David Shapiro · Tech & AI
TIER 4 Fri, 5 Jun 2026 11:56:32 +0000
Watch now (26 mins) | Bernie Sanders wants the federal government to own half of OpenAI. Half of Anthropic. Half of Elon Musk's xAI. ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ | | ---|---|--- | | | Forwarded this email? Subscribe here for more --- --- --- | | Watch now --- # Bernie Wants Half of Big AI ### Bernie Sanders wants the federal government to own half of OpenAI. Half of Anthropic. Half of Elon Musk's xAI. | | David Shapiro --- | Jun 5 --- | --- --- | | | --- | | --- | | --- | | --- | | READ IN APP --- # Who Should Own the AI Windfall? > The problem is not the idea that the public should share in AI's upside. The problem is trying to seize half of today's model labs while missing the rest of the value chain. Bernie Sanders has proposed a new American AI Sovereign Wealth Fund, funded by a one-time 50% tax paid in stock by the largest AI companies. Not a 50% tax on profits. Not a windfall tax on revenue. A 50% equity tax. Under the proposal, companies like OpenAI, Anthropic, xAI, and other major AI firms would transfer half their stock into a federally managed wealth fund. The public would then own a major stake in these companies. The government would also get voting shares and board representation, giving it influence over the future direction of AI. The money generated by the fund would eventually flow back to Americans, first through direct payments and later through broader public benefits like health care, education, and housing. The moral argument is straightforward: AI companies trained their models on the collective output of humanity--books, code, journalism, art, research, conversations, images, and the open internet. If AI creates trillions of dollars of wealth from that shared inheritance, Sanders argues, the public should receive a share of the upside. That is the strong version of the case. And honestly, it is not crazy. If AI becomes a general-purpose wealth engine, the question of who owns the gains is going to become one of the defining political fights of the next decade. If a handful of firms can automate large portions of white-collar labor, absorb huge quantities of public data, draw heavily on public infrastructure, and generate massive private valuations, it is reasonable to ask whether the public deserves more than cheaper chatbots and vague promises of future abundance. But Sanders's actual proposal is not a serious version of that idea. It is a messaging bill. And once you see it that way, the whole thing makes a lot more sense. ## The bill is designed to start a fight, not close a deal A 50% stock tax with government board seats is not where you start if you are trying to write passable legislation. It is where you start if you are trying to force a conversation. That does not mean Sanders is confused. Quite the opposite. Sanders has been doing this for decades. Medicare for All, the $15 minimum wage, wealth taxes, student debt cancellation--his political style is to introduce maximalist demands that have little short-term chance of passage but shift the debate. This AI fund fits the same pattern. The headline is simple and explosive: the public should own half of the AI companies. That is guaranteed to generate attention. It is also guaranteed to trigger every possible counterattack: socialism, nationalization, confiscation, capital flight, bureaucratic control, government overreach. But that is the point. The proposal is not optimized for markup. It is optimized for salience. Sanders wants to drag the AI debate away from "How do we help companies innovate faster?" and toward "Who owns the wealth created by AI?" As political theater, it works. As legislation, it has massive problems. ## The biggest flaw: it targets the wrong part of the AI economy The proposal focuses on frontier model companies: OpenAI, Anthropic, xAI, and similar labs. That makes sense rhetorically. These are the companies most associated with scraping the internet, training large models, and building systems that could reshape the labor market. But if the goal is to capture the economic windfall from AI, the target is far too narrow. The real AI profit stack is much broader than model labs. It includes Nvidia and AMD selling the chips. It includes Microsoft, Google, Amazon, Oracle, and Meta building the cloud infrastructure. It includes data center operators, utilities, cooling systems, power companies, enterprise software platforms, and firms that use AI to expand margins by replacing labor. In other words: the biggest winners may not be the companies building the models. They may be the companies selling the picks and shovels. That is the fatal design problem. If you tax OpenAI and Anthropic but leave Nvidia, Microsoft, Google, Amazon, Meta, Oracle, CoreWeave, and the energy infrastructure untouched, you are not taxing the AI boom. You are taxing a visible subset of the AI boom. Even worse, you may be taxing the less profitable layer. Model labs burn enormous amounts of cash on training, inference, talent, and distribution. The chipmakers, hyperscalers, and infrastructure owners may capture cleaner and more durable rents. So Sanders's proposal risks doing the worst possible thing: punishing the most obvious AI companies while missing the companies that actually capture the largest and most reliable profits. That is not a minor technical issue. It is the whole ballgame. ## The second flaw: government ownership creates conflicts Sanders's proposal does not just create a public investment fund. It gives the government voting power and board representation. That is where the idea becomes especially messy. The government already regulates these companies. It buys from them. It investigates them. It sues them. It approves or blocks mergers. It sets export controls. It shapes procurement. It funds research. It regulates energy, data centers, chips, privacy, labor, and national security. If the government also becomes a major shareholder, it becomes both referee and investor. That creates obvious conflicts. Would regulators crack down on companies the public fund owns? Would antitrust enforcement weaken if the government has a financial interest in incumbent dominance? Would future administrations pressure AI companies to make politically useful decisions? Would board seats become partisan spoils? Even if you support democratic oversight of AI, direct government control over corporate boards is a blunt and dangerous tool. A sovereign wealth fund should capture upside. It should not turn the state into a conflicted co-manager of frontier AI firms. ## The third flaw: it is legally and mechanically chaotic A 50% equity tax sounds clean until you ask basic implementation questions. How do you value private companies like OpenAI or Anthropic? What counts as an "AI company"? What happens to Microsoft, where AI is central but not the whole business? What about Google, Amazon, Meta, Nvidia, Oracle, and Broadcom? What about foreign investors? What stops companies from restructuring to avoid classification? What happens if a company dilutes the public stake later? Sanders has acknowledged that companies where AI is only part of the business are complicated. But that complication is not some footnote to be handled later. It is the central administrative problem. If you cannot define the tax base clearly, the policy becomes arbitrary, gameable, and politically indefensible. ## The better version: tax AI rents wherever they appear A serious AI wealth fund should start from a different principle: Do not target named AI companies. Target AI rents wherever they show up. That means the policy should be broad, measurable, and tied to economic activity rather than to a short list of current winners. A better bill might have five parts. First, a broad AI windfall levy on large firms that earn extraordinary returns from AI-related activity. This would apply across the value chain: model labs, chipmakers, hyperscalers, data centers, enterprise software firms, and major deployment winners. Second, a compute-based levy on very large-scale AI training and inference. Compute is the physical bottleneck of the AI economy. Taxing massive GPU clusters, frontier training runs, AI accelerator purchases, or high-intensity data center use could capture value without needing to decide whether a company is "really" an AI company. Third, public warrants or revenue-sharing when companies benefit from public bottlenecks. If a firm receives federal subsidies, tax credits, procurement contracts, grid priority, federal land, public data access, export privileges, or fast-tracked permitting, the public should get upside. Fourth, a professionally managed national AI dividend fund. The fund should be independent, diversified, and insulated from day-to-day politics. If it holds equity, that equity should generally be nonvoting or voted by independent fiduciaries under strict rules. Fifth, automatic public distribution. The fund should not become a vague Washington slush fund. A defined share of returns should go directly to households as an annual AI dividend, with another share potentially funding worker retraining, health care, education, housing, or local communities bearing data center costs. That version is much harder to dismiss. It preserves the core moral insight: if AI generates extraordinary wealth from public knowledge, public infrastructure, and broad social disruption, the public should share in the upside. But it avoids the worst parts of Sanders's proposal: the arbitrary target list, the 50% equity seizure, the government board seats, and the overconcentration in a few speculative companies. ## The real debate is just beginning Sanders's proposal is bad policy design. But it is not meaningless. Its value is that it forces the right question into the open: who owns the AI windfall? The status quo answer is simple: founders, investors, chipmakers, hyperscalers, data center owners, and the companies best positioned to automate labor. Sanders's answer is also simple: the public should own half. The better answer is more complicated: the public should capture a broad, durable, carefully designed share of AI-generated rents without turning the federal government into the co-owner and co-manager of a few named companies. That means taxing the whole stack, not just the model labs. It means dividends, not vague promises. It means nonvoting public upside, not politicized board control. It means treating AI less like a company-specific scandal and more like a structural economic shift. The AI boom is not just happening inside OpenAI, Anthropic, and xAI. It is happening in chips, cloud, power, data centers, enterprise software, logistics, finance, medicine, defense, and every company that uses AI to reduce labor costs or expand margins. If the public is going to claim a share, the policy has to follow the money. Sanders has opened the debate. Now someone needs to write the serious version. You're currently a free subscriber to David Shapiro's Substack. For the full experience, upgrade your subscription. Upgrade to paid --- | | | Like --- | | Comment --- | | Restack --- (C) 2026 David Shapiro 548 Market Street PMB 72296, San Francisco, CA 94104 Unsubscribe