The Sanders AI Sovereign Wealth Fund Act Is a Death Sentence for American Technology Leadership

What would be the quickest way to destroy America’s global leadership position in artificial intelligence (AI) and advanced computation and undermine life-enriching innovation in the process?

Sen. Bernie Sanders (I-Vt.) appears to have the answer. He recently introduced the “American A.I. Sovereign Wealth Fund Act,” a measure that would give the federal government significant control over computation in America. The bill represents the most radical and potentially destructive piece of technology legislation ever proposed. It would politicize the American technology industry, upend markets, deter innovation and job creation, and hand China the lead in the race for global supremacy in AI and advanced computation.

It might be tempting to dismiss the proposal as another wild-eyed idea from Sanders, who has well-known socialist views. Unfortunately, however, other versions of this idea are being considered. “The idea of giving the public a stake in AI has recently drawn interest from figures as ideologically diverse as President Donald Trump and OpenAI CEO Sam Altman,” notes The Associated Press. A recent R Street essay identified the general problems with these variations of AI quasi-socialization. The Sanders bill potentially softens the ground for the advancement of these proposals, all of which should be rejected.

Government Control of Computation

The new Sanders bill proposes a 50 percent tax on firms with $200 million or more in annual gross receipts that operate data centers, computing infrastructure, AI services, and advanced robotics. The money would be used to create a new $7 trillion sovereign wealth fund that would make direct payments of at least $1,000 to every citizen.

While the bill would not nationalize AI outright, it would give the government considerable control over firms involved in advanced computation. To deal with the problem of firms that have both AI and non-AI businesses, the measure requires structural separation of their AI business, such that the government could control the AI side of those firms.

This move would be massively destabilizing to markets and would sever the internal cross-subsidies that those firms use to invest in advanced computation systems and facilities. In some cases, it is not even clear how such a separation would work because their systems are not neatly compartmentalized; “AI” is baked into almost all their services now. The reach of this measure would extend beyond leading AI labs and hit firms like Dell, IBM, Tesla, Nvidia, Waymo, and many others.

For firms like OpenAI and Anthropic, which are more exclusively focused on AI model development, the tax would hit them directly. These firms remain unprofitable, however. While their revenues are growing, their R&D expenditures are growing faster. According to recently leaked financial reports, OpenAI suffered net operating losses of 237 percent in 2024 and 160 percent in 2025. Anthropic is struggling to achieve profitability.

These leading AI labs are also currently engaged in a price-cutting battle to compete with each other and also ensure that cheaper Chinese open source AI systems do not undercut them in global markets. In sum, they currently do not have any profits to share. But the Sanders bill instead levies the tax on their gross receipts, which would mean that the AI firms would lose their ability to reinvest that money into the further development of AI systems in the hope of eventually becoming profitable enterprises and meeting the competition domestically and globally.

Soviet-Style Thinking

The Sanders bill would also significantly politicize the AI ecosystem by putting DC bureaucrats in a position to heavily influence market decisions. The new wealth-sharing fund would be managed by the “Independent Commission for Democratic AI,” which would include seven members representing labor interests, large fund management, AI safety concerns, privacy interests, and public safety interests.

Sanders, who famously spent “a very strange honeymoon” with his wife in the Soviet Union in 1988, seems to have borrowed some ideas from communist central planners when designing this bill. During the Cold War, the Soviet Union and other Eastern Bloc states created ministries and planning bodies for all major industrial sectors thought to be of strategic significance to the state. Communist party members sat on all planning bodies and made strict demands of all affected sectors. The result was massive political corruption and economic misery.

One important distinction, however, is that Soviet apparatchiks very much wanted to ensure the robust development of the sectors and technologies under their control. That does not seem to be Sanders’s primary goal with this new AI socialization effort. In fact, in late March, Sanders and Rep. Alexandria Ocasio-Cortez (D-N.Y.) introduced the “AI Data Center Moratorium Act,” which proposes a federal ban on data centers until multiple laws are passed that ensure AI systems “are safe and effective.” This would create a regulatory standard for AI that will be impossible to satisfy, effectively blocking further development. It is strange that Sanders believes his new AI sovereign wealth fund will generate an estimated $7 trillion when he is simultaneously looking to block the very engine of AI development.

While Sanders is motivated by a desire to share the wealth, it is important to remember that private AI innovators are already providing America with “a massive private sector stimulus program,” which many economists argue is the fundamental driver of economic growth today. In January, the Federal Reserve Bank of St. Louis reported that, “recent investments in AI-related categories have contributed significantly to the real GDP growth in 2025,” and the current AI boom, “has surpassed the contribution of IT components to the real GDP growth made during the dot-com boom, both in levels and as a share of GDP.” They project that these AI firms “are likely to remain significant drivers of investment well into 2026 and beyond.”

Moral Hazard Dangers

While these AI firms and their workers and investors are making massive bets on the future, there is no guarantee their systems or approaches will pan out. Some skeptics believe the current AI investment book is just a massive bubble that could pop at any moment.

This points to a different problem with the Sanders bill and the idea of the government taking stakes in AI firms because it means the public might get left footing the cost of failing companies the government partially controls. The government’s track record in managing private enterprise is dismal. Even partial public ownership distorts the incentives of companies both through weakening market discipline and as political considerations inevitably come to predominate in their operation. While Sanders claims his fund would not bail out failing firms, a government stake in AI companies sends a signal to market participants that, should the company get into financial trouble, the government will likely step in to protect its interests.

If the expectation of a bailout is credible, these participants—lenders, shareholders, financial analysts—will spend less time and effort monitoring company management and its financial risk. This provides greater leeway for management to make riskier decisions, taking on more debt, making riskier investments, or maintaining unprofitable lines of business, keeping the upside for themselves, while losses are absorbed by the government (i.e., taxpayers). This “moral hazard” problem has been observed in a variety of settings, from banking to state-owned enterprises in developing and Eastern Bloc countries to the various government-sponsored enterprises (GSEs) here in the United States. Indeed, the case of the GSEs is instructive. Much like Sanders’ proposal, the federal government sought to address the bailout expectations among investors in GSE debt by explicitly stating on their debt that they are not obligations of the United States. Nevertheless, the government has repeatedly intervened to assist these enterprises whenever they have encountered financial difficulties.

In theory, moral hazard could be addressed through oversight by the proposed commission; however, the effectiveness of prudential regulation is questionable. In the first place, regulators tend to prefer a one-size-fits-all approach to measuring and constraining risk-taking in a regulated industry. While such an approach may reduce the burden of regulation for both bureaucrats and regulated firms alike, it nevertheless leaves margins for firms to exploit gaps and engage in “regulatory innovation” that ultimately leaves their overall level of risk unchanged.

Petrostate Politics

In addition to the difficulty of regulation, there is an inherent conflict in tasking regulatory agencies with both constraining the risk-taking behavior of an industry while simultaneously ensuring that it delivers the financial performance required to meet political goals. However, whereas the Civil Aeronautics Board, Interstate Commerce Commission, or other previous regulatory bodies depended on their particular industry’s performance to maintain their political legitimacy, the proposed AI commission would be on the hook to meet a set dividend to voters on a recurrent basis.

In the AI commission’s case, regulatory actions that address excessive risk-taking on the part of firms may nevertheless harm their performance and thus affect the fund’s ability to provide the benefits promised by its political sponsors. From the standpoint of an incumbent president, they would face a strong incentive to ensure that companies achieve the financial performance to at least deliver the dividends promised to voters. Indeed, the incentive to obtain or exceed this performance would become more acute as the election approached, potentially resulting in officials pushing companies to pursue short-sighted or unsustainable policies, the risks of which would likely become a liability for taxpayers.

While sovereign wealth funds (SWFs) come in many varieties, studies have identified several of their common failings, including a lack of transparency and accountability and a tendency to suffer from political pressures that “favor(s) short-term economic policy goals… at the expense of longer-term maximization of returns.” Most SWFs are formed around revenues derived from natural resources like oil that are abundant in some nations. AI systems are nothing like this and, as one analyst notes, it would be dangerous to apply a petrostate model to the computational economy because the resources in this case would be “pried out of a working economy rather than pumped out of the ground.” AI systems are not just there for the taking; they must be created. Government efforts to seize or control them will discourage further development.

Say No to Central Control of AI

While the Sanders bill is unlikely to attract widespread support due to its highly destructive directives, even variants of these ideas should give policymakers pause. The AI market is still developing, and there is no guarantee that any particular models or systems will succeed. Many will fail. Preemptive schemes to have Washington take greater control over this sector through ownership shares and wealth-sharing arrangements would only make that scenario a self-fulfilling prophecy, but then also leave taxpayers potentially picking up the tab. Lawmakers should reject the seizure of private property and centralized control schemes like the Sanders bill. It would represent the demise of American technology leadership on the global stage.