ServiceNow's ~20% growth and cybersecurity strategy. Sharp, balanced enterprise-software analysis. Also a pointed critique of the FTC blocking Amazon-iRobot, which ended in a Chinese acquirer.
A reported feature on how Palantir-alumni startups like Peregrine are selling 'Anduril-ized' data-analytics software to local police and public-safety departments, amid a $1.2B funding boom (a16z, Sequoia) for police tech. Substantive look at a fast-growing defense/govtech-adjacent startup category and the relationship-driven government sales playbook, opened via a Shaq anecdote.
Amir Efrati's Applied AI essay using the reaction to Salesforce's LLM-limitation comments to frame 2025's dueling AI narratives: Nvidia/cloud/chatbot revenue booming on one side; flat agent adoption, benchmark-vs-reality gaps, pushed-out AGI timelines, and a hyper-reactive headline-driven stock market on the other. Lasting reference value as a balanced, framework-level state-of-AI synthesis ('normal technology,' skill-issue, the megadeal merry-go-round) heading into 2026.
Applied AI analysis of how consumer web platforms are reacting to a future where AI agents, not humans, are the primary site visitors: Expedia (unlike ad-protective Amazon) is building its own MCP server and 'trip agent' to greet ChatGPT/Gemini agents, with Booking, DoorDash and Uber similarly embracing chatbots. A second segment explains why mixed AI licensing models (per-seat vs per-usage) create a cost minefield for enterprise buyers. Useful framework on agent-commerce strategy and AI cost governance.
An Applied AI analysis framing the next enterprise-software battleground as the 'agentic control plane' - software for managing AI agents - where incumbents (ServiceNow, Atlassian, Microsoft, Salesforce) bet their workflow/process knowledge beats OpenAI's Frontier and Anthropic in owning the dashboard. It lays out the strategic stakes with on-record executive quotes and ties in Databricks CEO Ghodsi's crash-warning, making it a lasting reference on enterprise-AI competitive dynamics.
Applied AI explains why chip-design-software maker Cadence is resisting the SaaS sell-off: workload-based (not seat-based) pricing, deep semiconductor exposure, agentic tools that expand engineer output, and complexity too high for AI to displace; opens with Amazon's internal 'Clarity' AI-usage tracking tied to promotions. A useful explainer of which software business models are AI-resilient, plus a concrete look at corporate AI-adoption surveillance.
An incisive read on how the 'SaaSpocalypse' is sharpening enterprise-software rivalries: Salesforce unveiled an 'Agentic Work Unit' metric and a 'SaaSquatch' rallying cry, Workday's Bhusri called rivals tapping its data 'parasites,' and HubSpot/Snowflake moved to meter and monetize agent access to customer data. It crystallizes the emerging fight over who controls and charges for AI-agent data access, with stark layoff anecdotes (one customer went from 15 humans to 2.5 plus 20 agents). Substantive sector analysis with framework value.
This Applied AI piece dissects Microsoft's new E7 bundle-pricier than Office 365 plus Copilot a la carte-as a bet that per-seat subscription pricing survives the AI-agent era, enabled by Microsoft owning its own cloud and thus better margins than usage-priced rivals like Cursor. A substantive contribution to the central per-seat-vs-usage AI pricing debate that ServiceNow and Salesforce are watching.
This Applied AI analysis examines Oracle's bet that it's immune to the 'Saaspocalypse' by giving away AI agent features for free, leveraging its own cheap cloud, while rivals like Salesforce charge for Agentforce. It frames the strategic and pricing tension incumbents face from superagents that access enterprise data, with skepticism that Oracle's AI value is showing up in its decelerating numbers.
An exclusive analysis showing that while Figma, Workday and HubSpot CEOs publicly downplay AI threats, their SEC filings increasingly cite AI agents as a competitive risk-with 27 software firms disclosing such risk this year versus seven a year earlier, per an AlphaSense review. The gap between executive messaging and legal disclosures is the sharp, data-grounded thesis. Useful, original framing of the SaaS-disruption fear quantified through filings.
Cohesity CIO Brian Spanswick details how his team is using Claude Code and custom AI agents to replace pricey add-on automation tools from ServiceNow and Splunk, expecting to roughly halve that spending, and argues IT hiring will shift toward business analysts over technical staff. The piece reframes the SaaS-disruption debate from outright replacement to AI capping incremental spend. A concrete, well-sourced enterprise-software case study with named cost mechanics.
A profile of Asha Sharma, new CEO of Microsoft's $24 billion gaming business, courting major and indie game publishers at GDC to revive Xbox and grow Game Pass while disavowing 'soulless AI slop.' It signals Microsoft's strategy and tone for Xbox after layoffs and margin declines in the division. Substantive personnel-and-strategy feature on Big Tech gaming.
Peers builds a sharp analytical case that software firms' profits are overstated by excessive stock-based compensation (median 13.8% of revenue vs 1.1% for all Russell 1000 firms; Snowflake at 34-41%), which drains real cash via buybacks and leaves less free cash flow than headline numbers suggest. Cites KeyBanc data and names companies trimming (ServiceNow targeting sub-10%). A reusable financial framework for reading SaaS quality, with lasting reference value.
Applied AI analysis built around an Arcade.dev ranking of which enterprise apps are most open vs. closed to external AI agents (Slack, Workday, Meta ad tools, WhatsApp most closed; GitHub and Figma most open), illuminating the strategic fight over whether SaaS vendors will let customers' agents bypass their UIs. Frames the core threat AI agents pose to enterprise-software monetization, with named vendor responses. A useful, framework-bearing explainer despite the source's own conflict of interest.
Applied AI deep-dive on how Novo Nordisk uses AI agents (built on Celonis software over Anthropic/OpenAI models) to detect trial risks, choose trial locations, and cut clinical-trial timelines by weeks-to-months, with headcount savings and a candid note that one earlier AI tool was 'scaled back' for poor ROI. A concrete, named enterprise-AI case study with real ROI detail and a useful 'stay in Excel' caution. Substantive explainer value.
A substantive Applied AI survey of the emerging 'guardian AI' category, apps from ServiceNow, Salesforce, Holistic AI, CredoAI, Wayfound and Avon AI that monitor other AI agents for misbehavior, with concrete setup mechanics, pricing models, and the paradox that guardians often run on the same models they police. A useful map of a new enterprise-software niche born from agent reliability problems.
Exclusive that Singapore's GIC is in talks to lead a ~$500M round in database startup Supabase at about a $10B valuation, double its level from the prior fall. A concrete funding scoop signaling continued investor appetite for AI-adjacent developer infrastructure.
Analysis proposing R&D-as-percent-of-revenue as a lens for sorting which enterprise-software firms can survive AI disruption, with Atlassian and Figma standing out as heavy investors. Offers a usable framework amid the indiscriminate software selloff rather than just reporting the panic. Worth reading for the metric-driven argument.
Applied AI deep dive on LinkedIn's Hiring Assistant, a surprise enterprise hit growing ~36% weekly and outpacing every prior LinkedIn launch, with AI-drafted outreach getting ~50% higher response rates than human messages. Also covers CIOs bracing for higher Claude bills under consumption pricing and Salesforce's 'Headless 360' opening its data to third-party agents. Concrete enterprise-adoption case studies.
Briefing synthesizing multiple bullish AI signals: TSMC's 40.6% Q1 growth and raised 30%+ guidance, big-tech stock rebounds, and evidence of real enterprise AI spend (Anthropic's revenue, Uber maxing its AI budget, LinkedIn agent adoption, consumption-based pricing shifts). Argues consumption pricing is a positive longevity sign and presages more AI-driven layoffs. A useful demand-side thesis despite the digest format.
Applied AI piece on Adobe moving to outcome-based pricing for its new CX Enterprise agents, charging by completed jobs (e.g., ad campaigns delivered) rather than tokens, with president Anil Chakravarthy arguing 'tokens don't equate to value' and customers overpay under usage models. A clear explainer of the next phase of AI pricing evolution (Sierra/Zendesk/Salesforce-style), with a sharp thesis on why outcome pricing favors vendors as model costs fall.
Exclusive that major insurers (Berkshire Hathaway, Chubb, Travelers) have won state-regulator approval to exclude AI-related damages from corporate liability policies, meaning claims from misbehaving AI agents (e.g., copyright misuse in auto-generated ads) may go uncovered. A landmark, under-covered development: insurers pricing AI as an uninsurable tail risk is a structural signal that could throttle enterprise AI deployment regardless of model quality.
Applied AI dispatch from Google Cloud Next arguing the conference theme shifted from model capability to adoption help, as customers hit roadblocks on data prep, training and 'agent sprawl'; Google announced a $750M consulting fund and a revamped Gemini Enterprise Agent Platform, and Merck signed a $1B deployment deal. A well-grounded explainer on the widening gap between AI capability and enterprise implementation capacity.
Reports that dozens of enterprise software firms are abandoning flat per-seat pricing as AI threatens the seat-based model, with 79 of 500 tracked companies (HubSpot, Adobe, Salesforce, now Atlassian) adding usage-based AI fees by end-2025, more than double the 2024 figure. Matters as quantified evidence of the structural pricing shift reshaping SaaS economics in the AI era.
Aaron Holmes' Applied AI analyzes Microsoft's pivot to consumption-based pricing - GitHub Copilot usage dragged cloud gross margins down ~5 points, so Nadella declared every per-user business will become 'per-user and usage' - and warns CIOs of the 'false sense of security' such hybrid pricing creates. Also debunks AWS's 'stateful AI' exclusivity claim, noting rivals already offer it. A meaty explainer on the AI pricing shift squeezing enterprises.
Laura Bratton's Briefing makes a substantive case that Palantir is better positioned than Salesforce/ServiceNow/SAP against the AI-driven software selloff because it long ago sold data-consolidation software and pioneered outcome-based pricing (charging on milestones/margin gains) plus forward-deployed engineers. Includes a useful earnings-week preview table for Palantir, AMD, Uber, Arm, CoreWeave and others.
ServiceNow unveiled 'Action Fabric,' a metered layer that charges customers when outside AI agents access data in its apps, joining HubSpot, Workday, DataDog and SAP in erecting software 'tollgates.' The Applied AI piece frames a real industry debate: whether taxing third-party agents protects incumbent SaaS revenue or backfires as usage-based AI pricing already strains enterprise budgets.
An Applied AI analysis of Microsoft winding down redundant Copilots (Gaming Copilot, Windows entry points) after customer complaints they were 'functionally useless,' as Nadella refocuses on high-value Copilots (Office 365 Copilot grew 33%, now partly Anthropic-powered). Matters because culling product bloat also eases margin pressure from costly-to-run AI features.
An Applied AI piece on the industry debate over outcome-based AI pricing: a survey finds 31% of enterprise software firms expect to charge primarily by AI task outcomes by 2029 (up from 5%), with HubSpot, Adobe and Salesforce adopting it, while ServiceNow's COO argues outcomes are unmeasurable and create blame disputes. It matters because AI's high run-costs are squeezing the per-seat SaaS model and pricing is the unresolved battleground.
Applied AI report on how enterprise customers (National Life Group, Conga) are flipping the power balance with software vendors-negotiating opt-out clauses, shorter one-year contracts, and written AI-feature commitments so they can cut and run if vendors fall short on AI. Also covers Intuit shifting to consumption-based AI pricing. A concrete, structurally important read on changing enterprise-software contract dynamics.
Earnings-preview Briefing reframing the 'is AI killing enterprise software' debate ahead of Salesforce, Snowflake, and Asana results. Its sharpest insight: new AI revenue (e.g. Workday's $500M annualized from agents) isn't lifting overall growth, implying AI money is cannibalizing existing revenue, while customers shift to shorter contracts and opt-out clauses. A genuinely analytical setup rather than a pure roundup.
Analysis contrasting Snowflake's 34% sales-metric growth (no legacy app baggage, AI-on-databases as a clean business) against Salesforce, which must drag a huge legacy app base into the AI age and saw AI revenue fail to move overall growth. The Briefing frames who survives the SaaSpocalypse and flags Salesforce's vagueness on Headless 360 monetization and a cash-flow guidance cut-a sharper-than-usual nightly column with a clear thesis.
Applied AI explainer on how Lowe's uses semantic layers and knowledge graphs to give its OpenAI-built agents shared business context, improving accuracy for order tracking, inventory coordination, and invoice verification-and potentially cutting agent compute costs. Concrete enterprise case study amid the broader vendor fight (Microsoft, Databricks, SAP) over the semantic layer, plus analysis of Salesforce splitting its subscription revenue to spotlight AI growth.
Explainer on how the Palantir-coined 'forward-deployed engineer' role has spread across the industry (Meta, Google Cloud hiring hundreds) as vendors race to get customers actually using their AI tools. A useful read on a structural shift in enterprise AI go-to-market, where deployment help is becoming as scarce and valued as model research talent.
A genuinely useful explainer cataloging the five most common tactics enterprises use to cut AI costs: model routers, prompt engineering to reduce 'thinking,' per-role usage caps, knowing when to avoid AI entirely (e.g. Novo Nordisk reverting to Excel), and pushing for discounts/escape-hatch contracts. Concrete named examples (Snowflake, UiPath, Zscaler, Mars/Gemini) give it lasting reference value as AI-spend ROI becomes a central enterprise concern.
Covers Trump's quietly-signed AI executive order, which made compliance voluntary, explicitly barred a licensing regime, and softened the pre-release model-access window from 90 to 30 days after lobbying by tech firms and David Sacks. The lede also surveys the crowded enterprise-AI sales scramble (OpenAI's NY pitch event, Microsoft's hardware) as vendors deploy 'forward-deployed engineers' to win corporate buyers.
An Applied AI piece on how Snowflake CIO Mike Blandina cut engineering headcount in January specifically to force remaining staff to adopt AI coding tools and 'prove the concept' of productivity gains, with senior engineers proving more resistant than juniors. Pairs with a useful read on Microsoft's cost-conscious enterprise AI pitch (mid-weight models, ROI dashboards) as customers grow anxious about runaway spend.
Applied AI feature detailing Snowflake's internal AI rollout: agents that prep earnings-call Q&A in minutes, flag accounts deviating from spend projections and draft sales emails, with leadership messaging that employees must use AI daily or fall behind. The thesis - dogfooding AI internally makes Snowflake better at selling it - is a useful enterprise-adoption playbook, backed by a jump to 13,600 AI-using customer accounts. ---
5 tier-5 · 61 tier-4
Silicon is the physical bottleneck of the AI boom, and these pieces map the contest to own it. Nvidia sits at the center - its cash hoard deployed as a circular-financing flywheel (CoreWeave backstops, Lambda, Groq, the Intel stake, equity in OpenAI) that entrenches it across the supply chain - but the most consequential reporting is about the cracks: Google landing Meta as a TPU customer and pitching its chips to Nvidia-focused clouds, internal Oracle data showing GPU-rental gross margins of just 16%, and a wave of challengers (Groq, Cerebras, SambaNova, Broadcom, Marvell). The cluster reads as a single question - whether the GPU monopoly that funds the whole AI economy is durable, or whether custom silicon and thin rental economics are quietly loosening Jensen Huang's grip.
An exclusive that AI chipmaker Groq slashed its revenue projections soon after sharing them with investors-from over $2B to over $500M for 2025 (per related coverage)-citing data-center capacity shortfalls, while raising at a $6B valuation. A notable AI-chip-startup reality-check scoop on the gap between projection and delivery among Nvidia challengers. Body paywalled; the projection cut is the news.
An exclusive that Nvidia-challenger Cerebras is seeking up to $1 billion in private funding, potentially delaying its planned 2025 IPO. A meaningful semiconductor-startup financing scoop signaling AI-chip challengers favoring private capital over public markets. Body paywalled; the funding move and IPO delay are the substance.
Editor's note framing the scoop that Nvidia challenger Groq cut its 2025 revenue guidance from over $2B to roughly $500M, attributed to data-center capacity shortfalls, while raising at a $6B valuation. A telling case study of how capacity constraints can gut AI chip startups' projections.
Martin Peers's Briefing argues a US government stake in Intel is less crazy than it sounds-comparing it to the 2008-09 GM/Chrysler bailouts and framing Intel's self-manufacturing as a strategic jewel amid Taiwan-invasion risk-while warning Trump's micromanagement tendencies could backfire, plus analysis of the Trade Desk's Amazon-driven slump. A genuinely analytical take on government intervention in chips that rises above a routine digest.
An investigation into robotics' open secret: high-profile Tesla and Nvidia robot demos are often remotely teleoperated by human 'puppeteers' rather than running autonomously. A worthwhile reality-check feature that punctures autonomy hype in the humanoid-robotics race.
Feature on how AI is forcing incumbent software companies into existential reinvention, anchored by Airtable CEO Howie Liu's 'paranoia'-driven pivot to an AI app-builder after fearing his company was becoming a dinosaur. Captures the broader anxiety and opportunity sweeping SaaS founders. A substantive trend piece on the incumbent's dilemma in the AI era.
Nvidia CEO Jensen Huang's backing has revived investor and industry interest in silicon-photonics startups that aim to move AI data-center interconnects from electrons to light. Matters because optical interconnect is emerging as a key bottleneck (and opportunity) for scaling AI compute. A useful semiconductor/infrastructure explainer on an under-covered niche.
An exclusive on how Nvidia anointed a startup (Vast Data) as the favored player in AI data storage, illustrating Nvidia's power to make or break vendors across the AI infrastructure stack. Insightful on the increasingly important storage layer and Nvidia's kingmaker role.
SoftBank has signaled to bankers that its aggressive AI- and chip-related dealmaking will continue, even as analysts question whether the company can afford the spending spree. A useful read on the financial sustainability of one of AI's biggest capital allocators.
Scoop: Nvidia acquired Solver (formerly Laredo Labs), a small AI-coding-agent startup, the latest in a swelling run of AI acquihires/acquisitions; also notes Kleiner Perkins making its first model-maker investment via Anthropic's round. A solid deal scoop signaling Nvidia moving up the stack into software/agents.
Exclusive: Google is shifting strategy to place its TPU AI chips inside third-party cloud providers (deal with Fluidstack, talks with Crusoe and CoreWeave) that have historically been Nvidia-centric, aiming to chip away at Nvidia's dominance and grow its TPU business. Matters as a concrete escalation of the Nvidia-vs-custom-silicon competition in AI compute.
Exclusive scoop: Lambda, the Nvidia-backed GPU cloud renting chips to AI firms, has hired Morgan Stanley, J.P. Morgan and Citi to prep a U.S. IPO as early as H1 2026 after ~60% revenue growth to >$140M in Q2 (with ~$16M loss). Frames Lambda as a leaner CoreWeave-style 'neocloud' test of public-market appetite for AI infrastructure plays.
The underlying exclusive scoop that Nvidia signed $1.5B in contracts (a four-year $1.3B deal for 10,000 chips plus a $200M deal for 8,000 more) to rent its own AI chips from cloud ally Lambda, making it Lambda's top customer ahead of a potential IPO. A concrete, original report documenting the circular AI-chip financing the later Briefing dissects.
Briefing built around a sharp Apple-iPhone analogy to dissect Nvidia's circular practice of investing in cloud startups (Lambda, CoreWeave), selling them chips, then renting the chips back, boosting both startups' pre-IPO revenue and Nvidia's own, and argues for far more disclosure of these related-party arrangements. A genuinely analytical take on AI's circular financing, plus a secondary read on Atlassian's $610M Browser Co. acquisition.
Editor's pick fronting the scoop that Google is escalating its TPU push by approaching smaller Nvidia-renting cloud providers to host its own chips, already striking a deal to place TPUs in a Fluidstack New York data center with Google acting as a $3.2B financial backstop. A meaningful read on Google directly challenging Nvidia and the diversification-from-Nvidia trend, with substantive editor framing plus the article lede.
The original exclusive scoop that OpenAI raised revenue projections but now expects to burn $115B through 2029, far above prior plans, on compute and an own-chip/data-center strategy. A landmark financial disclosure on the cost of frontier AI; this is the source scoop the later digests derive from, hence ranked above them.
Exclusive that Nvidia is pulling back from building its own cloud service to rival AWS-a notable strategic retreat that eases tension with the big cloud customers (AWS, Microsoft, Google) who are also Nvidia's largest chip buyers. Clarifies how Nvidia is choosing not to compete with its own customers. Substantive semiconductor/cloud scoop despite paywalled body.
The Information AM digest led by a scoop on deal mechanics: Nvidia's first $10B of its planned $100B OpenAI investment buys ~2% at the $500B valuation, with later tranches at future (higher) valuations to limit dilution. Strong roundup overall-also covers the Google ad-tech antitrust remedies trial, Coinbase's 'super app' bank-replacement ambitions, and Oracle's CEO succession.
Deep analysis of how Jensen Huang is deploying Nvidia's enormous cash hoard - the CoreWeave GPU backstop, $700M Nscale investment, Enfabrica acqui-hire, $5B Intel stake, and OpenAI funding - to entrench Nvidia at the center of the AI economy. A landmark map of the circular-financing strategy reshaping the entire AI supply chain.
The market-moving scoop: internal Oracle documents reveal renting out Nvidia chips earned just ~16% gross margin (vs ~70% company-wide), with Oracle losing nearly $100M on new Blackwell rentals in one quarter. Landmark reporting that exposed the thin economics of the GPU-cloud rental business and dragged AI cloud stocks down market-wide.
Martin Peers argues the industry may be gripped by a mass delusion that AI is good for business: so far only chip/server makers (Nvidia, TSMC, Dell) clearly profit while cloud renters like Oracle see margins collapse (14% on GPU rentals vs ~70% overall) and app/model builders show scant profit. A sharp framing of the AI profitability gap, paired with a Tesla Optimus-vs-Model-Y segment. More analytical than a routine digest.
Exclusive that SoftBank's Arm is involved in OpenAI's Broadcom custom-chip effort, meaning OpenAI's in-house silicon push directly benefits one of its largest shareholders. An original scoop adding a key dependency to the OpenAI-SoftBank-Arm-Broadcom web (lede partly gated behind subscription pitch in the email).
Peers tallies OpenAI's six big announcements in three weeks-three chip deals (Nvidia, AMD, Broadcom) totaling 26GW-and argues that with 800M weekly users mostly on the free tier, the real fix for the supply crunch is pricing: only what people will pay reveals true demand. A pointed economic critique of the 'biggest joint industrial project in human history' framing and its financial-system fragility.
Exclusive (lede gated behind subscription pitch) on the intensifying competition among cloud providers to rent out Nvidia GPUs, the supply-demand dynamics and thin-margin economics tied to OpenAI's massive capacity needs. A flagship multi-reporter piece central to the AI-cloud-buildout beat, even though the email body shows mostly the paywall.
Applied AI scoop that a majority of AWS's Bedrock AI service now runs on Amazon's in-house Trainium chips rather than Nvidia GPUs-a strategically significant move to lift AI cloud margins via cheaper custom silicon and commoditize the AI stack. Strong original reporting on the economics of cloud AI (paired with a lighter Bay FC AI-coaching anecdote).
Exclusive reporting that Oracle's AI data-center business (mostly Nvidia GPU rentals to OpenAI, Meta and others) ran ~16% gross margin over five quarters versus the 30-40% co-CEO Magouyrk now promises investors. A concrete, original data point at the heart of the AI-cloud-profitability debate that itself moved Oracle's valuation.
Analysis of how Nvidia, AMD and Broadcom each carry different exposure from their OpenAI chip-supply deals, given OpenAI's plan to burn $115B through 2029 and questionable ability to honor commitments-with AMD most at risk after a 40%+ stock rally on its deal. A substantive risk-mapping of the circular AI chip-financing dynamic.
Daily AM digest led by a substantive scoop: Anthropic will use up to 1 million Google TPUs (a gigawatt of capacity next year, implying tens of billions in spend), a notable vote of confidence in TPUs as a Nvidia GPU alternative and a deepening of Google's strategic stake in Anthropic. Roundup also covers Trump calling off the SF National Guard after tech-CEO calls, Globalstar's stock pop, and Synthesia's $4B GV-led round. The TPU lede has lasting strategic relevance; the rest is aggregation.
Exclusive that AI-chip startup SambaNova (last valued at $5B in 2021) is exploring a sale after failing to close a funding round, having struggled to grow sales against Nvidia. A concrete data point in the broader thesis that non-Nvidia AI-silicon startups are getting squeezed and consolidating. Original scoop with semiconductor-market relevance.
Inaugural Sunday look-ahead Briefing previewing a heavy week: Meta/Google/Microsoft/Apple/Amazon earnings, Nvidia's Washington GTC, and the Trump-Xi meeting, with consensus revenue/EPS estimates and a clear watch-list (cloud growth rates, Meta capex, Google search-ad drift to ChatGPT, iPhone 17 mix). The framework of what to monitor across the megacaps gives it reference value beyond a single day. A genuinely useful earnings-season explainer.
Exclusive scoop that RF-chip maker Skyworks held talks to acquire $8B rival Qorvo, with the consolidation driven by both companies warning of weakening sales to their biggest customer (Apple) as competitors win business. Signals stress in the Apple smartphone-chip supply chain and a possible defensive merger among RF suppliers. Single-story exclusive with M&A and supply-chain relevance.
The Information AM digest leading with the landmark restructured OpenAI-Microsoft terms: Microsoft gets a 27% stake valued at $135B and IP access through 2032 while ceding its compute right-of-first-refusal in exchange for a $250B Azure commitment, plus OpenAI's recap into a public benefit corp. Also: Nvidia's $500B in orders through 2026 and a Nokia investment, and Skyworks buying Qorvo for $10.6B.
A Briefing that goes beyond the Nvidia headline to make the load-bearing observations: ex-China revenue doubled, the $500B+ Blackwell/Rubin backlog, and the sustainability question around Nvidia's circular deals (investing in customers like Anthropic who then buy its chips). The skeptical read on vendor-financed demand plus the TPU/AMD competitive threat gives it analytic value above the AM digest. Also includes a tidy breakdown of Adobe's $1.9B Semrush buy as an SEO-business-in-decline bet.
Martin Peers analysis showing Nvidia's free cash flow exploded from $3.8B (FY2023) to an estimated $96.5B (FY ending Jan 2026), a ~194% CAGR faster than any three-year stretch by a major tech firm since 1990 (rivaled only by post-iPod Apple), raising the question of how Nvidia deploys that cash. A useful framing piece on Nvidia's unprecedented cash-generation and the strategic leverage it confers.
Nightly Briefing with an incisive thesis: Nvidia neutralizes competitive threats by investing billions into customers (OpenAI, Anthropic) to lock in chip loyalty, and now has more free cash flow (~$97B) than even Google, ironically funded by the same hyperscalers buying its chips. A clear analytical framework for why Nvidia is unlikely to lose share despite Google's TPU push.
Exclusive scoop that Marvell is in advanced talks to acquire pre-revenue Santa Clara optical-interconnect chip startup Celestial AI in a cash-and-stock deal that could top $5 billion including milestone earnouts. The deal underscores how strategically valuable photonic/data-movement chip technology has become for AI data centers, with a major acquirer paying billions for a company with no revenue.
An earnings-preview column laying out the AI businesses of Oracle and Broadcom ahead of their reports: Broadcom's AI revenue heading to ~$19.9B in FY2025 (vs $12.2B prior year) against Tan's $60-90B 2027 market forecast, and Oracle's cloud growing 50%+ toward a booked $144B by 2030. It frames the financing and customer-commitment questions (OpenAI dependence) that will drive both stocks. A genuinely useful explainer/setup with concrete numbers, above typical digest fare.
Briefing makes a sharp contrarian case that the OpenAI-Disney deal is lopsided in OpenAI's favor: Disney pays $1B for equity it can ill afford, becomes a paying OpenAI customer, and takes licensing compensation as equity warrants tied to OpenAI's future valuation rather than cash. Notes the pattern of OpenAI's splashy not-yet-finalized announcements (Nvidia precedent). Original, well-argued deal analysis; also dissects Broadcom's AI vs non-AI revenue split.
Single-article promo email for The Information's deep-dive analysis of why Nvidia paid ~$20B (roughly 3x Groq's $6.9B valuation months earlier) to license Groq's inference-chip tech and hire its founders. The framing piece behind the week's biggest semiconductor story; worth reading for the strategic rationale, though the email itself is mostly the lede plus related-article links.
The Information AM digest leading with its own scoop on the Nvidia-Groq ~$20B nonexclusive licensing-and-hiring deal, structured to dodge regulatory review (mirroring the Enfabrica playbook). Explains the strategic logic: Groq's on-chip-memory inference design reduces Nvidia's dependence on the HBM supply chain, while Groq's 75% revenue-projection cut underscores how hard it is to challenge Nvidia. Lead item is original Information reporting with real analytical substance, lifting it above a routine digest.
Briefing connecting a holiday cluster of AI deals into one argument: Meta-Manus and SoftBank-DigitalBridge aside, the standout is Nvidia's $20B 'nonexclusive license' for Groq, hiring ~90% of staff and paying out investors at that valuation, the biggest of the acqui-hire-style structures (alongside Google-Character, Amazon-Adept, Microsoft-Inflection) designed to dodge antitrust review. Sharp, pointed analysis of regulatory-arbitrage dealmaking with a clear thesis regulators should notice.
The Information AM digest reporting Meta's acquisition of Manus (Butterfly Effect), gaining the viral agent's $125M run-rate and access to Meta's user base, a big win for Benchmark/ZhenFund/HongShan/Tencent backers off a $500M round. Also covers SoftBank buying data-center investor DigitalBridge for $4B (Stargate exposure) and Nvidia completing its $5B Intel investment for a $2.9B paper profit. Dense with consequential AI-infrastructure and consolidation deals.
Year-in-review Briefing using Sandisk (up 559% as NAND demand shifts from phones to AI data centers) as the entry point for a clean map of who won 2025's AI boom: memory (Micron +240%), power gear (GE Vernova, Rolls-Royce), and software's split between surging Palantir (+135%) and slumping Salesforce/ServiceNow (-20% to -30%). Includes a full Mag-7-plus scorecard, making it a useful reference snapshot of AI-driven equity divergence.
Exclusive scoop that OpenAI is lining up up to $60B from its biggest partners-Nvidia (up to $30B), Amazon (potentially $20B+ as a new investor), and Microsoft (under $10B)-as part of a raise targeting $100B. A consequential, market-moving funding story even though the email itself is a single-scoop teaser.
Amir Efrati's Applied AI essay uses Jensen Huang's 'software is a tool' defense to lay out the central question for the SaaS selloff: if super-agents access all apps from the OS/browser layer, the value contest shifts to the agentic aggregation layer and per-seat pricing erodes. A framework piece that names winners, losers, and the timeline (agents aren't trusted yet) with lasting reference value for the AI-vs-SaaS debate.
A trend feature documenting the rise of the 'do-it-myself CEO,' as founders use Claude Code, NotebookLM and Replit to personally prototype and take over coding, design and other work once delegated to their teams. It captures a genuine management-culture shift driven by AI tooling, with named executives illustrating the productivity-and-intensity ethos.
Exclusive that OpenAI is locking in its ~$100B round at an $830B valuation, anchored by SoftBank's $30B (in three $10B installments), with Amazon up to $50B, Nvidia up to $30B, and Microsoft in the low billions. Matters as the definitive cap-table breakdown of the largest private funding round, showing how chip/cloud suppliers are doubling as equity backers.
Meta has signed a multi-year, multibillion-dollar deal to rent Google's TPUs to train AI models, with talks to buy TPUs for its data centers as soon as next year. The marquee customer win materially strengthens Google's bid to build a TPU business and directly threatens Nvidia's GPU dominance; Google is also setting up investor-funded JVs to lease TPUs more broadly. A landmark scoop that reshapes the AI-chip competitive map.
Peers casts Jack Dorsey's 40% Block layoff as a watershed moment-the first major CEO to openly blame AI for cutting nearly half his staff-and argues it will fuel the political backlash against AI and data centers even as Wall Street cheered the stock up 24%. He contrasts this with Nvidia's puzzling 5.5% post-earnings drop and Netflix's exit from the WBD bidding. A sharp, thesis-driven column on the AI-jobs reckoning with broader resonance.
Nvidia-backed UK cloud upstart Nscale (customers include OpenAI and Microsoft) is in talks to take control of a large, permit-cleared West Virginia data center site ahead of its IPO, which would instantly make it a major US AI-infrastructure player. An exclusive scoop illustrating how the neocloud land grab for power-ready sites is reshaping the GPU-cloud landscape.
A technically rich Briefing previewing Nvidia's GTC reveal of a Groq-based server system-its first time integrating another company's processor-driven by the shift toward inference, with OpenAI expected as a buyer. It details the unusual architecture (256 Groq LPUs per rack, Intel-managed interconnects, Samsung foundry production) and a future plan to fuse Groq's LPU with the Feynman GPU. Substantive semiconductor analysis with real engineering specifics and strategic read.
Martin Peers' column scrutinizes Meta's surging third-party cloud spend-a $27B Nebius deal and contractual commitments jumping from $32.8B to $131B year over year-arguing this off-balance-sheet operating cost helps explain falling operating margins and possible 20% layoffs. A second item unpacks the confusion around Nvidia's $1T chip-revenue projection. Genuinely useful financial analysis connecting Meta's AI capacity crunch to its margin and headcount pressures.
AM roundup led by Arm's break from its neutral-IP role: it will produce its own 'Arm AGI CPU' AI server chip with Meta, OpenAI, Cloudflare, and SAP as committed buyers, a strategic move to lessen reliance on Nvidia GPUs. Also covers software stocks dropping on AWS internal-AI fears, Meta's $9T-target executive stock plan, and a $375M New Mexico child-safety verdict. The Arm pivot is a meaningful semiconductor-strategy story.
Martin Peers spotlights a striking valuation dislocation: Amazon trading at its cheapest earnings multiple since the 2008 crisis (below Walmart for the first time ever) and Nvidia at its lowest forward multiple in seven years despite ~71% projected revenue growth, framing it as a possible 'selective AI-wariness' mispricing. A genuinely analytical valuation argument with concrete comparisons, plus a note on Apple's vibe-coding crackdown.
Exclusive that Google is in talks with Marvell to develop two new chips for efficient inference: a memory processing unit to pair with its TPU, and a new inference-specific TPU. Underscores surging demand for inference silicon for autonomous agents, paralleling Nvidia's Groq-licensed language processing unit. Concrete custom-silicon strategy reporting.
Martin Peers synthesizes the 'Super Bowl' of tech earnings (Google, Microsoft, Amazon, Meta), arguing Google was the clear winner - cloud up 63%, backlog doubling to $460 billion, TPU sales to outside buyers - while Meta's rising $145 billion capex without a cloud business spooked investors and Microsoft's 40% Azure growth underwhelmed. A genuinely useful cross-company read of how AI compute demand is splitting the cloud field.
With sales exploding and servers strained, Anthropic is in talks to buy inference chips from London startup Fractile (available next year) to add a fourth chip source beyond Google, Amazon and Nvidia. The exclusive shows how Anthropic is diversifying suppliers and gaining leverage as its chip-and-server spend heads toward tens of billions annually.
Anthropic has committed to spending $200 billion with Google Cloud over five years, including Google's TPU chips alongside its cloud capacity. The scale of the deal underscores how AI labs' compute spending is reshaping cloud economics and deepens Anthropic's dependence on Google as a key supplier and investor.
Core Automation-founded in late March by ex-OpenAI researcher Jerry Tworek-is targeting a $300-500M raise at a ~$4B valuation just weeks after a $100M round at $1B, with Nvidia having invested in the earlier round. Illustrates how Nvidia's investment spree and its halo are letting brand-new model startups raise back-to-back rounds at breakneck speed.
Peers uses Cerebras' 68% IPO pop (valuing it ~$94B against ~$800M revenue) to argue the AI trade is booming and that now is the time for Anthropic and OpenAI to go public-while flagging Cerebras' heavy OpenAI dependence and comparing its multiple to Nvidia's. Sharp secondary sections cover OpenAI's odd public complaint about Apple after xAI's collusion suit, and the AT&T/Verizon/T-Mobile 'end dead zones' JV as a defensive move against Starlink and Amazon satellite plans. Good market-and-strategy analysis.
A reported feature on why embryo-editing startups intriguing Silicon Valley keep collapsing: Bootstrap Bio ran out of money because regulatory bans on edited-embryo pregnancies cap the addressable market, while Manhattan Genomics fell apart over co-founder disputes. It illustrates the gap between biotech ambition and commercial/regulatory reality. A substantive off-the-main-beat analysis with lasting context value.
Exclusive scoop that Analog Devices is in advanced talks to buy 12-year-old Empower Semiconductor for ~$1.5 billion, a deal reflecting demand for power-management chips that deliver electricity efficiently to energy-hungry AI processors. It expands ADI's AI-supporting portfolio and highlights power delivery as an emerging chokepoint in the AI hardware stack. A solid M&A scoop in the semiconductor beat.
A Briefing previewing Microsoft's Build conference as a coming-out for its homegrown AI models post-OpenAI 'uncoupling,' aiming to convince developers the models can replace Anthropic/OpenAI for simpler tasks and protect GitHub's coding franchise. A strong second section on Computex argues the AI hardware race has shifted from speed to manufacturability and supply, with networking (NVLink Fusion, co-packaged optics) becoming the next battlefield.
Exclusive that Nvidia acquired Kumo AI, a five-year-old predictive-AI software startup, for $400M+, expanding its roster of enterprise models optimized for Nvidia hardware. A signal of Nvidia moving up the stack from chips into enterprise model offerings to deepen customer lock-in.
Exclusive scoop that OpenAI is in advanced talks to lease a 10 GW Ohio data-center campus on federal land - potentially $500B+ if fully built - with possible Nvidia financial backing and a first phase in 2028. A concrete, high-stakes example of the scale and circular financing of OpenAI's compute commitments and the Nvidia-anchored buildout.
Politics feature on freshman congressman Sam Liccardo refusing an unsolicited endorsement from Leading the Future, the $125M a16z/Brockman/Lonsdale AI super PAC, because the AI industry has become politically 'radioactive' ahead of the midterms. Captures the backlash against AI money in politics as voters blame the sector for job losses and rising electricity prices. ---
5 tier-5 · 36 tier-4
China coverage doubles as the live-fire test of US export controls. The scoops here are load-bearing: DeepSeek training on smuggled Blackwell chips and pivoting some training to Huawei silicon, Nvidia's stalled H20/H200 restart, Beijing ordering its champions to halt Nvidia orders on security grounds, and a survey of China's ten-plus serious domestic chip-design firms that reframes Huawei-as-sole-threat. Alongside the semiconductor cold war runs the competitive story of Chinese AI and hardware itself - Zhipu, DeepSeek's fundraising, ByteDance vs. Alibaba in AI cloud, XPeng's EVs and robots worrying Detroit. Together they argue that the bifurcation of the global AI stack is already happening, and that the question is no longer whether China builds an independent stack but how fast.
An exclusive revealing that after Trump cleared Nvidia to resume H20 AI-chip sales to China, Nvidia has told customers (ByteDance, Alibaba) it holds only limited H20 stock and does not plan to restart production-dampening the buying frenzy. It exposes a key supply-chain and geopolitical wrinkle in the US-China chip-export saga with lasting reference value.
An analysis of how China became a major market for AI coding tool Cursor-partly because it offers Chinese developers access to U.S. AI models-and why that tailwind may not last. A useful, original angle connecting the AI-coding boom, China tech, and U.S.-model access/geopolitical risk. Body paywalled; the China-dependency thesis is the takeaway.
Chinese AI developer Zhipu AI -- which has drawn OpenAI's attention -- expects revenue to grow as much as fourfold in 2025. A notable China-tech scoop on a rising domestic foundation-model contender's commercial momentum.
Stephanie Palazzolo analyzes why OpenAI's release of open-weight models is a risky but strategic bet, its first open LLM since GPT-2 in 2019. Worth reading as a framing of OpenAI's competitive logic against DeepSeek/Meta/Llama in the open-source arena; the email carries the analytical hook though the full piece is paywalled.
Martin Peers frames Trump's 'government-as-banker' approach-a 15% cut of Nvidia/AMD China chip sales, possible downgraded-Blackwell exports, TikTok deal stakes-as a transactional new regime where deals clear if Washington gets paid, raising export-tax constitutionality and China-perception risks. A genuinely analytical column with a clear thesis on tech-policy dealmaking, above the typical Briefing.
An Information exclusive (Qianer Liu) reporting China's internet regulator has ordered local companies to freeze Nvidia chip orders on security grounds-a fresh snag for Nvidia's China sales just as U.S. export terms loosen. Significant original China-tech scoop with real consequences for Nvidia's biggest geopolitical battleground.
Scoop revealing Nvidia ordered component suppliers to suspend H20 production after Beijing told local firms to stop buying the China-tailored chip on alleged security grounds. Shows Nvidia's China foothold is in limbo from both sides-U.S. export policy lifting and Chinese government pressure now blocking demand. A consequential semiconductor/geopolitics scoop with real market impact.
HongShan, the renamed former Sequoia China unit, has deployed only about a quarter of the $8.8 billion it raised in 2022 amid a slumping Chinese VC sector, and is now hunting for deals outside China. Signals how China's tech-investing climate has cooled and how a marquee firm is pivoting its geographic focus. Useful read on the state of China/cross-border venture capital.
The Information AM leads with Nvidia's earnings: 56% revenue growth to $46.7B, a 53.8% growth projection, a $600B hyperscaler capex figure and a claimed Trump-administration demand for a 15% cut of China sales. Detailed, analytically rich coverage of the most-watched earnings of the season (plus ByteDance buyback and Snowflake items), well above a typical roundup.
A scoop that DeepSeek, a leading Chinese AI developer, is now using Huawei's chips to train some models, signaling a concrete shift away from Nvidia under Beijing's self-reliance push. A landmark datapoint on the bifurcation of the global AI chip stack and the viability of domestic Chinese alternatives.
Daily 'Information AM' digest, but an unusually high-signal one: OpenAI weighing a large India data center under an expanded Stargate; Meta's AI org discussing using Google Gemini and OpenAI models to power Meta AI (admission of in-house model struggles); OpenEvidence at $6B; US revoking Samsung/SK Hynix/Intel authorizations to use American tools in China fabs; xAI suing an ex-employee for taking secrets to OpenAI. Strong cluster of AI-infra, China-chip and Big Tech items.
Exclusive (Qianer Liu) reporting that Alibaba and Baidu have begun training AI models on their own in-house chips, partly displacing Nvidia. A meaningful signal of China's semiconductor self-sufficiency push under export controls and a potential dent in Nvidia's China demand; substantive scoop despite the paywalled email body.
An explainer laying out what's settled and what's still unknown about ByteDance's nearing sale of TikTok U.S. A useful, organized synthesis of a fast-moving and confusing deal that cuts through the noise on ownership, control, and approvals.
The Information AM digest led by reporting that Trump will sign an executive order detailing the TikTok U.S. deal-a U.S.-controlled JV with Oracle retraining/monitoring a licensed copy of ByteDance's algorithm. A strong roundup also covering H-1B visa-fee chaos, an Oracle-Meta ~$20B compute deal, and OpenAI's $450B server-rental projection through 2030.
Reports that the U.S. is losing its draw for Chinese AI talent as China's domestic industry matures and U.S. visa/immigration policy pushes researchers to consider leaving. A substantive China-tech and talent-flow story with strategic implications for the AI race, though this email is a gated single-article teaser.
A profile of XPeng's He Xiaopeng arguing the once-dismissed 'Tesla copycat' has become an unignorable force: sales on track to double to 400K+ units in 2025 driven by the $17K Mona M03, a new $30.7K luxury sedan, and a likely first profitable quarter, potentially making it a top-7 global EV seller. A lasting-reference read on the competitive threat from Chinese EV/robot makers to Detroit and Silicon Valley.
Applied AI essay reporting Airbnb CEO Chesky won't yet let users book in ChatGPT and is leaning on Alibaba's Qwen open-weight model for 'faster and cheaper' customer-service tasks-a window into how enterprises mix model providers and view OpenAI's app-integration play. Pairs with Khosla's diagnosis that enterprises fail at AI by using unqualified staff instead of forward-deployed engineers (Palantir/Distyl/Sierra model). Substantive read on real-world enterprise AI adoption patterns.
A Briefing that digs into Anthropic's report of a China-linked espionage campaign that weaponized Claude agents to autonomously attack 30 global targets at machine speed, and skewers the industry's 'our tool is dangerous but you need it for defense' logic with a gun-lobby analogy. A pointed take on the central AI-agent security dilemma that future readers will find useful. Also covers the accelerating implosion of traditional TV via Disney's earnings.
A Briefing that dissects the Meta antitrust ruling well: the judge's 'must prove the law is being violated now' standard, why TikTok alone defeated the FTC case, and the irony that Trump's own decision to keep TikTok alive handed Meta its defense. Adds a sharp Waymo-vs-Tesla section (Waymo expanding to 20+ cities, Tesla still using safety drivers) and notes the unprecedented Anthropic cap table with Microsoft, Nvidia, Amazon and Google all as shareholders.
Analysis arguing that three years after U.S. export controls, China has flipped the situation and will have more AI chips than it needs for the next five years, per Jefferies estimates, thanks to domestic chipmakers catching up plus stockpiling via open and black markets. A meaningful reframing of the export-control debate with concrete supply estimates.
Analysis of how ByteDance has emerged as a major force in Chinese cloud computing, threatening Alibaba's long-held dominance by leading in 'model-as-a-service' AI cloud offerings, prompting Alibaba CEO Eddie Wu to make selling AI models a top priority. A substantive look at the shifting competitive structure of China's AI cloud market between two giants.
Standalone exclusive: China's policymakers gathered Alibaba, ByteDance and Tencent in emergency meetings to gauge demand for Nvidia's H200, which China has no domestic alternative for, after Trump's export approval complicated Beijing's self-sufficiency goals. The piece surfaces the core dilemma between enabling AI development now and forcing adoption of homegrown chips. Substantive single-article scoop on a pivotal US-China chip-policy flashpoint (note: same story re-sent as a digest lede in #0555).
An exclusive scoop reporting that DeepSeek is training its next major model on several thousand smuggled Nvidia Blackwell chips that US export controls forbid for China, sourced to six people. The smuggling scheme routes servers to permitted countries, then dismantles them and imports the chips in pieces. Landmark reporting that directly bears on the efficacy of US export controls, China's frontier-AI capability, and the chip-smuggling enforcement debate.
Single-article scoop that Tencent, long a passive bystander in China's AI race, is now aggressively poaching ByteDance researchers (reportedly doubling salaries) and reorganizing to compete with DeepSeek, Alibaba and ByteDance on frontier models. Signals a strategic shift by the WeChat owner. Substantive China-tech reporting on the intensifying domestic model race and talent war.
Applied AI piece on Fed Chair Powell calling AI 'part of the story' behind worsening unemployment while cautioning it's not yet showing in jobless claims, and his striking admission that this AI wave 'may be different' from prior tech transitions with no policy tools to manage it. Pairs it with the 'bargain-basement AI' shift - Pinterest, Airbnb and others citing 90% cost savings from open-source/open-weight models (DeepSeek, Qwen). Substantive on both the labor-impact debate and the open-source cost-disruption trend.
Single-article scoop that Ford is halting the all-electric F-150 Lightning after weak sales, relaunching it as a 700-mile extended-range hybrid, taking a $19.5B writedown, and expanding use of CATL's licensed Chinese battery tech for grid/data-center storage - a politically charged move. Situates Ford within an industrywide EV pullback (VW, GM). Substantive reported analysis of the EV retreat and China battery-tech dependence.
Martin Peers' year-end Briefing arguing Google's 2025 turnaround (light antitrust remedy, Gemini 3, +62% stock) flowed partly from sequencing: the ChatGPT-as-search-threat perception softened Judge Mehta's remedy, after which Google grew bolder distributing Gemini. Also frames Musk and ByteDance as 2025 winners, with a sharp read on how the TikTok 'sale' lets ByteDance keep the revenue-generating parts. Substantive cross-company analysis above typical digest level.
The Information AM digest whose lede adds real reporting to the Manus deal: Meta's acquisition values the agent startup at $2B+, handing Benchmark a 4x return in eight months and seed backer ZhenFund a stake bought at a $14M valuation. Also covers Tesla pre-releasing analyst delivery consensus, and a notable China item: MiniMax ($538M) and Zhipu ($559M) launching Hong Kong IPOs to become the first publicly traded foundation-model makers, testing appetite for unprofitable AI. Strong VC and China-AI signal.
Briefing connects China telling firms to halt Nvidia H200 orders to broader geopolitical leverage-possible tit-for-tat over US moves on Venezuelan oil, plus China's potential cards on TikTok's data-arm sale and review of Meta's Manus acquisition. A genuinely useful framework for reading China's chip stance as one node in a multi-front US-China bargaining game; also flags WBD's contradictory leveraged-buyout argument.
Exclusive analysis of China's rapid-fire AI and semiconductor IPO wave-two LLM developers and four GPU designers raising $4B+ on Hong Kong and Shanghai exchanges since December, with Baidu's Kunlunxin and Tencent-backed Enflame next. A substantive look at how China is financing its domestic chip/AI buildout, with reference value on the China-tech capital markets.
Alibaba is outpacing US rivals in turning AI agents into shopping assistants, handling ~200M Qwen-app orders during Lunar New Year and spiking daily active users to 73.5M from 17M. Its structural edge is owning the full stack (in-house models, e-commerce, Alipay payments, Amap, Fliggy travel), though Taobao integration and reliable bookings remain unfinished. A useful comparative analysis of why China may leapfrog the US in agentic commerce.
ByteDance paused the global launch of its hyperrealistic Seedance 2.0 video model after copyright disputes with major Hollywood studios, following viral clips that depicted recognizable actors and franchise characters. It's a concrete signal that IP/likeness liability is now a gating constraint on frontier video-generation rollouts, even for the TikTok parent.
An 'Information AM' digest led by Jensen Huang's projection of $1 trillion in Blackwell/Rubin revenue across 2025-2027 (doubling last year's $500B guidance), with strong supporting items: OpenAI's Simo telling staff to refocus on enterprise as Anthropic surges, Alibaba consolidating all AI under a new CEO-led 'Token Hub' division, and Nvidia's Groq-based LPX server system. Unusually news-dense for a digest, with several genuinely significant items. Worth reading for the Nvidia guidance and Alibaba reorg alone.
Exclusive that DeepSeek is raising outside capital for the first time at $10B+, after previously self-funding via parent hedge fund High-Flyer and turning down top Chinese VC and tech-giant offers. The shift signals the costs of staying competitive in frontier-model development. A notable scoop on China's most prominent AI-model startup.
A landscape feature arguing the Huawei-as-sole-threat narrative misses the picture: China has 10+ companies actively designing and shipping AI chips, from state-backed institutions to startups founded by ex-Nvidia/AMD/Intel engineers. A reference-grade survey of China's domestic AI-chip ecosystem with lasting analytical value on the US-China semiconductor race.
Exclusive that Tencent and Alibaba are in talks to invest in DeepSeek at a $20B+ valuation, double the $10B target from days earlier, as the High-Flyer-owned lab raises outside capital for the first time after intense investor demand. Matters as a marker of China's two biggest tech firms consolidating around the country's flagship AI lab to hedge their own model efforts.
DeepSeek is raising its first-ever round-up to 50B yuan (~$7.35B), the largest by a Chinese AI company-with founder Liang Wenfeng writing the biggest check, prompting the lab to accelerate revenue generation and speed model releases toward industry tempo. Significant signal that the research-first Chinese lab is pivoting to commercialization and scale.
Kuaishou plans to spin off its Kling AI video-generation unit ahead of a 2027 IPO, holding pre-IPO talks at a $20B valuation to capitalize on investor appetite for AI stocks. Kling competes with ByteDance, Google and Alibaba in video models, making this a notable marker of China's AI-video commercialization race.
Exclusive that Junyang Lin, former lead researcher of Alibaba's Qwen models, is raising several hundred million dollars for a new AI lab at a ~$2B valuation-nearly unheard of for a brand-new Chinese startup-with Gaorong Ventures and HongShan in talks. Lin left Alibaba in a rare public falling-out and is credited with pushing Qwen to the front of global open-source AI. A meaningful China-tech talent-and-funding scoop.
Scoop that battery giant CATL plans to join DeepSeek's first-ever funding round (~50B yuan / $7.35B, valuing DeepSeek above $51B), as CATL aggressively pushes power equipment to AI data centers. A meaningful China-tech signal tying battery/grid players to AI capital, illustrating the cross-sector circular financing pattern inside China.
Exclusive single-article scoop revealing Amazon CEO Andy Jassy - a major Anthropic investor and vendor - was among tech leaders who lobbied senior Trump officials about security risks in Anthropic's Mythos 5 and Fable 5 models, helping trigger the export restrictions that forced Anthropic to suspend model access. Notable for showing an investor turning on its own portfolio company over national-security concerns. ---
4 tier-5 · 32 tier-4
Anthropic goes from challenger to co-protagonist across this cluster. The competitive reversal is the headline - revenue growing 5x in five months to a ~$45B run rate, overtaking OpenAI on coding monetization while projecting actual operating profit - but the more distinctive reporting is the friction. The 'Mythos' security model that makes AI the best zero-day hunter in the world is also a budget-buster; the company's unpredictable, telemetry-poor billing aggravates enterprise customers; and its principled refusal of Pentagon demands triggers a supply-chain ban from Defense Secretary Hegseth that splits the labs on military work. The arc runs through a $900B raise and a confidential IPO filing - a company trying to win the enterprise on price and capability while picking the fights its rivals avoid.
An Applied AI explainer on how Databricks and Snowflake's new AI migration services let companies extract data from legacy Oracle/SQL Server/Teradata databases and translate SQL 'dialects' so old code runs on new systems, lowering enterprise software switching costs (DoorDash is cited as a user). A second section covers Anthropic imposing usage limits on heavy Claude Code subscribers in August, framing AI-coding subsidies as a market-share land grab that will give way to price hikes. Substantive enterprise-AI analysis with a clear thesis on falling lock-in.
Anthropic's revenue run-rate is nearing $5 billion as it lines up investors for a major new funding round. A meaningful dealmaker scoop quantifying the rapid commercial scaling of OpenAI's chief rival.
A feature on how Anthropic's unusual behavioral interviews and culture-driven hiring have helped it retain AI talent better than rivals amid an industry-wide talent war. Useful window into the non-financial levers labs use to hold onto scarce researchers.
Full Applied AI essay framing the central margin debate for AI app companies: Replit CEO Amjad Masad bets model prices (esp. Anthropic's coding-dominant models) will never fall and plans to pass costs to enterprises, directly contradicting a16z's 'prices will come down' thesis. Pairs it with the de-skilling finding (a Lancet study showing colonoscopy doctors got worse at detection after AI use). A sharp, lasting framework on the application-layer unit-economics fight.
Analysis examining why Anthropic CEO Dario Amodei's prediction that AI would be writing 90% of code 'by now' hasn't materialized, probing the gap between AI-coding hype and real-world adoption/capability. A useful reality-check on AI coding claims, though the captured email body is largely the paywall stub.
Big Read profiling how Anthropic is deliberately taking a cooler, more distant posture toward the Trump White House while rival tech leaders court favor-a distinctive positioning bet with real policy and competitive stakes for the AI-safety-branded lab. A substantive feature on AI-industry political strategy. Body paywalled, but the framing is a lasting-interest angle.
Substantive Applied AI analysis arguing incumbents Salesforce and Microsoft are struggling to prove their AI add-ons justify premium prices, and remain dependent on frontier models from OpenAI/Anthropic rather than cheaper commodity models. Notes Anthropic research showing 77% of business users automate tasks (vs. augment), and that leaders like Jassy, Lütke, and Benioff now frame AI explicitly as headcount reduction. Good explainer on the enterprise-AI margin squeeze.
Aaron Holmes reports Microsoft's new Office Agent ('vibe working') is largely powered by Anthropic rather than OpenAI, an implicit admission OpenAI's tech couldn't reliably build PowerPoint/Excel even though Microsoft gets it free. Also covers Stripe/OpenAI's Agentic Commerce Protocol for agent payments. A substantive read on Microsoft's model diversification and the Copilot quality gap.
Aaron Holmes examines Anthropic's claim that Claude can vibe-code clones of enterprise apps like Slack, reviving CIO dreams of ditching SaaS, with Musk's 'Macrohard' xAI effort making a parallel pitch. Balanced with skepticism from AMD and startup founders that large enterprises won't rip out SAP/ServiceNow, suggesting the real disruption is at the user-interaction layer. A useful, substantive read on the SaaS-replacement debate.
A substantive Applied AI deep-dive on a concrete enterprise use case: Cribl uses Zip's procurement AI agent (built on OpenAI/Anthropic models plus licensed Vendr/Tropic pricing data) to auto-draft vendor emails haggling down small contracts, saving ~$3M/year, and raises the question of dueling buyer-vs-seller negotiation agents. A second segment covers ex-AWS exec Baskar Sridharan joining Trase Systems and the forward-deployed-engineer model for regulated-industry AI.
An Applied AI analysis on whether AI agents over data lakes (Databricks, Lakeside) can finally displace Salesforce/ServiceNow/Workday, with concrete customer examples including one firm that left Workday for a bespoke AI model. It usefully tempers the hype: replacement requires the security, data-management and workflow-integration heavy lifting that underpins enterprise apps. Includes a separate vignette on General Catalyst/Anthropic/Percepta embedding consultants in Maryland government.
AM digest led by a major funding scoop: Anthropic plans to raise $10B at a $350B valuation (nearly double its September round), led by GIC and Coatue, against revenue guidance of ~$15B this year rising toward $70B by 2028 and a $60B three-year compute bill that may force a 2026 IPO. Loaded with reference-grade financials, plus Google passing Apple as #2 by market cap and the Musk-OpenAI suit heading to a spring jury trial.
Applied AI essay on the Microsoft-Anthropic frenemy dynamic: Microsoft now spends $500M+/year on Anthropic models to power Copilot even as Anthropic's new white-collar-automation product Cowork competes head-on with Copilot for the same enterprise customers. Uses the Cowork-vs-Copilot collision (and Microsoft's earlier OpenAI/Bing history) to dissect the coopetition shaping enterprise AI. Strong analytical framing of how alliances and rivalry coexist among AI leaders.
Exclusive scoop that Anthropic projected a 40% 2025 gross margin-ten points below earlier expectations-because inference costs to run Claude on Google and Amazon servers ran 23% higher than forecast, even as revenue surged. The data point matters as a concrete read on frontier-lab unit economics and how unpredictable inference cost makes margins for AI sellers and their enterprise buyers. Single-story exclusive with lasting reference value on AI economics.
Amir Efrati's Applied AI essay argues the 'AI kills SaaS' thesis is overstated: JPMorgan plans to keep and increase spend on legacy enterprise apps, and even Anthropic runs on Workday/Salesforce/NetSuite and partners with Intercom. A well-evidenced contrarian take on the build-vs-buy debate that's useful reference value for the SaaS-disruption narrative.
A hands-on review of Anthropic's Cowork, which extends Claude Code's agentic abilities to non-programming tasks. The reporter completed a long-deferred chore and automated a weekly task, but found the experience bumpy and concludes the product won't reach a general audience until the learning curve smooths out. Useful first-person assessment of where agentic consumer AI actually stands.
Applied AI case study of IDX Advisors, a ~10-person asset manager that replaced Morningstar/FactSet with Claude Code-built in-house tools ($150K saved) and cut legal billable hours 70-80% with agents ($1M saved), plus Cloudflare positioning itself as the network agents run on. A concrete, quantified look at how small firms are displacing incumbent software with AI coding tools.
AM digest leading with Amodei publicly refusing the Pentagon's demands ('we cannot in good conscience accede') ahead of the Friday deadline. High-signal companion items: Jack Dorsey's Block cutting 40% of staff (to under 6,000) on an AI bet, Netflix dropping out of the WBD bidding, and an exclusive that daily Claude chatbot signups tripled since November. Several substantive, lasting data points across the roundup.
Exclusive that Anthropic is negotiating a joint venture with a PE consortium (Blackstone, Hellman & Friedman) to sell Claude plus hands-on integration consulting into the firms' portfolio companies, Palantir-style. Notable as a new go-to-market model for AI labs and an example of how the DoD conflict is rippling into commercial deals.
The original Information scoop that Anthropic executives have discussed a Q4 IPO that bankers expect to raise $60B+, which would be the second-biggest deal after SpaceX. A genuine exclusive with market-moving specifics, though delivered as a teaser email pointing to the full article.
AM roundup with a substantive lead: Anthropic eyes a Q4 IPO that bankers think could raise $60B, while prepping flagship model Claude Mythos (and a larger 'Capybara' tier) said to carry 'unprecedented cybersecurity risks.' Also covers Anthropic's court injunction against the Pentagon blacklisting, OpenAI passing $100M annualized ad revenue, and Microsoft's hiring freeze. Several genuinely consequential AI-business items packed into one digest.
A substantive Applied AI essay on how Anthropic's forthcoming 'Mythos' model, now the best cybersecurity researcher in the world per Wiz's CTO, lets AI find and exploit zero-days faster than defenders can respond, with vendors fighting AI with 'ungated' versions of the same models. It crystallizes the offense/defense inflection in AI cybersecurity with concrete demonstrations (Claude Code finding a Ghost zero-day in hours) and named expert sourcing.
Exclusive that Anthropic acquired AI-biotech startup Coefficient Bio for roughly $400M to seed its healthcare/life-sciences group, building tools for drug discovery, clinical regulatory strategy and R&D planning. A notable signal of Anthropic's vertical expansion into applied AI for biotech.
After an 8-year, 48% decline in new App Store listings, new-app submissions jumped 30% in 2025 and 84% year-over-year in Q1 2026, which Aaron Tilley ties to the rise of AI vibe-coding tools like Anthropic's Claude Code and OpenAI's Codex. A data-grounded read on how AI coding tools are concretely reshaping software supply, with a useful quantified trend.
Applied AI deep-dive on GitHub straining under AI-agent load: commits jumped from 1B/year to 275M/week (~14x), AI-agent pull requests rose from 4M to 17M, and Claude Code's public commits grew ~25x in six months, per COO Kyle Daigle-driving outages and an Azure migration. Notes GitHub captures little of the third-party-agent traffic in revenue and faces new rivals (ex-CEO Dohmke's startup, OpenAI's internal GitHub). Concrete, quantified window into agentic coding's infrastructure impact.
Applied AI deep-dive on research from Sequoia-backed Buzz showing an agent built from off-the-shelf Anthropic/OpenAI/Google models autonomously exploited 103 of 122 known vulnerabilities-most in under an hour (React2Shell in 22 minutes)-proving the defender-patch window is collapsing even without Anthropic's Mythos. A bonus section details ServiceNow's fee-based 'Context Engine' for the customer-data wars. Landmark, well-evidenced reporting on the offense-defense asymmetry in AI cybersecurity.
Profile of Claude Code head Boris Cherny, who nearly quit Anthropic for Cursor before returning to build the coding agent into a megahit-revenue rose from $1B (Dec) to $2.5B (Feb) and helped lift Anthropic to $30B annualized. Connects one product and one person to the broader software-stock decline and reshaping of engineering work. A substantive, lasting backstory on a pivotal AI product.
Applied AI interview where Uber CTO Praveen Neppalli Naga says surging Claude Code use maxed out the company's full-year AI budget within months, with ~11% of live backend code now written by AI agents (up from a fraction of a percent in three months). He envisions a shift to 'agent software engineering' with supervisor agents checking other agents. A vivid enterprise data point on AI-coding ROI and cost dynamics.
Exclusive that Google assembled a strike team of researchers/engineers to improve its AI coding models, prompted partly by Anthropic releases, with DeepMind viewing Anthropic's coding tools as exceeding Gemini's. Aim is to automate more of Google's own coding and ultimately its AI research. Signals competitive pressure in the coding-model race and the self-improving-AI thesis.
Exclusive reporting that Google signed a deal letting the Pentagon use its AI models for classified work under broad 'any lawful government purpose' language, over objections from 600+ employees. Matters as a marker of Big Tech's re-entry into military contracting and the recurring tension between defense revenue and AI-ethics commitments after Google's earlier Project Maven retreat.
Laura Bratton's Applied AI details AWS launching 'Amazon Quick,' a $20/month desktop superagent (no AWS account required) competing with Claude Cowork, framing the crowded race to build digital co-workers that operate enterprise apps autonomously. Also covers IBM's coding tool 'Bob' (model-routing for cost) and Salesforce's move to outcome-based AI pricing via its Agentic Work Unit metric. A useful survey of the superagent and AI-pricing landscape.
Original enterprise reporting on why Anthropic's costs are hard to control: it withholds the granular per-user telemetry data and SLAs that incumbents like ServiceNow, SAP and Microsoft provide, leaving customers (ServiceNow blew through its full-year budget) unable to police 'tokenmaxxing.' A second segment details Google, OpenAI, Anthropic and others all hiring Palantir-style forward-deployed engineers, underscoring how hard AI adoption remains. Substantive, decision-relevant analysis of the AI enterprise economics squeeze.
Applied AI piece marshaling evidence that AI hype is outrunning reality: Uber's COO says no clear productivity gain from AI coding tools while token costs balloon, and case studies of failed deployments at Pizza Hut (Dragontail delivery system, $100M lawsuit), Starbucks (scrapped inventory tool), and Salesforce's walked-back Agentforce claims. A grounded, skeptic's-side enterprise-AI-ROI explainer with concrete examples.
Single-article deep dive showing Anthropic's Claude Mythos found 2+ dozen critical vulnerabilities in Palo Alto Networks' code in three weeks-about 5x existing tools-but burned through $1M+ in tokens 'very quickly.' Crisply captures the capability-vs-cost tradeoff defining frontier AI deployment in cybersecurity.
Applied AI feature on Palantir CEO Alex Karp seizing the moment as enterprises (Uber, ServiceNow, Snowflake) recoil at runaway Claude Code token costs, pitching intermediaries and forward-deployed engineers as the way to control AI spend and avoid vendor lock-in. Frames the emerging software-vs-foundation-model battle and the FDE hiring boom, with the wrinkle that Anthropic's revenue still dwarfs Palantir's. A useful read on enterprise AI-cost backlash.
Exclusive scoop on how Anthropic blindsided design partners Figma and Canva: weeks before launching Claude Design, it invited them as showcase 'partners,' then quietly reworked the product to compete directly with them - prompting Figma to drop out and CPO Mike Krieger to leave Figma's board. A concrete case study in foundation-model labs cannibalizing their own customers. ---
2 tier-5 · 44 tier-4
Musk's interlocking empire gets treated as one financial organism, and the reporting's value is in pricing the parts the hype obscures. The anchor scoops are the SpaceX-xAI megamerger - rammed through by Musk as controlling shareholder of both to give cash-burning xAI a lifeline off SpaceX's planned IPO - and a rigorous sum-of-the-parts valuation that puts SpaceX near $678B against the $1.75T bankers want, the gap being 'just because it's Elon.' Around them sit the SpaceX IPO machinery, xAI's fast-but-costly data-center build-out turning it into a cloud firm, and the unsentimental Tesla coverage: Optimus's unsolved hands, the robotaxi gap between promise and reality. The cluster is a running audit of where Musk's narrative outruns the numbers.
Theo Wayt's Briefing dissects Tesla's accelerating revenue decline (-12%) against Musk's flippant earnings-call demeanor and his claim robotaxis will reach half of Americans by year-end, arguing Tesla risks being the Lyft to Waymo's Uber. Contrasted with Google 'firing on all cylinders' yet facing the $85B-capex ROI question, plus Trump's mused Nvidia breakup; a sharp earnings-week analysis.
The full exclusive (paywalled teaser here) revealing Tesla's Optimus program is running well behind Musk's 5,000-robot-2025 goal, hampered by hand-actuator problems and wrapped in unusual internal secrecy. Substantive investigative reporting on a humanoid-robotics bet central to Tesla's bull case; this is the standalone article of the story flagged in 1155.
Editor's-pick framing of a deep investigative scoop revealing Tesla is far behind Musk's goal of 5,000 Optimus robots this year (only hundreds made), with the robotic hand a key technical blocker and the program hidden from internal org charts. Substantive reporting on a closely watched robotics bet, though this email is the curated teaser version of the full piece (see 1165).
An explainer on the specific engineering challenges of building human-like hands for Tesla's Optimus humanoid robots-dexterity, actuation, and sensing being among the hardest unsolved problems in robotics. A focused technical deep-dive with lasting explainer value on a bottleneck for the whole humanoid-robot field, not just Tesla. Body paywalled; the framing of hands as the key constraint is the thesis.
An exclusive showing xAI has hired dozens of technical staff from Google-19 just in 2025-revealing the company's deep talent pipeline from the search giant. Concrete data point on how Musk's AI lab is staffing up in the talent war and its reliance on Google alumni.
An exclusive profile of Alan Clarke, the Tesla veteran Ford poached three years ago to build a new low-cost EV line meant to compete with Tesla and Chinese automakers, with Clarke arguing the effort is on track. Substantive look at a legacy automaker's strategy to survive the EV transition against cheaper rivals.
A profile of how former oilman Brent Mayo became the public face of xAI's race to build AI supercomputers like Colossus. A substantive narrative feature illuminating the people and energy/permitting politics behind xAI's data-center buildout.
A deep dive on how Musk's intense focus on the Optimus humanoid robot is running into hard engineering hurdles, notably replicating the human hand, even as he sidelined Tesla's affordable mass-market EV. Matters for gauging the gap between Tesla's robotics hype and reality, and the strategic tradeoff against its core auto business.
Stoke Space, a startup developing fully reusable rockets, is in talks to raise new funding at a valuation near $2 billion. Signals continued investor appetite for SpaceX challengers in the reusable-launch market.
Editor's-pick framing of Theo Wayt's investigation into Tesla's Optimus program, which abandoned plans to build thousands of robots this year due to persistent technical problems, especially with the robots' hands. Matters because Musk's pay package and ~80% of Tesla's projected future value ride on Optimus, yet the current bot is indoor-only and would need a full thermal redesign for the promised Mars role. Substantive behind-the-scenes reporting on a trillion-dollar bet hitting prosaic engineering walls.
Editor's-pick promotion of a reported feature: SpaceX has tweaked its next-gen Starlink satellite design to support the same spectrum Apple's iPhone satellite features use (currently via Globalstar), while Globalstar's chair has floated selling for $10B+, suggesting an Apple-SpaceX satellite deal may finally be back on. Substantive analysis of a strategically important connectivity shift, though wrapped in heavy subscribe-CTA framing.
An Applied AI essay contrasting Xpeng's CEO He Xiaopeng with Elon Musk on 'physical AI': Xpeng's Iron humanoid, robotaxis, in-house Turing chip, and flying-car roadmap mirror Tesla's ambitions but with sober, near-term framing (humanoids as receptionists/guides, not factory workers). It matters as a grounded read on how a fast-scaling Chinese EV maker is positioning as an AI company and where China's robotics/eVTOL timelines realistically land.
Analysis piece on the sudden tech-elite chorus (Musk, Bezos, Schmidt, Huang) hyping orbital data centers, including Google's 2027 prototype-satellite plan and Starcloud training an LLM in space. Argues the enthusiasm partly reflects a pessimistic belief that US compute demand can't be met on the ground - while noting Musk's own ulterior motives. Substantive thesis-driven explainer connecting AI-compute scarcity to the space-DC trend.
A reported Briefing essay using xAI's sponsorship of a Memphis Christmas light show as a lens on the local political backlash to gas-powered data centers (county-mayor and congressional races turning on it). The original argument: AI CEOs' sudden interest in space-based data centers makes sense less as engineering than as a hedge against Earth-bound political opposition. A genuine analytical take rather than a news roundup.
A substantive year-end Briefing essay cataloging Musk's missed 2025 deadlines across Robotaxi (still only Austin/SF with safety drivers vs Waymo's five driverless cities), Optimus production (5,000 goal slashed to 2,000), Grok 5 (slipped to Q1 2026), and X Money (no NY regulatory approval). The thesis: the gap between promises and reality is widening just as Tesla/SpaceX/xAI valuations have ballooned, raising the stakes. More analytical than a typical Briefing.
Applied AI piece built on a departing xAI engineer's podcast account of 'human emulators'-AI employees placed on the org chart and treated as staff, sometimes confusing real workers, with plans to scale to a million instances (partly via idle Tesla compute) under Musk's enterprise-software effort 'Macrohard.' Vivid, concrete reporting on the messy reality of automating white-collar work, plus a sidebar on Gemini API demand doubling. Worth reading for an unusually candid look inside an AI lab's automation ambitions.
Peers delivers a skeptical analysis of the SpaceX-xAI merger, arguing it's financially motivated (giving xAI access to IPO cash) rather than the space-data-center vision Musk pitches, and questioning whether even SpaceX's ~$15B revenue can fund xAI's burn. Strong sidebars on Oracle's $45-50B AI-data-center raise (and circular-financing risk) and Palantir's Nvidia-like 70% growth quarter.
The Information's own scoop on the SpaceX-xAI merger ($250B for xAI, $1T for SpaceX), reporting that Musk rammed the deal through as controlling shareholder of both to give cash-burning xAI a fundraising lifeline via SpaceX's planned $50B summer IPO. A definitive, multi-byline exclusive on one of the year's biggest deals-the financial logic behind a headline-grabbing combination.
Peers reports hands-on that Google's Gemini app can now quietly execute real transactions (ordering Uber/DoorDash) on Pixel and Samsung phones, an AI agent actually delivering, and argues Google's low-key 'ship quietly' approach contrasts favorably with OpenAI's announce-then-abandon pattern. Also flags EchoStar as a retail proxy for SpaceX exposure. A genuinely useful product-state observation plus a clean strategic contrast.
Information exclusive that SpaceX aims to confidentially file its IPO prospectus within a week or two, targeting a June listing and potentially raising $75B+ (above the prior $50B estimate) against its $1.25T valuation. A market-moving original scoop on what could be the largest US IPO ever, delivered as a teaser email.
AM roundup whose strongest item is the landmark social-media addiction verdict finding Meta and YouTube liable, a precedent for thousands of pending cases and possibly AI-chatbot suits. Also covers space stocks rallying on the SpaceX IPO-filing report, a ruling letting an expert testify that Microsoft could owe Musk $25B in the OpenAI charitable-trust case, and SpaceX bankers dropping the 'lead left' convention. Dense with precedent-setting legal and IPO news.
A single-article promo email for an analysis piece showing that space stocks (Rocket Lab, AST SpaceMobile) have surged on SpaceX IPO enthusiasm even though those same companies are most threatened if SpaceX cements its launch and satellite-internet lead. A useful contrarian framing on a momentum trade, though the email is a teaser rather than the full body.
Martin Peers's nightly column argues that SpaceX's hyped $75B+ IPO faces a sober reality check: its ~$16B revenue (up from $4.6B in 2022) implies far slower growth than Anthropic or OpenAI, making Musk's $1.25T valuation hard to justify, especially with Starlink as the core business. Frames whether the market can absorb SpaceX, Anthropic, and OpenAI IPOs in one year. A genuinely analytical valuation argument that goes beyond the day's headlines.
Peers argues OpenAI should stop issuing far-out revenue forecasts, dissecting its ad projections ($2.4B this year, $11B next, $102B by 2030) against Snap's decade-long slog to $5.2B and warning the targets set OpenAI up to disappoint. A sharp second section asks whether xAI is being quietly wound down as its executives exit and it's absorbed into SpaceX. Substantive skeptical analysis with comparative benchmarks.
Exclusive revealing SpaceX lost just under $5B on $18.5B+ revenue last year, with the loss driven by xAI (acquired February) AI spending-consolidated figures not previously reported. Hard, original financial disclosure ahead of the SpaceX IPO. High reference value for the offering's bear case.
Exclusive on previously unreported SpaceX financials showing its IPO success hinges on Starlink growth, since the rocket-launch and AI segments burn cash and don't justify Musk's sought valuation. Substantive original reporting that underpins the broader SpaceX-IPO coverage. Single-story scoop, high reference value for anyone tracking the offering.
Exclusive from SpaceX's confidential IPO prospectus: Musk bought $1.4B of stock from current/former employees via his trust to increase his pre-IPO control, plus a new plan to award him tens of millions of shares if SpaceX hits a $6.6T market cap. Original reporting from a non-public document on governance and control structure ahead of a landmark IPO.
Reporting from SpaceX's confidential IPO prospectus that its debt jumped nearly two-thirds to ~$23B at end-2025, largely from the xAI acquisition, including $4.5B tied to a related-party lease deal with Valor Equity Partners (run by SpaceX board member Antonio Gracias) for AI infrastructure. Matters as concrete prospectus-derived financial detail and a related-party-governance flag ahead of the year's biggest IPO.
Briefing arguing that SpaceX's move to buy Cursor implicitly admits xAI cannot catch up in coding, casting doubt on the $250B xAI acquisition, and dissecting Musk's $6.6T market-cap incentive grant as a device to make SpaceX IPO investors believe in a far higher valuation rather than to motivate Musk. A genuinely analytical column with a clear thesis on Musk's overlapping bets, plus an Apple-valuation-pressure section ahead of the Ternus CEO transition.
Exclusive reconstruction of why Cursor sold to SpaceX for a potential $60B: late-stage investors balked at funding it after pouring billions into OpenAI and Anthropic, with 'so much Anthropic fear' and doubts Cursor could keep up on compute. Matters as a revealing case study of how Anthropic's Claude Code dominance is starving rival coding startups of capital and forcing consolidation.
Two-year-old orbital-data-center startup Starcloud is raising at least $200 million at a ~$2.2 billion valuation, double its $1.1 billion round from a month earlier, riding investor excitement that Musk has stoked by making space data centers central to the SpaceX IPO pitch. The exclusive captures a fast-inflating funding cycle around a still-unproven LEO-compute concept that many in the industry doubt.
Ahead of SpaceX's ~$1.5 trillion-valuation mega-IPO, the largest mutual funds are debating which tech stocks to trim to free up cash for SpaceX shares, after a SpaceX-branded jet ferried nearly 200 Wall Street investors to Texas for a multiday pitch. The analysis flags a market-structure effect: a single offering big enough to pull billions out of other tech names.
SpaceX is touting xAI's speed and cost advantage in building data centers as part of its IPO pitch, but the reality is more complicated and the build-out carries hidden costs. The exclusive scrutinizes a central claim underpinning SpaceX/xAI's valuation narrative ahead of its mega-IPO.
Peers argues xAI is becoming an accidental cloud provider (renting capacity to Anthropic and Cursor) precisely because Grok isn't getting traction and xAI doesn't need all the compute it built-an awkward story ahead of SpaceX's IPO, which Musk reframes with orbital-compute hype. Sharp skeptical read on xAI's positioning, plus a segment on ESPN's underwhelming streaming switch.
AM digest led by Anthropic agreeing to take 100% of xAI's Colossus 1 Memphis data center (300MW, 220,000+ Nvidia chips), with both eyeing future orbital compute via SpaceX. Strong companion item: Dario Amodei saying an '80x' jump in Q1 revenue/usage blindsided the company (annualized revenue $9B to $30B), explaining its compute scramble; plus Uber maxing out its AI budget.
Exclusive that BlackRock has discussed a $5-10 billion order in SpaceX's IPO (which could raise up to $75B), a vote of confidence from the world's largest asset manager and a reflection of the Musk-Larry Fink relationship. It signals growing Wall Street willingness to back the largest IPO ever despite a stratospheric valuation and near-zero investor governance rights. A meaningful scoop on the demand side of the SpaceX offering.
A week-ahead preview anchored by Nvidia's fiscal-Q1 earnings (growth expected to accelerate to ~80% amid a GPU compute crunch), the SpaceX Starship test launch and IPO filing, and Meta's planned 8,000-person layoffs. Peers argues Starship matters most because most of SpaceX's business plans depend on it, and notes Starship is five years behind schedule. Also covers Ackman building a $2B Microsoft stake while Gates exits. Useful setup-the-week analysis with concrete Nvidia growth numbers.
A chart-driven explainer knitting together SpaceX's disconnected segments-rocket launches, Starlink, social media, AI models, data centers, defense, Mars-ahead of the largest IPO ever, surfacing huge losses alongside impressive growth. The visual breakdown helps investors parse a sprawling and contradictory financial story. Worthwhile reference framing for understanding the SpaceX numbers.
SpaceX's draft prospectus foregrounds the 400-foot Starship rocket as the single most important piece of its investor pitch-and names its delay or failure as the company's top risk factor. Full reusability and unprecedented mass-to-orbit underpin SpaceX's future Starlink and data-center ambitions. The piece frames the high-stakes coupling of a critical test launch with the largest IPO ever.
SpaceX's IPO prospectus discloses that Anthropic agreed to pay $1.25 billion/month to rent capacity from SpaceX's two Memphis data centers (Colossus and Colossus II), a deal worth ~$40 billion through May 2029 and one of the largest AI infrastructure deals on record. The arrangement would prop up SpaceX's suddenly slowing revenue growth but comes with risk caveats. It links two of the most-watched private companies and shows how AI compute demand is reshaping SpaceX's business model.
Briefing tying together Nvidia's blowout quarter (85% growth, 95% forecast, $48.6B FCF), the SpaceX and OpenAI rush to IPO, and a contrarian warning that long-term bond yields at 19-year highs threaten the AI-financing euphoria. Also dissects SpaceX's science-fiction prospectus ($28.5T TAM, Mars/asteroid disclaimers) and the Vox Media breakup as a digital-media-decay marker-a wide-ranging column with several genuinely analytical threads.
A standout sum-of-the-parts valuation of SpaceX off its newly public prospectus, pricing Starlink (~$150B), launch (~$300B), X/advertising (~$3B), AI cloud (~$165B), and Cursor segment by segment to reach roughly $678B-far below the $1.75T bankers seek, with the gap being 'just because it's Elon.' A rigorous, original analytical framework with lasting reference value for assessing the SpaceX IPO.
Ahead of SpaceX's mid-June IPO, this Briefing argues the near-$2 trillion valuation implies paying 'for 2030 economics at 2026 prices'-107x trailing revenue, justifiable only if launch, Starlink, and xAI businesses all mature, which won't happen soon. It frames the offering as fine for risk-tolerant young traders but dangerous for retirement-aged investors given expected post-IPO volatility from the unusually high retail share allocation. A useful valuation-discipline lens plus a second item on memory chipmakers (Micron, Samsung, SK Hynix) propping up Anthropic.
AM digest led by Anthropic's $65B raise at a $900B pre-money valuation (bringing total funding past $130B, with memory-makers Micron, Samsung and SK Hynix joining), paired with the launch of flagship Claude Opus 4.8 and a $47B annualized revenue figure. Also carries the Blue Origin New Glenn explosion, Illinois' landmark AI safety bill mandating third-party audits, and Dell's ~40% share jump on AI server sales.
Theo Wayt explains why SpaceX is offering condolences rather than gloating over Blue Origin's New Glenn launchpad explosion: while it cements SpaceX's lead (and relieves Starlink competitive pressure from Amazon's Leo and AST), any space mishap reminds investors of sector-wide risk weeks before SpaceX's IPO, and SpaceX has been in the same position before. A smart read on competitive dynamics and investor psychology in the space sector.
Martin Peers picks apart the implausibly aggressive 2030 revenue projections from SpaceX (Goldman's $474B forecast, two-thirds from AI) and OpenAI ($284B), noting SpaceX's 'AI' revenue today is essentially X ad sales and citing Musk's dismal record of meeting forecasts. A sharp, skeptical framing of how IPO-era hype maps to credible numbers, plus the Nadella 'addiction' rebuke. ---
2 tier-5 · 29 tier-4
A grab-bag by sector but a coherent one by stance - this is where The Information watches money and attention chase the AI story sideways. The standout is the macro argument that the AI bubble is a Wall Street phenomenon, not a Silicon Valley one: tight credit spreads, $3T of private debt and PE dry powder, and crypto-treasury plays all pumping air into AI's cash needs. Around it: the stablecoin price wars and crypto's leverage-driven $700B rout, the prediction-market boom (Polymarket, Kalshi) and its regulatory fights, the media consolidation chess (Paramount, Warner, Netflix, Disney's one-sided OpenAI deal), and the creator-economy edges where generative video cuts game-development costs 99% and AI music startups tangle with record labels.
An exclusive on the fintech and crypto industries lobbying the Trump administration to block banks from charging fees for access to customer data-a fight over open-banking economics with real stakes for data-access business models. Substantive policy scoop, though the email body is paywall with the reporting behind the link.
An exclusive profile of how Roku CEO Anthony Wood is reversing course on home-screen ad policies as a sharp slowdown in the streaming-TV market squeezes the device pioneer. Substantive business-strategy reporting on a maturing streaming player, though the full analysis is behind the paywall.
Martin Peers' Briefing column analyzes the FCC's approval of Skydance's $8B Paramount takeover, framing it as Larry Ellison bankrolling son David's media adventure and tracing the shifting control/voting structure in FCC filings. A sharp media-M&A read with a useful weekly stories recap appended, elevated above routine digests by its original argument.
An exclusive that Walmart renegotiated its deal with The Trade Desk to make it nonexclusive, putting one of the ad-tech firm's most valuable clients in play as Amazon pushes to become the web's primary ad broker. Market-moving reporting (the stock fell on the news) on the competitive squeeze in programmatic advertising.
Sara Germano analyzes David Ellison's $7.7B, seven-year UFC rights deal days after closing the Paramount-Skydance merger, signaling aggressive streaming ambitions and a sharp turn from Paramount's prior tightfistedness. The piece argues UFC alone won't be enough to make Paramount a streaming heavyweight, situating it in the broader live-sports rights arms race.
News publishers are pushing AI licensing toward usage-based payments that scale with how much of their content AI firms actually use, potentially raising what AI companies pay. A substantive media/AI policy story on the evolving economics of content licensing deals.
A 'Big Read' on Snap's deepening struggles, including discussions to raise outside money-possibly from sovereign wealth funds-or even spin off its Spectacles AR-glasses unit as it falls behind Meta, Google and Samsung, with younger users not adopting the app. Combines a fundraising scoop with a deeper diagnosis of Snap's business decline. A meatier feature than the daily news items.
An exclusive reporting that Google staff serving publishers have begun pitching ad agencies directly, positioning the ad-tech unit to operate as a stand-alone business ahead of a court ruling that could force a spinoff. Concrete evidence of how Google is hedging against an antitrust remedy, and that its ad-tech tech has fallen behind rivals.
Nasdaq is tightening scrutiny of public companies that want to remake themselves into crypto-treasury 'crypto stocks,' the latest pushback on a financial-engineering trend where firms load balance sheets with bitcoin and other tokens to juice their share prices. Matters because it signals exchange-level friction against a wave of corporate crypto adoption and the regulatory boundaries around it.
Analysis arguing that corporate-card startups Brex and Ramp received software-style valuations that misprice them, since they are fundamentally fintechs with interchange-driven economics rather than SaaS firms. A useful valuation-framing piece for VC/fintech watchers, wrapped here in a paywall promo.
Yueqi Yang reports that crypto exchange Hyperliquid has set off a stablecoin price war that could squeeze issuers' profits, with Circle among those hit. A substantive crypto-economics scoop on how exchange-driven competition pressures stablecoin margins.
Pinterest CEO Bill Ready argues in an op-ed that the judge in Google's ad-tech antitrust case should impose behavioral remedies rather than order a breakup, and outlines alternatives. A substantive named-author opinion piece on a live antitrust remedy question with industry-insider weight.
A deep Applied AI essay on how generative video/world models are slashing game-development costs-e.g., Genvid cut a 20-minute cutscene from $2M to $1,500-and how studios like Square Enix and startups like Decart and World Labs are racing toward fully AI-generated games. A concrete, well-reported framework piece with named economics and a credible prediction of a viral AI-built game within months.
Analysis catching the contradiction in Google's ad-tech antitrust defense: in court it argues the open web is in decline to justify keeping its ad business intact, while elsewhere it tells a different story. A sharp framing of the rhetorical strategy underlying the DOJ remedies fight.
Two months after the new stablecoin law passed, it is already driving fresh competition and squeezing issuer profits. Matters because the regulatory shift reshapes the economics of the stablecoin market and who can capture its margins.
Exclusive tracing crypto's ~$700B sell-off to over-leveraged borrowed money compounded by a glitch at the world's largest exchange, exposing structural fragility in the market. A substantive risk-analysis piece on hidden crypto-market plumbing (most of the email body is the subscription wall, but the thesis is clear from the lede).
Ken Brown's thesis-driven argument that the AI bubble is fundamentally a Wall Street phenomenon, not a Silicon Valley one: frothy markets, near-record-tight credit spreads, $3T in private debt chasing yield (CoreWeave debt at 8.5%), $2.5T of PE dry powder and crypto-treasury plays all pump air into AI's cash needs. A genuinely original framing with lasting reference value on the bubble's financial-system plumbing.
Briefing combining OpenAI's ChatGPT Atlas launch with a sharp media-M&A analysis: Warner Bros. Discovery is on the block for the third time in nine years with unsolicited bids from Paramount Skydance, Comcast and Netflix, and the column sums-of-the-parts the studio (~$25B+), HBO Max (~$60B+) and cable channels to argue a breakup may beat a sale. Also covers the Musk-vs-Duffy NASA feud. Genuinely useful valuation framework for the WBD situation.
Briefing using Trump's pardon of Binance's Changpeng Zhao to frame a thesis: offshore crypto giants (Binance, Tether, Polymarket, OKX) are exploiting Trump-family crypto ties for a once-in-a-lifetime path into the US market, disadvantaging compliant domestic players like Coinbase and Kalshi. A coherent argument about crypto's regulatory regime shift, with the next prize being US IPOs. Substantive crypto-policy analysis above the usual nightly fare.
Analysis explaining how Netflix sustains ~15% revenue growth-double or more its rivals Disney and Warner Bros. Discovery-despite being the largest streamer with the most subscribers, an inversion of normal scale dynamics. Offers a substantive competitive read on streaming-market structure. Worth reading as a media/streaming explainer.
Profile of Klay, a secretive AI music startup founded by Ary Attie that became the first to strike licensing deals with all three major record labels (Universal, Warner, Sony). The deals matter because the music industry has resisted AI, making Klay a notable test case for licensed, label-sanctioned AI music.
Single-article analysis showing prediction markets (Kalshi, Polymarket) are increasingly driven by sports betting-the majority in Kalshi's case-now drawing DraftKings, FanDuel, Coinbase, Robinhood and Crypto.com into the space. A useful explainer on the convergence of sportsbooks, crypto exchanges and prediction markets and the regulatory gray zone it opens.
A Briefing explaining that despite 2025's crypto wins (trust charters for Circle/Ripple, Erebor Bank, friendlier SEC and Fed), crypto and stablecoin startups still can't easily get full-service bank accounts because trust charters bar deposits and lending. The structural point: too many stablecoin firms depend on one small Kansas City bank (Lead Bank), which is now tightening controls-showing banks stay cautious even under a hands-off Trump regime. A useful explainer of the sector's persistent banking bottleneck.
Erin Woo's deep explainer on the first bellwether social-media-addiction trial (Meta, YouTube, TikTok), the novel product-liability theory that sidesteps Section 230, and why a plaintiff win could reshape internet-service design. The Section 230 framing and stakes analysis give this lasting reference value for media/policy watchers.
Standalone Martin Peers analysis explaining Reddit's 42% five-week drop: after an underpriced IPO and two years of 60%+ ad-revenue growth, the most-engaged U.S. user metric is now softening, undercutting the growth story. A useful single-company deep dive with original metric analysis and IPO-comparison context (Snap, Pinterest).
Peers analyzes the Wall Street rout in social-media stocks (Meta, Reddit, Snap down 8-10%) after a jury found Meta and YouTube liable in the social-media addiction case, stripping the Section 230 immunity internet firms have relied on. Argues the legal precedent creates lasting liability overhang extending even to AI chatbot makers, and that cash-poor Snap is most exposed. A solid analytical read on a precedent-setting verdict's market consequences.
Reporting that stablecoin issuer Tether is making a last-ditch push to raise at a $500B valuation, giving investors two weeks to commit amid skepticism that has stalled the round since late last year, with postponement likely if too few sign on. A meaningful crypto-finance scoop on whether the market will validate an extraordinary valuation.
An investment thesis arguing Mercado Libre is undervalued: shares fell 9% over 12 months despite 30%+ revenue growth, as investors overreact to Sea Ltd. competition in Brazil. The piece contends the Latin American e-commerce/fintech market has room for multiple winners and that MELI's management has balanced growth and profitability well. A standalone stock-analysis deep dive with reusable reasoning.
A profile of new CFTC chair Michael Selig, a prediction-markets enthusiast aiming to turn the agency into a power center by crafting federal rules that let betting startups (sports, elections, world events) flourish over state objections. A useful policy explainer on the federal-vs-state fight shaping the fast-growing prediction-market industry.
An exclusive that Polymarket's U.S. relaunch is faltering as rival prediction market Kalshi pulls ahead, with Polymarket's U.S. CEO conspicuously absent/'AWOL.' A substantive competitive scoop in the fast-moving prediction-markets space (truncated here behind a paywall/Pro upsell, but clearly original reporting).
A reported feature on 1155 F Street, the DC office tower that has become crypto's lobbying command center-housing the Blockchain Association, Solana Policy Institute, Hyperliquid Policy Center, Coinbase, and DCG advisers-amid the push for the Clarity Act. A colorful, original deep-dive into how the crypto industry is buying mainstream financial legitimacy in Washington. ---
1 tier-5 · 62 tier-4
If silicon is the bottleneck, the data center is the capital sink, and this cluster follows the money into the ground. The buildout has outgrown corporate balance sheets, so the real story is financial engineering: OpenAI reserving $100B of backup servers, private-credit shops (Blue Owl, BlackRock) and securitization underwriting CoreWeave, Crusoe, Nscale, Lambda and Switch, and hyperscalers spreading buildout risk through off-balance-sheet structures. Underneath runs the physical-constraint reporting - power, nuclear permitting, water, batteries, even space-based data centers - and the margin anxiety as cloud capex outruns revenue. Read as a set, it's the plumbing diagram of how a trillion-dollar compute habit gets paid for, and who is left holding the debt if demand blinks.
A profile of Crusoe, the OpenAI data-center provider building the Abilene Stargate site, and its ambitious bid to compete with hyperscale cloud giants. Substantive coverage of the AI-infrastructure buildout and a rising neocloud challenger.
Exclusive scoop that fintech Ramp is in talks to raise $350M at a ~$21B valuation just a month after a $16B round, with Iconiq in talks to lead. Notable because back-to-back step-up financings of this scale are common for hot AI startups but unusual in fintech, a signal about where investor appetite is spreading.
An exclusive that Fireworks AI-a startup renting Nvidia-chip servers to run and customize open-source models faster/cheaper than AWS or Google-is raising at a $4B valuation (7x last year's) led by Lightspeed and Index, on ~$200M revenue pace heading toward $300M. A concrete financing scoop on the fast-growing AI-inference-cloud segment with real revenue and valuation data. Body paywalled; the round is the story.
Idaho National Laboratory and Microsoft are building AI tools (using OpenAI models plus traditional semantic models) to automate nuclear reactor permitting, aiming to meet a Trump executive order to cut licensing from years to 18 months; early tests cut reactor-mesh generation from weeks to under a day and could lower permit costs over 20%. The issue also reports an agent-hacking contest where Claude proved hardest to trick and Llama easiest, plus a Microsoft paper ranking white-collar jobs most exposed to AI. Useful applied-AI explainer tying AI to data-center power buildout and agent security.
Martin Peers dissects Figma's 250% IPO pop (a 55x forward-revenue multiple on an ~8% float) as bubble-like froth spilling over from the AI boom, drawing a Snowflake-2020 parallel and warning first-day buyers. Also analyzes AWS's stalled growth vs Azure/Google and Apple's surprise capex jump. A sharp, analytical Briefing.
Martin Peers makes a contrarian case that the AI-driven shift toward cloud will compress Microsoft's and Google's overall margins (cloud is far less profitable than their software/ad businesses) while lifting Amazon's, since AWS margins dwarf e-commerce. A genuinely analytical Briefing with a non-obvious investment thesis on the cloud margin mix.
Argues that Big Tech's massive AI capex is finally being matched by accelerating cloud revenue growth, easing fears that the spending is outrunning monetization. A useful analytical frame for judging whether the AI infrastructure buildout is paying off.
Miles Kruppa reports that AI startups chasing big checks and cheap energy from Gulf states are finding those deals come with significant strings attached (geopolitical, control, and dependency conditions). A substantive analysis of the tradeoffs in Middle East AI financing; the email is the promo-stub but the thesis and beat are clear and lasting.
Martin Peers makes the strongest case in this batch that Meta is taking the biggest AI gamble of Big Tech-$100B+ annual capex with no AI cloud to monetize it, a cash balance down 40% in H1, and analyst revenue projections that would require Meta to capture all global social-ad growth through 2030 (WPP forecasts $23B less than analysts expect from Meta alone). A genuinely original financial argument with hard numbers, well above a typical roundup.
An exclusive that Base Power, the home-backup-battery startup co-founded by Zach Dell, is in talks to raise up to $1B at a $4B valuation-roughly 4x its April mark-as revenue jumps on deals with housing developers, with plans to build its own batteries and sell to data centers. A concrete energy-startup fundraising scoop at the intersection of grid storage and AI power demand.
An exclusive that cloud-hosting startup Vercel, which hosts AI apps, has been approached by investors looking to triple its valuation to $9B. A concrete fundraising data point illustrating how AI-infrastructure startups are being bid up in the current frenzy.
Steve LeVine's deep dive on the collapse of Northvolt, Europe's would-be Tesla answer that raised $13B and booked $55B in orders before declaring bankruptcy in March, and how a tiny California startup ended up acquiring its assets. A substantive postmortem on one of the largest tech failures ever and the broader battery-industry downturn.
Martin Peers argues AI investment has become so pervasive and interlinked (SoftBank-Foxconn's Lordstown EV-plant-to-AI-server conversion, Google-TeraWulf-FluidStack, PE and 401(k) exposure) that data center construction now rivals consumer spending as a driver of US GDP growth, raising systemic-risk questions if the boom reverses; a secondary section flags Starboard's renewed Salesforce activism. A genuinely useful macro/systemic-risk framing on AI's economic footprint.
Data center developer Crusoe, which builds AI infrastructure for Oracle and OpenAI, is discussing raising over $1 billion at a $10 billion valuation to finance its cloud ambitions. The scoop is a useful data point on the capital intensity and valuations of the picks-and-shovels layer underpinning the OpenAI/Stargate buildout.
Exclusive on Google landing a $10B+, six-year cloud deal with rival Meta-one of the largest in Google Cloud's 17-year history-covering compute, storage, networking and Nvidia GPU access for Meta's AI ambitions. Demonstrates Google Cloud's ability to win even its fiercest ad competitors as customers and Meta's escalating AI capex. A high-impact Big Tech scoop with lasting strategic significance.
Ken Brown's Briefing argues the data-center debt boom (Meta's $29B, Vantage's $22B, xAI, CoreWeave) is exploding-market doubling to ~$60B in 2025-as private credit chases yield, while flagging the real risks: falling AI prices, rising power costs, and CoreWeave's stock down ~50% as an early warning. A genuinely analytical lead essay on AI-infrastructure financing risk, above the typical Briefing bar.
Exclusive analysis of tech giants' emerging playbook for financing the massive AI buildout by spreading the capital risk across partners, SPVs and off-balance-sheet structures rather than carrying it alone. A useful framework piece on AI-infrastructure finance, though this email is a paywall wrapper around the lede.
Martin Peers's Briefing unpacks Oracle's 'Nvidia moment': a 359% backlog jump and a projection that cloud-server revenue hits $144B by fiscal 2030 (from $10B), driven by AI deals likely including OpenAI's $300B commitment, while capex jumps to $35B. He stress-tests how solid the projections are if AI demand weakens, plus an 'Apple's Back to Basics' essay on the AI-light iPhone event. A meaty, analytical Briefing.
Ken Brown's Briefing argues this may be remembered as the moment of peak AI euphoria: Oracle's 36% jump (on a 359% backlog increase from a likely-OpenAI $300B deal) spread froth to energy and data-center names, while the whole edifice rests on OpenAI raising tens of billions yearly to pay the bills. A sharper-than-usual Briefing with a memorable bubble thesis, plus a Bending Spoons/Vimeo aside.
Amir Efrati analysis on how OpenAI's enormous compute commitments (Oracle, Microsoft, CoreWeave and others) are reshaping cloud-market share and forcing hyperscalers to spread data-center risk. A useful framing of the OpenAI-driven capex cycle's structural effect on cloud; email is paywalled but the underlying piece is substantive market analysis.
Single-article email (Efrati and Muppidi) on OpenAI's roughly $350B computing-cost obligation, the looming gap between its compute commitments and its ability to fund them. Part of the run of deep reporting on OpenAI's cash-burn and infrastructure spending that anchors the AI-bubble question; the email body is paywalled but the scoop is a substantive 'True Value' analysis.
Exclusive funding scoop: prediction-market Polymarket has fielded an offer valuing it at $9B while rival Kalshi nears a ~$5B financing round, marking a sharp re-rating of the prediction-market sector. Concrete deal data with clear signal value on where private-market capital is flowing. Substantive startups/VC scoop despite paywalled body.
OpenAI aims to spend $450 billion renting servers over five years, including $100 billion specifically on backup/spare capacity reserved for potential AI research breakthroughs. A striking exclusive that quantifies the scale and strategic logic of OpenAI's compute spend, with lasting reference value for the AI-infrastructure debate.
A reference-style rundown of the top lenders financing the AI data-center buildout-useful 'follow the money' reporting on which institutions are bankrolling the infrastructure boom. This email is a gated teaser, but the underlying piece has reference value for tracking AI capital flows.
Standalone scoop email for the Nvidia-OpenAI chip-leasing story-Nvidia exploring leasing its GPUs to OpenAI rather than selling them outright as part of the $100B partnership. A consequential original report on how AI infrastructure gets financed, though this email is the gated teaser version of the same scoop carried in issue 0820.
The Information AM digest led by a genuine scoop: Nvidia and OpenAI may structure their megadeal around OpenAI leasing rather than buying chips, with OpenAI's total compute ambitions now implying ~$1 trillion of spending this decade. The lede story is original and consequential (a new financing model for AI infrastructure); the rest is a roundup (Stripe buyback, Microsoft publisher-pay marketplace, Stargate expansion).
Analysis arguing AI's water footprint is a major overlooked constraint, with data-center infrastructure projected to consume as much water annually as every household in California. Frames a physical-resource bottleneck on AI buildout beyond the usual power and chip discussion.
Editor's-pick repackaging of a scoop that OpenAI and Nvidia are discussing leasing chips rather than buying them, which could cut OpenAI's server costs 10-15% and shifts who bears obsolescence risk. Points to a new business model for selling expensive AI hardware to capital-constrained buyers, layered on Nvidia's $100B OpenAI funding pledge.
An analysis of Fermi America's IPO, a revenue-less data-center developer building a gas/nuclear-powered Texas site, framed as a leap of faith on debt, AI data-center demand, and Trump ties. Substantive scrutiny of speculative AI-infrastructure listings, a recurring beat worth tracking.
OpenAI's new no-code Agent Builder / AgentKit makes assembling agents easier but locks customers into OpenAI models, breaking the model-agnostic norm and pressuring rivals like Zapier, Anthropic, and Google. Also notes Deloitte's $229K refund over an AI-hallucinated government report. A substantive analysis of OpenAI's lock-in play in the agent-tooling wars.
Detailed breakdown of Oracle's investor day: it declared itself a 'hypergrowth company' projecting 53% revenue growth by FY2028 and $225B revenue / $21 EPS by 2030, with co-CEO Magouyrk forecasting AI data-center gross margins rising from a reported 16% toward 35%-but Peers flags the missing capex, segment-size and overall-margin disclosures and the $7.5B/yr projected cash burn. A genuinely analytical read of whether Oracle's AI cloud economics hold up.
Peers contrasts Oracle's aggressive, capex-heavy AI cloud buildout (stock fell 7% despite bullish 2030 projections lacking margin/capex detail) against Microsoft's deliberately judicious data-center strategy, with investors favoring Microsoft's discipline. Useful framing of the central tension in the AI-cloud buildout-financing risk and the Palantir-style government-pivot subtext behind Benioff's Trump overtures.
Daily AM digest led by a WSJ-sourced report that Nvidia is discussing guaranteeing some of OpenAI's data-center loans-part of the broader $100B Nvidia-OpenAI chip-leasing partnership-a notable escalation of the circular AI financing that exposes Nvidia to OpenAI's repayment risk. Roundup also covers NASA opening the Artemis III contract to SpaceX alternatives, Benchmark's new GP, and Sakana AI's $2.5B raise. The loan-guarantee lede has real significance for the AI-financing debate.
Daily AM digest led by a scoop that Anthropic is in early talks for a 'high tens of billions' Google cloud-compute deal (Anthropic recently raised $13B at $170B and is nearing $7B annualized revenue), underscoring the compute arms race alongside OpenAI's $300B Oracle commitment. Roundup also covers the OpenAI ChatGPT Atlas browser launch, Netflix's 17% growth, and Warner Bros. Discovery opening strategic review. Strong lede plus the Atlas item give it above-roundup value.
Briefing contrasting the three Big Tech earnings reports: Meta's spending spree squeezed free cash flow by a third (stock down 8%) while Google (accelerating search and cloud, cloud profits nearly doubled) and Microsoft (40% Azure growth) absorbed similar AI investments comfortably because their cloud capex is demand-driven. Also covers Fiserv's 42% stock collapse and a Snowflake exec's unauthorized $10B guidance gaffe.
Briefing with substantive analysis of AWS reaccelerating to 20% growth ($132B run-rate), joining Azure and Google Cloud in faster growth and easing fears it was losing ground-with careful caveats on scale-distorted growth-rate comparisons and Amazon's $125B capex burn. Second section dissects CoreWeave's failed $9B Core Scientific acquisition and its debt-heavy, vertical-integration ambitions.
Analysis using September-quarter earnings to rank the Big Tech giants by AI-spending risk: Google and Microsoft are best positioned because they throw off more cash and devote a smaller share of it to AI capex, while Meta is the worst positioned. A clean comparative framework for assessing who can sustain the AI capex arms race.
A Briefing built around a substantive framework: BlackRock's Tony Kim argues the 10 biggest tech firms generate ~$1T annual EBITDA with no net debt and could borrow trillions (1-3x net debt/EBITDA) to fund AI capex, turning cash-printing giants into disciplined, leveraged spenders. It walks through what Google's balance sheet would look like under that scenario and weighs the risk of debt-funding an uncertain-return bet-a useful lens on how AI's price tag gets financed.
'Big Read' profile of 29-year-old Zach Dell (Michael Dell's son) and his Austin battery startup Base Power, which raised $1B in October at a $4B valuation (nearly 5x its level six months earlier) and is building two battery factories to disrupt the half-trillion-dollar U.S. power industry. A substantive founder/energy-startup feature relevant to the AI-driven power-demand story.
Applied AI argues AWS's new networking interconnect to Google Cloud (and planned Azure link) is a full reversal of its long-standing single-cloud stance, driven by customers bypassing weak Nova models to reach Gemini and OpenAI-some even migrating data off AWS to use Gemini. It pairs this with re:Invent (Nova 2 Omni, Trainium3) and OpenAI's entry into AI roll-ups via a Thrive Holdings stake. Substantive analysis of AWS's competitive concession and the AI-rollup trend.
A sharp comparative analysis contrasting Anthropic's disciplined, profit-focused enterprise strategy under Amodei with OpenAI's sprawling 'YOLO' expansion under Altman (chips, data centers, devices, browser, shopping) atop $1.4T in compute commitments and no profitability. Peers frames Altman's 'code red' as a possible forced narrowing and floats whether OpenAI needs its own Amodei. A genuinely useful strategic framing of the two-lab divergence, above routine digest value.
AM digest led by a meaty Oracle scoop: the firm burned ~$10B in the November quarter on AI data centers, raised FY2026 capex to $50B, and saw shares fall 11% as lenders grow wary of its debt and thin Nvidia-rental margins - with $523B in signed contracts hinging on whether OpenAI can pay. Other notable items: Trump insisting any WBD sale include CNN, Pinterest's 90% open-source AI savings, and Oracle floating customer-owned chips. The Oracle lead is genuinely substantive analysis of AI-capex risk.
Exclusive single-article email on OpenAI's internal financials showing compute margin jumping to ~70% (October) from 52% and 35% earlier, even as it discusses raising up to $100B. The thesis matters because inference profitability is the central question for whether frontier-AI economics can work; original reporting on a company that rarely discloses such numbers. Email is the lede plus related links.
Exclusive reporting that Google's business selling Gemini model access has surged over the past year as model quality improved, with developer API spend rising sharply-a tailwind for Google Cloud's core server business since Gemini runs only on Google's cloud. Matters as evidence Google is converting model momentum into enterprise revenue and closing ground on AWS/Azure in the AI race. Focused single-story analysis on a key Big Tech competitive shift.
A reported deep dive on how AWS CEO Matt Garman has restructured the cloud unit's AI product leadership over the past year-reshuffling teams and poaching execs to accelerate enterprise agent and search tools-accompanied by an updated org chart of 37 AWS executives. Substantive org/strategy reporting on how the cloud leader is reorganizing around the agent bet.
Peers breaks down Alphabet's strong Q4-search ad growth reaccelerating to ~17% and Cloud up 48% to a $71B run rate at 30% margins-arguing AI is now lifting Google's business and putting search-disruption fears to rest, while flagging the capex-vs-cash-flow strain. A genuinely analytical earnings read with the standout point that Google is becoming a rare dual consumer-and-enterprise winner.
A Martin Peers analysis showing big tech's projected 2026 capex ramp will nearly wipe out free cash flow at Amazon, Google and Meta, forcing hard choices on buybacks versus more borrowing - while noting they each have capacity to borrow hundreds of billions more. A useful, numbers-grounded explainer of the financial squeeze from the AI-infrastructure buildout.
An Applied AI feature on ex-GitHub CEO Thomas Dohmke's new startup Entire ($60M seed at $300M valuation, led by Felicis with Microsoft participating), which builds AgentOps tooling (Checkpoints) to monitor AI coding agents' step-by-step actions as developers stop reading generated code. It substantively frames the emerging 'don't-look-at-the-code' workflow and the AgentOps category around it.
A 'Big Read' on Blue Owl Capital co-CEO Marc Lipschultz, who is working on $100B+ in new data-center financing on top of $5.6B equity and $64B debt deployed last year, including previously unreported talks with Crusoe (OpenAI Stargate/Abilene) and Google. It substantively illuminates the private-credit machinery bankrolling the AI data-center buildout.
Miles Kruppa reports from the asset-securitization conference where veterans of the 2008 'Big Short' era are scrutinizing data-center debt deals, with securitizations of digital infrastructure projected to hit $60B and bankers now eyeing chip- and power-generator-backed bonds. A useful, original on-the-ground explainer of how AI buildout risk is spreading to mainstream credit markets and everyday savers.
Analysis of the legal and business stakes after Hegseth's 'supply chain risk' designation, which could sever Anthropic from cloud partners Amazon, Google and Microsoft and is likely to trigger a court battle. It maps the probable next moves in a confrontation with existential implications for the company. Substantive forward-looking analysis of a landmark dispute.
A clear explainer of how OpenAI's planned 'stateful' agent runtime on AWS (agents that remember customer context) routes around Microsoft's exclusive rights to sell 'stateless' OpenAI models, letting Amazon monetize OpenAI tech directly. It frames a structural cloud-rivalry shift and OpenAI's bet that agent/non-API products will exceed API sales by 2028. Genuinely useful for decoding the OpenAI-Amazon-Microsoft three-way dynamic.
This piece reports that the Iran war is jeopardizing $300B+ in planned Gulf (UAE, Saudi) AI data-center and chip spending, threatening a key funding source for power-hungry US tech firms (xAI, OpenAI, Microsoft, Oracle, Google) drawn there by cheap energy. A substantive look at how Middle East geopolitics intersects with the global AI-infrastructure financing picture.
Exclusive that OpenRouter, which routes developers to hundreds of models via one API, is in talks to raise $120M at a $1.3B valuation led by Alphabet's CapitalG, more than doubling its prior ~$500M mark. The scoop captures the rise of multi-model 'router' infrastructure as apps mix and match models, a structurally interesting layer of the AI stack.
Exclusive that Anthropic quietly shifted enterprise pricing to charge business customers by amount of AI used rather than flat fees, meaning heavy Claude users pay significantly more. Framed as how Anthropic manages booming coding/agent demand and the rising cost of running that AI amid a compute crunch. A consequential pricing-model story for the whole enterprise-AI market.
Exclusive on the OpenAI-Cerebras circular-financing arrangement: OpenAI pays $20B+ over three years for Cerebras servers (double the prior figure), receives warrants that could reach 10% of the firm as spend rises, and provides ~$1B to fund Cerebras data centers. A clear case study in the circular AI-financing structures reshaping the chip ecosystem and reducing Nvidia dependence.
Profile of Anthropic CFO Krishna Rao, who quietly became a key force by pushing the company off single-cloud reliance toward multiple chip and cloud partners, with investor Byron Deeter crediting him with the insight that more partners would accelerate growth. A people-and-strategy deep dive explaining the financial architecture behind Anthropic's multi-cloud compute deals.
A reconstruction of the recent Nadella-Altman negotiations that defused a legal dispute over OpenAI's surprise deal to sell its models through Amazon Web Services, which Microsoft argued violated its exclusive cloud-resale rights. Matters because it shows how the foundational Microsoft-OpenAI partnership is being renegotiated as OpenAI's commercial interests diverge from its longtime backer's.
Martin Peers analyzes Apple's standout March quarter (17% revenue growth, 22% iPhone growth, 34% R&D jump) and a notable strategic shift: CFO Parekh abandoning the 2018 'net cash neutral' target and halving buybacks, suggesting Apple wants to stockpile cash for memory-price spikes or acquisitions while rivals burn reserves on capex. A genuinely original read on a quiet but meaningful capital-allocation pivot.
Peers dissects CoreWeave's high-stakes balancing act: revenue doubled to $2B but capex hit $7.7B and it burned $4.7B cash in the quarter, projecting up to $35B capex on $12-13B revenue (3x revenue) against $24.8B debt and only ~$3B cash, much of it from Nvidia. A clear-eyed risk anatomy of the leveraged AI-cloud model and what happens if customers pull back.
Peers dissects why Anthropic publicly announced a confidential IPO filing that commits to no timetable, reading it as a signal meant to contrast its readiness against OpenAI's lagging, fluid plans-while flagging the reputational risk if Anthropic later backs off. The lede also frames Google's surprising $80B equity raise (atop $85B of debt) as proof AI development is as much a financing challenge as a technical one.
Exclusive that data-center developer Switch is in early talks with Brookfield, KKR and other PE/institutional investors to raise billions at a $50B+ valuation, potentially setting up an IPO as soon as next year. A concrete data point on the scale of capital flowing into AI infrastructure buildout.
An AM digest led by a strong exclusive: Satya Nadella sent a sharp internal rebuke after an exec's memo proposed making users 'addicted' to Microsoft's Scout AI agent, revealing the stakes Microsoft attaches to its Copilot/Scout rollout. The digest also carries the Switch $50B+ data-center raise, the S&P 500's decision not to fast-track SpaceX, and reports of US officials weighing equity stakes in AI firms. ---
1 tier-5 · 18 tier-4
The most technically grounded cluster steps back from the business to ask what the models can actually do - and where they stop. The landmark is the inside account of OpenAI's rocky path to GPT-5: dwindling high-quality data, diminishing pre-training returns, and the evidence that scaling laws are bending. Around it sit the post-training frontier (why reinforcement learning and 'universal verifiers' became the new lever, why human experts can no longer teach frontier models, where LLMs still fall short) and the security inflection where off-the-shelf agents autonomously exploit known vulnerabilities faster than defenders can patch. Together they form the reality check beneath every revenue projection: the capability curve, not the capital, is the variable everything else depends on.
Stephanie Palazzolo examines why xAI and other frontier labs are pouring resources into reinforcement learning, using AI models to accelerate the model-training process itself. A useful explainer on a frontier AI-engineering trend with lasting technical relevance, despite the thin email teaser.
An AI Agenda piece drawing on a marquee AI conference to highlight the current gaps and limitations of large language models. A useful state-of-the-art reality check on where the technology still falls short.
An exclusive that Meta has poached three Google DeepMind researchers who worked on the Gemini model that achieved IMO gold-medal performance, bringing Meta's DeepMind hires to at least six. It documents the escalating AI-talent war and Zuckerberg's aggressive superintelligence-team buildout.
An AI Agenda analysis on how AI developers are experimenting with more advanced forms of reinforcement learning to push model capability beyond current RLHF approaches. A useful technical-trend explainer on where frontier-model training is heading.
Reports the backstory on why the White House's AI Action Plan conspicuously omitted copyright recommendations, revealing internal tension over how to treat training-data IP. Useful policy reporting on a question central to AI's legal future.
A Big Read by Kalley Huang and Erin Woo on how immigration-visa friction is stranding foreign AI researchers (illustrated by Windsurf staff left in limbo during Google's $2.4B acqui-hire) and why US immigration restrictions threaten its AI lead. Substantive feature connecting the talent war to immigration policy.
The Applied AI newsletter unpacks a University of Chicago/Rotterdam study showing AI-conducted job interviews led to 12% more offers and higher retention than human interviewers, with 78% of candidates preferring AI; a second segment covers backlash over AWS Kiro's coding-tool pricing. A substantive, data-grounded explainer on AI in hiring plus a real signal on AI coding-tool economics.
Exclusive: model makers including OpenAI, xAI and Anthropic have discussed deals to acquire or license proprietary coding data from startups like Cursor's maker, underscoring how valuable real-world coding interaction data has become for training frontier coding models. Notable signal on the emerging market for training-data deals.
Analysis of why chatbots' tendency to endorse users' delusional beliefs ('AI psychosis') is structurally hard to eliminate-the sycophancy that makes models endorse delusions is entangled with the engagement-optimizing behavior companies want. A thoughtful look at an AI-safety/product tension, though the captured body is mostly paywall stub.
Single-article promo for a feature reporting that AI researchers are increasingly recording all of their work so models can learn to perform their jobs. It's a substantive look at data-collection practices feeding the automation of knowledge work, though this email is mostly a subscribe wrapper around the lede.
Single-article email on a surprisingly simple inference-time method that sharply boosts the math performance of OpenAI, Google, and xAI models. Relevant to understanding test-time/reasoning techniques and how much headroom remains via prompting/scaffolding rather than retraining. Body paywalled here, but the lede points to a substantive technical-capability angle.
Feature on how the AI talent war is reshaping academia-professors now weigh whether prospective PhD students will stay in their programs or bolt early for lucrative corporate AI jobs, complicating university recruiting and retention of NLP/ML researchers. A useful look at the second-order effects of the AI hiring frenzy on the research pipeline that feeds the field. Substantive enough lede to read.
Reports that AI developers are tapping insurance-company assets as a new financing channel to meet their enormous capital needs. A fresh angle on the financial engineering behind the AI capex boom that matters for understanding how labs and infrastructure buildouts are funded. Lede only here, but the thesis is a distinct, non-obvious funding-structure story.
Taser-maker Axon is in advanced talks to buy an emergency-response software startup backed by General Catalyst and Andreessen Horowitz for about $800 million. A clean M&A scoop on public-safety software consolidation and a notable VC exit.
Reported look inside Scale AI's expansion, on the heels of its Meta deal, into labs that generate training data to teach robots to think and move. Signals the data-labeling giant's pivot toward physical AI and robotics foundation models.
An exclusive scoop reporting that a longtime Meta AI executive mused about quitting as staffers chafed at a new process for publishing research, signaling internal turmoil in Meta's AI lab amid its superintelligence reorg. Original reporting on a topical fault line in Big Tech's AI talent wars, though the email body is largely paywall (lede carries the story).
Exclusive revealing that data-labeling startups are booming on AI labs' appetite for expert human graders: Handshake's AI-training annualized revenue leapt to nearly $1B from $550M in January and ~$5-10M a year ago. Concrete figures illuminate the unglamorous human-contractor layer beneath the AI boom. Useful original data point on a fast-growing market segment.
An original-reporting analysis of how AI labs mine their own employees for training data: Microsoft uses 100K engineers' proprietary code and Xbox studio source; Meta tracks mouse movements and browser activity via its 'Model Capability Initiative' (drawing internal resistance); xAI paid staff $420 for tax returns. The kicker thesis is that employee data is the easiest to obtain because workers can't really say no. A second segment details Salesforce's $300M Anthropic bill and Benioff's plan to route most tokens to cheaper models. Substantive, lasting insight into an underexamined industry practice.
Peers identifies a paradox: cybersecurity execs (CrowdStrike, Netskope, Palo Alto) tout AI-driven 'structural demand,' and stocks are up 50%+, yet reported growth rates barely improved or slowed, suggesting investors got ahead of themselves. Also reframes SpaceX's 'record' $75B IPO as smaller than Google's $84.75B equity raise the same week. ---
0 tier-5 · 33 tier-4
Beyond the named giants, this is the venture ecosystem in full deal-rush mode. The reporting tracks the money - solo-GP mega-funds, declining dry powder, unsolicited term sheets, insurance capital as a new AI funding source - and the eye-watering marks: Cursor at $30B, Thinking Machines testing $50B with a talent exodus, Polymarket and Kalshi, Ramp, Vercel, Supabase, Cohere, Anduril's cash-burning defense valuations. It also catches the frontier oddities that signal where capital thinks the next platform is - humanoid-robot developers, physical-AI location startups, longevity and gene-editing bets. The recurring tension is whether this is durable platform formation or a deal market detached from the underlying economics, a question the skeptical valuation pieces keep pressing.
A Big Read feature on Joby Aviation, the eVTOL flying-taxi startup nearing commercial takeoff with manufacturing and capital support from Toyota. Substantive deep-dive on a hard-tech sector reaching an inflection point, though the email body is mostly paywall and the analysis sits behind the link.
A deals feature compiling 50+ AI startups that could be acquisition targets as buyers increasingly chase products and technology, not just talent (a shift from acqui-hire). A useful reference list mapping the M&A landscape, though the email surfaces only the premise before the paywall.
Exclusive that Elad Gil is raising a $1.5B fund, among the largest ever for a solo GP, stretching the definition of solo-capitalist venture investing. A notable VC-structure scoop signaling how concentrated AI-era capital and brand-name solo investors are reshaping fund formation; the email is the paywalled teaser.
Scoop that Meta is pursuing deals -- licensing or acquisition -- with AI video-generation startups, in talks with Pika and having explored Higgsfield and Runway, as part of its aggressive AI buildout. Shows Meta extending its talent-and-tech land grab into generative video.
Video-AI startups Runway and Luma, after early traction in Hollywood, are pursuing a new revenue stream selling synthetic-video/world-model tech to robotics companies for training data. Signals how generative video is pivoting toward the physical-AI market.
Aaron Holmes's Applied AI column shows how M&A brokerages are using ChatGPT Pro/Deep Research to replace the buyer-research work of junior bankers, with one boutique founder saying the output beats 60% of junior analysts and lowers barriers to entry for the industry. A parallel section on McKinsey deploying 12,000 AI agents while cutting headcount ~11% extends the thesis to consulting cannibalization. Concrete, well-sourced reporting on AI displacing white-collar advisory work.
An exclusive scoop that video-AI startups Runway and Luma are in talks to raise billions more as their revenue picks up. Matters as a marker of momentum and capital concentration in generative video; the email is the promo-stub but the funding news and revenue-rise framing are clear.
Jessica Lessin's editor's pick spotlights Cognition offering buyouts (nine months' salary) to ~200 Windsurf employees just three weeks after acquiring them-after laying off 30-illustrating the brutal limbo facing rank-and-file staff in AI acqui-hires where founders leave and equity gets cashed out at a discount. A pointed window into the human cost of AI consolidation, drawn from substantive Mascarenhas/Huang reporting.
Natasha Mascarenhas reports on the AI funding frenzy-unsolicited term sheets and dwindling VC dry powder as rapid startup revenue growth eases investor skepticism. A useful inside look at deal-market dynamics driving the current AI bonanza, though the email is a teaser with the body paywalled.
An exclusive revealing that before its splashy $34.5B unsolicited Chrome bid, Perplexity had quietly courted smaller browser makers-offering ~$1B for Brave and talking to The Browser Co. and DuckDuckGo, all without a deal. Reframes the Chrome bid as part of a deliberate browser-acquisition strategy by the AI search startup.
Jessica Lessin's editor's pick frames Natasha Mascarenhas's reporting on the AI funding frenzy: VCs sending unsolicited term sheets before rounds even open, dry powder potentially hitting its lowest since 2019, and founders dictating valuations and governance. A useful synthesis of the structural shift in AI-era startup fundraising dynamics.
A year after its founders left for Google, Character.AI is exploring a sale or new fundraise as it struggles with the cost of running its companion chatbots. A solid exclusive on a notable consumer-AI startup's strained economics and strategic options.
Apple's cultural and structural aversion to large acquisitions and partnerships is hampering its ability to catch up in AI, where rivals are spending and dealmaking aggressively. With Siri lagging and Apple weighing outsourcing the model behind it, the piece argues that its conservative M&A posture is a strategic liability. Matters because it frames Apple's whole AI predicament as a self-inflicted constraint, not just a tech gap.
Peers makes a sharper-than-usual argument that Tim Cook is right to resist Eddy Cue's push for big AI acquisitions (Mistral, Perplexity), citing the long history of value-destroying megadeals versus cheap home runs like PA Semi and Instagram, and contending Apple is better off pitting AI labs against each other for access to its user base. A well-reasoned strategy essay, though wrapped in the nightly digest.
A startup billing itself as 'ChatGPT for doctors' (OpenEvidence) is weighing investment offers at a $6 billion valuation, a sharp markup that captures investor appetite for clinical-AI tools. It is a useful datapoint on how fast healthcare AI valuations are climbing as OpenAI and others push into the same space.
Jessica Lessin's editor's pick spotlights a deep feature on how Apple's build-in-house, small-deals-only culture is keeping it on the AI sidelines while Meta and Google spend billions on talent and licensing-with Eddy Cue pushing bolder M&A (even past pitches to buy Netflix and Tesla) but repeatedly overruled by Cook. The framing email is promo, but it carries the substantive thesis and standout details of a strong original feature.
Martin Peers argues Google effectively 'won the antitrust war'-Judge Mehta accepted Google's proposed remedies 'in full' with minor tweaks, rejected Chrome/Android divestiture, and crucially acknowledged generative AI 'changed the course of this case,' a sign of a more pro-business Trump-era regulatory climate. Also dissects Klarna's unusual no-economic-rights dual-class supervoting structure ahead of its IPO. Substantive analysis of a pivotal ruling.
An opinion piece by investor Vinod Khosla arguing that while most AI valuations are 'bonkers,' sitting out the bubble entirely means missing the handful of massive winners. A named-investor framing of the bubble-vs-power-law tension that's worth reading for the perspective, though brief.
September's packed IPO calendar is finally delivering distributions to venture limited partners after a three-year cash crunch in the asset class. Solid VC-market analysis on how the IPO reopening eases LP liquidity pressure and what it means for fund dynamics.
Scoop that humanoid-robot startup 1X is targeting $1 billion in new funding to compete with Tesla and Amazon in bringing humanoid robots into the home. A meaningful signal on robotics-funding momentum and the consumer-humanoid race, though this email is the gated single-article teaser.
A 'Big Read' profile probing whether ex-OpenAI CTO Mira Murati's $10B-valued lab (Thinking Machines) can deliver on its enormous hype with little public product or track record. Substantive examination of one of the most richly valued and least-understood new AI labs.
A TITV interview with Anduril CEO Brian Schimpf, who says the $30B-valued defense startup will double revenue this year while addressing the perception gap between Silicon Valley enthusiasm and traditional-defense skepticism, government shutdown impact, and space warfare. Substantive founder Q&A on defense-tech growth and the new-vs-legacy contractor divide.
Peers uses Jeff Bezos's declaration of an AI bubble to argue the bubble is now conventional wisdom, reframing the real questions as when it bursts and who survives, with TSMC and Nvidia best positioned and undifferentiated AI startups most at risk (only ~a dozen beyond OpenAI/Anthropic exceed $100M revenue). A clear-eyed framing of AI bubble dynamics plus a week-in-review of The Information's scoops.
Peter Thiel's Founders Fund is shifting from caution to concentrated, large-scale bets on AI, wagering that scale will determine winners in the AI race. Notable as a strategy reversal from a marquee VC and a read on how top investors are positioning for the AI cycle.
Editor's-pick deep dive on defense-tech firm Anduril burning $800-900M in cash this year to chase Pentagon contracts while becoming the hottest private-market stock-secondary investors pay nearly double the $30.5B last institutional round, implying $70B+. A useful case study in the disconnect between cash burn and private-market appetite, with talk of a $50B+ IPO within a year. Matters as a read on defense-tech valuations and secondary-market froth.
A Briefing built around Waymo opening freeway driving in three metros and its multi-city 2026 expansion (4.3M monthly rides, up from 3.5M), making the case that Waymo is 'living the dream' while Tesla still runs safety drivers and faces a possibly-never-profitable robotaxi unit tied to Musk's $1T pay package. A meaty read on the AV competitive gap, plus a sharp riff on the crowded humanoid-robot field (Andy Rubin's new startup, Figure, Xpeng).
Exclusive on the unraveling at Mira Murati's Thinking Machines Lab: CEO Murati fired co-founder/CTO Barret Zoph for poor performance and talking to competitors, while researchers Luke Metz and Sam Schoenholz quit in the same all-hands, as the ~40-person lab seeks a $50B-valuation round. A consequential scoop on instability inside a top AI talent bet and its fundraising risk. Substantive single-story reporting.
A guest opinion essay by Niantic founder John Hanke arguing that AI capital should flow toward physical-world applications rather than attention-harvesting content and ad optimization, drawing a parallel to his earlier metaverse critique. It is an original point-of-view piece on AI's misdirected priorities, though more provocation than rigorous framework.
A deep on-the-ground look at Anduril's 866,000-sq-ft 'Arsenal 1' factory near Columbus, Ohio, the centerpiece of its bet to scale manufacturing capacity and deliver on goals pitched to investors during a $4 billion raise. The defense-tech startup envisions a Silicon Valley-style campus of a half-dozen buildings on 500 acres. Substantive feature on the industrialization of defense tech and Anduril's manufacturing-led strategy.
Analysis teaser arguing tech M&A has stalled because CEOs are paralyzed by uncertainty over how AI will reshape both their own businesses and acquisition targets, with dealmakers at the Tulane Corporate Law conference citing AI-driven 'fear of irrelevance.' A useful read on a structural dampener on dealmaking, sourced to named practitioners, though the email is a teaser.
A narrative feature following a tech CEO who used AI chatbots to research and pressure-test treatment options after his wife's brain tumor returned and doctors offered no cure. A substantive human-interest deep dive on AI's real and limited role in personal medical decision-making, illustrating both the promise and the boundaries of chatbot-assisted medicine.
Stealthy nine-year-old startup ZaiNar (backed by Steve Jurvetson and Jerry Yang, valued at $1 billion) is exiting the shadows with location-tracking tech it claims is accurate to within inches indoors and out, targeting $5 billion in deals as a GPS alternative for the physical-AI era. A deep-dive startup profile on infrastructure for robotics and physical AI.
Deep feature on Silicon Valley's well-funded longevity startups (NewLimit, Retro Biosciences, Life Biosciences) and the conundrum that despite raising billions to reverse aging, none can yet run trials for 'longevity' drugs themselves - regulators have no aging endpoint, so they must target specific diseases. A substantive explainer of the structural FDA roadblock facing the anti-aging biotech wave. ---
0 tier-5 · 24 tier-4
The incumbents' beat is increasingly a policy and antitrust beat. The spine is the Google antitrust endgame - barred from exclusive deals but allowed to keep Chrome, the unresolved search-data-sharing order, the ad-tech unit's forced independence - read alongside Meta's monopoly fight, Apple's deal-averse culture stalling its AI push, and Amazon locking down against rival shopping agents. Running through it is the Trump-administration overlay: equity stakes in chip and critical-minerals firms, DOGE turning AI on regulations, pressure on Microsoft, and the immigration squeeze on foreign AI researchers. The cluster is the map of where state power and platform power now collide, and how the giants are bending to (or resisting) it.
An Org Charts feature mapping 60+ of the top leaders and researchers running Meta's new Superintelligence Labs, the org built from Zuckerberg's aggressive AI talent spree. High reference value for tracking Meta's AI org structure and the talent war, though it's a directory/data artifact rather than narrative analysis.
An exclusive reporting that Apple's AI reorg, meant to fix its lagging efforts, is instead driving key researchers to rivals for reasons beyond compensation. It diagnoses Apple's structural and cultural disadvantages in the AI talent war.
Martin Peers argues Google should merge its Search bar and Gemini app, citing Pichai's earnings-call hints, ChatGPT's single-interface advantage in user numbers, and the internal-fiefdom/$200B search-ad-revenue reasons holding the split in place. A genuine strategy argument on Big Tech's AI-product structure, plus a note on Trump's copyright comments favoring AI firms.
An exclusive profile of Koray Kavukcuoglu, Google's new chief AI architect, and his plan to inject Gemini models across all Google products. The strategic-leadership angle on how Google operationalizes Gemini distribution is substantive, though the email body is mostly paywall (lede only). Worth reading for Big Tech AI-strategy watchers.
Catherine Perloff reports that advertisers say Google search ads have grown less effective (worse traffic quality, diminishing returns) yet they keep spending because alternatives remain limited, revealing the depth of Google's ad-market lock-in. A useful analysis of advertiser dependence amid the AI-search disruption threat; promo-stub email but the thesis is explicit.
Walmart pulled thousands of independent-merchant product listings from its marketplace as it pivots to court big-name brands, signaling a strategic reorientation of its third-party platform. The move matters because it reshapes who wins and loses on a top-tier US e-commerce marketplace and reflects Walmart's premiumization push against Amazon.
A counterintuitive analysis explaining why some Google rivals argue that forcing Google to share its search data as an antitrust remedy could actually hurt, not help, competition. A worthwhile explainer on the nuances of the search-monopoly remedy debate ahead of Judge Mehta's ruling.
Exclusive: DOGE is deploying AI to identify SEC regulations-including diversity rules-for possible elimination, a concrete look at AI being used inside government to drive deregulation. Matters as an early case study of AI in the federal policymaking/regulatory-cutting process, with implications for capital-markets oversight.
Standalone scoop on the Google search-monopoly remedy ruling: the judge rejected the DOJ's most drastic proposals (no Chrome/Android divestiture), barred exclusive distribution contracts, and ordered limited search-data sharing with rivals while allowing continued paid distribution (e.g., the ~$20B/yr Apple deal). The core news article behind the day's AM digest and Briefing-lasting reference value on the case outcome.
Wayne Ma and Juro Osawa report that Amazon is developing consumer AR glasses, entering the still-unproven smart-glasses market against Meta. A substantive Big Tech hardware scoop signaling Amazon's wearables ambitions; email body paywalled but the story is an exclusive.
Erin Woo reports that the limits of the federal judge's antitrust remedy ordering Google to share search data have become clearer, leaving smaller search rivals unsure the order will meaningfully help them compete. A substantive policy/antitrust piece on the gap between the Google remedy's promise and its practical effect.
Apple told at least two suppliers to raise entry-level iPhone 17 output by 30%+, signaling unexpectedly strong demand for the cheaper model. An exclusive supply-chain scoop that reads Apple's demand signals ahead of earnings.
Scoop that the Trump administration is increasingly demanding equity stakes in critical-minerals companies as a condition of releasing Biden-era grants and loans. Documents an expanding pattern of government taking direct ownership in strategic industries, paralleling its Intel stake.
Scoop that Meta has discussed with Google Cloud using Gemini models to improve its ad-targeting business - a notable instance of two ad rivals cooperating on core AI infrastructure. Signals that even Meta sees value in outside frontier models for its highest-stakes product.
An Electric exclusive by Steve LeVine reporting that Trump officials have embraced seafloor-mining firms like The Metals Co., and markets are betting the administration will take equity stakes as it has in other strategic-resource deals. Matters as original reporting on an emerging pattern of government equity grabs in critical-minerals and the geopolitics of deep-sea mining amid the rare-earths standoff with China. Substantive policy/resources scoop.
An exclusive on how Amazon's AI-tweaked shopping search is pushing merchants to rewrite product listings in a more conversational style to stay discoverable. Matters because it is a concrete, ground-level look at how generative search is reshaping retail SEO and seller behavior, an under-covered angle on AI's commercial impact. Substantive applied-AI reporting with practical implications for the e-commerce ecosystem.
Exclusive reporting that a Google contractor (working for Accenture in the Philippines) took ~2,000 screenshots and sensitive internal Play Store files-code, developer-account data, and fraud/malware review processes-and sent them to an outsider via WhatsApp/Telegram. Raises real questions about app-store guardrail integrity and contractor security. Original single-story scoop.
A Briefing making a clean bull case for Alphabet: stock up 50% YTD on Buffett's $4.3B stake, antitrust relief, TPU momentum and cloud growth, with the argument that Google is the already-vertically-integrated AI giant (model, lab, chip, cloud, browser, devices) that OpenAI is straining to become. The reversal-of-fortunes contrast with Meta's 20% slump sharpens the thesis. More original framing than a typical digest.
Sunday Insights analysis arguing big-tech headcount rebounds after layoffs-using Meta's ~10% Reality Labs cut as the opener and the 2022-23 layoff cycle at Meta, Alphabet, Amazon and Microsoft as evidence that workforces shrink only temporarily before resuming growth. A useful, data-backed counter-narrative to the 'AI is gutting tech jobs' framing. Worth reading as durable context on tech employment dynamics.
A profile of SEIU-UHW president Dave Regan, who filed a one-time 5% California wealth tax on billionaires and split his own labor coalition in the process. It's a substantive policy feature on a wealth-tax fight that directly targets tech fortunes and is rippling into a national debate.
Standalone analysis of the three challenges facing incoming Apple CEO John Ternus: reducing China manufacturing dependence amid geopolitical risk, fixing Apple's lagging AI efforts, and managing executive-chair Cook looking over his shoulder. A structured framework piece on a major leadership transition with durable reference value.
A Sunday Insights analysis of 100 public tech companies' earnings transcripts showing AI is cutting both ways on margins-boosting productivity and trimming headcount for many (Spotify, Uber, Airbnb) while the rising cost of running AI depresses margins for others. Useful framework piece on the unsettled economics of AI adoption across the tech sector.
Peers analyzes Amazon retiring its 'Rufus' shopping chatbot brand for the familiar Alexa name and merging its search bar with AI features-following Google's playbook of converging search and chatbot-as a bid to fend off ChatGPT and Gemini in shopping. Strong secondary section on the AI compute crunch (Nebius seeing 'four or more customers per GPU,' 684% revenue growth). Solid Big Tech strategy read with the GPU-shortage data point.
Scoop based on an internal memo from wearables VP Alex Himel: Meta plans to test an AI pendant within a year, expand AI glasses, and add a 'Wearables for Work' business service-aimed at reversing hardware-division losses and driving usage of Meta's AI models, subscriptions and the Hatch agent. A concrete look at Meta's hardware roadmap and its bid to own an AI-native device category. ---
0 tier-5 · 15 tier-4
The narrowest but fastest-moving cluster: the application layer where coding agents and consumer agents are being built and monetized. The coding-tools story dominates - Cursor's rise (lifted partly by China), the margin squeeze as demand soars, Claude Code blowing up AI budgets, the 'Cursor resistance' of engineers who distrust the tools, and OpenAI and xAI circling Cursor's training data. Alongside it runs the agent-platform jockeying: OpenAI making agents easier to build but harder to switch models away from, Andreessen backing the tools agents themselves use, and the chaotic consumer-agent moment after OpenClaw. It's the ground floor of the agent economy - where the abstractions the enterprise beat argues about get their first real code.
An exclusive on Amazon clamping down on AI shopping agents-particularly Google's-as e-commerce giants wrestle with who controls the agentic-buying layer between consumers and retailers. A substantive scoop on an emerging Big Tech battleground (agent-mediated commerce) with strategic implications for both Amazon's marketplace control and Google's AI ambitions. Body paywalled; lede carries the thesis.
Anita Ramaswamy's 'True Value' column argues that AI coding-assistant startups like Cursor have become overvalued given the real risks (thin margins, model dependency, competition) underlying the vibe-coding hype. A clear contrarian valuation thesis with lasting relevance to the AI startup funding debate.
An exclusive using previously unreported financials from Replit to show the high underlying costs (thin or negative gross margins) behind the surging revenue of AI coding-assistant startups. Matters because it quantifies the unit-economics problem facing the vibe-coding boom; the promo email carries the thesis even though the full data is paywalled.
Profile of a small startup that scrapes Google Search results and has quietly become critical infrastructure for OpenAI, Cursor, Perplexity and Meta, revealing how much top AI products depend on Google's index they can't legally access directly. A useful explainer on a hidden dependency in the AI search supply chain.
a16z led a $23M financing of a startup building tools designed for AI coding agents rather than human coders to use. It's an early signal of an emerging infrastructure layer where the customer of dev tooling is the agent itself, a notable shift in how software gets built.
An exclusive scoop by Natasha Mascarenhas that Anysphere, maker of the hit AI coding assistant Cursor, is weighing funding offers valuing it around $30 billion. Matters as a marker of how aggressively investors are repricing AI-coding leaders and Anysphere's rapid valuation climb. Original, single-focus funding scoop with clear startup/VC relevance.
A feature on engineers who resist mandates to use Cursor and other AI coding assistants, opening with two Mixus engineers who staged a mini-revolt and a frustrated founder demanding heavier adoption. Matters as a grounded counter-narrative to the AI-coding hype, surfacing the productivity-vs-craft and trust tensions inside small teams. Original reporting with a memorable lede and lasting relevance to how AI tools actually get adopted.
An analytical mapping of the web of partnerships among 10 AI firms, retailers, and payments companies forming temporary alliances to capture AI-driven shopping and challenge Amazon. The frenemy-landscape framing and partnership table give it useful reference value for the emerging agentic-commerce ecosystem.
Meta has formally wired AI usage into its performance and bonus system via 'Checkpoint,' an AI tracker measuring AI-generated code and 200+ data points, with bonus multipliers up to 300% for top performers-part of Zuckerberg's push to flatten teams. A well-sourced look (with internal recording/memo) at how a major firm is mandating AI adoption, plus a sidebar on Microsoft's blunt internal OpenClaw security review.
A feature on how OpenClaw's viral launch has triggered a 'mania within a mania' around consumer/personal AI agents, pushing startups like Jo to rush competing on-device-model products to market. It substantively captures the inflection point where agentic AI began going mainstream for home tasks and the competitive scramble it set off.
Aaron Tilley's column argues Apple's App Store guidelines (notably rule 2.5.2) are colliding with vibe-coding apps that let users spin up apps from prompts, with developers like Bitrig seeing updates rejected since November-possibly because such tools threaten Apple's app-distribution control. It uses a concrete developer case to frame a real policy-and-platform-power tension. Substantive analysis with a clear thesis on Apple's gatekeeping in the AI era.
Exclusive that Cisco is in talks to acquire Israeli AI-agent-security startup Astrix for $250-350M, a 25%+ premium over its ~$200M valuation, as exploding agent adoption drives incumbents to buy guardrail tech. Concrete M&A scoop signaling a wave of consolidation in agent security. Useful market-direction signal for the AI-security beat.
A feature on OpenClaw, one of the fastest-growing open-source AI agent projects ever, wrestling with whether to professionalize from a volunteer hacker project into reliable enterprise tooling without losing its experimental ethos. Matters as a case study in the governance and sustainability strains facing viral open-source AI infrastructure as it confronts real production demands.
Scoop that Meta is building a consumer OpenClaw-style agent ('Hatch,' internal testing targeted by end of June) trained in sandboxed simulations of sites like DoorDash, Etsy, Reddit and Yelp, plus a separate agentic shopping tool for Instagram. Concrete look at Meta's push to monetize its AI investment through consumer agents.
Exclusive that Meta is weighing up to $200/month for its planned OpenClaw-based consumer AI agent 'Hatch,' with a $199.99 premium tier, signaling an ambition to compete head-on with top-tier offerings from OpenAI and Anthropic. Concrete evidence of Zuckerberg's push to build new revenue streams to fund Meta's enormous AI infrastructure spend. ---