Bernie Sanders' American AI Sovereign Wealth Fund proposal

What separates Sanders' proposal from a conventional windfall tax is the payment instrument. Rather than collecting cash, the bill imposes a one-time 50% tax that AI companies satisfy by handing over shares, depositing that equity directly into a public sovereign wealth fund ^[1]. The mechanism sidesteps the obvious objection that frontier labs are cash-poor: OpenAI and Anthropic remain unprofitable, so a cash levy would be hard to pay ^[2]. Paying in stock instead converts paper valuations into public ownership without touching operating cash.

It also reshapes the relationship from taxpayer-and-treasury to shareholder-and-company: the fund's 7-member Independent Commission for Democratic AI would wield voting shares to block harmful corporate decisions, and mixed companies would have to carve their AI divisions away from non-AI businesses so the public stake attaches only to AI operations ^[1]. That is governance leverage, not just revenue.

The most surprising feature of this debate is who agrees. Sanders frames the bill as a check on Big Tech oligarchs, but the idea of government taking AI equity already has traction on the right. The Trump administration has been pursuing voluntary, passive stakes in AI firms and is in active talks with OpenAI about one ^[5], Commerce Secretary Lutnick favors a sovereign wealth fund approach, and Treasury Secretary Bessent wants AI equity to seed Trump Accounts ^[2]. Steve Bannon went further, arguing the government should force AI firms to give up 50% of their equity ^[5].

The online reaction underscored the crossover: a prominent right-leaning tech figure, while professing no love for socialism or arbitrary confiscations of wealth, publicly conceded he could see why Sanders' proposal resonates, including with many on the right. The convergence is real but not identical: the Trump-side version is voluntary and passive (no board seats, no voting rights), while Sanders' is mandatory and explicitly activist. The Cato Institute's Tad DeHaven even argues Trump's February 2025 executive order opened the door Sanders is now walking through ^[3].

The proposal's scale is what makes it arresting. The headline figure is a roughly $7 trillion fund at current valuations, seeded by a one-time 50% tax paid in stock and triggered for any AI company with $200 million or more in annual AI gross receipts ^[1]. From that base the bill projects a 5% initial annual dividend, which Sanders' office translates into more than $1,000 per American per year ^[1]. The governance layer is a single 7-member Independent Commission for Democratic AI ^[1].

The math, however, leans on valuations that are still speculative: OpenAI alone is valued above $850 billion and reportedly eyeing an IPO as early as September 2026 ^[6], while the same firms underpinning the dividend math are not yet profitable ^[2].

The objections cluster into three families. First, feasibility: the $1,000 dividend rests on shares of companies that are not yet profitable, so converting equity into a reliable income stream is far from guaranteed ^[2]. Second, conflict of interest: Cato's Tad DeHaven warns that a government holding equity in firms it also regulates, taxes, and investigates undermines impartial oversight, comparing it to referees in the NBA Finals having a financial stake in the teams on the court ^[3]. He and the World Socialist Web Site both attack Sanders' Norway analogy from opposite directions, with WSWS noting Norway's fund is invested entirely abroad while its population owns and controls none of it ^[7].

Third, constitutionality: community debate on Reddit repeatedly raised Fifth Amendment takings concerns about confiscating private equity, a thread that ran alongside arguments over whether the plan amounts to nationalizing frontier labs. Reason's Tosin Akintola adds the innovation objection, arguing AI already benefits ordinary Americans and government control would slow breakthroughs ^[4].