AI as a Public Service: The Debate No One Wants to Have, But Everyone Needs To

There is something poetically cyclical about Senator Bernie Sanders proposing that the American public receive direct dividends from artificial intelligence companies. According to Tom's Hardware, Sanders introduced a bill suggesting that 50% of the U.S. AI giants be publicly owned, distributing about $1,000 annually per citizen. The curious detail? The response came not only from the expected left, but from the right itself: Vice President J.D. Vance indicated that the Trump administration would support giving the American people a stake in these companies, though he prefers the "pre-distribution" of opportunities rather than handing out cash. When left and right agree that something needs to be taxed or redistributed, you know the business model in question has hit a structural limit.

The market's knee-jerk reaction will be to cry "nationalization." But that misses the point. What is at stake is not the ownership of a recommendation algorithm or a chatbot. It is the computing infrastructure that underpins them. We are talking about massive data centers, dedicated power grids, and priority access to cutting-edge semiconductors. Generative AI is no longer just a consumer product; it is the computational foundation of the next decade. Treating it like an ordinary software startup is an anachronism.

Tech capitalism has always operated under the premise that private risk buys private reward. If you fail, the loss is yours; if you build the next Google, the profit is yours. The problem is that the scale of capital and natural resources required to train frontier models have made this equation unsustainable. When a company needs dedicated nuclear power plants to run its servers, it has crossed the fine line separating a private business from a public service.

The proposals from Sanders and Vance, each in its own way, are clumsy attempts to solve a real problem: how to capture public value from infrastructure that was built with private capital but relies on common resources—from the electromagnetic spectrum to the national power grid. The idea of giving the public "a stake" is not necessarily about social justice; it is about recognizing that AI has become a strategic asset, akin to railroads in the 19th century or telecommunications in the 20th.

Opponents of these ideas will argue that any form of public ownership or dividend redistribution will kill innovation. It is a valid argument, but an incomplete one. AI companies already operate in a de facto state capitalism regime: federal subsidies, military contracts, and privileged access to infrastructure. The question is not whether the government will be involved, but on what terms. If the public assumes the systemic risk—whether by funding basic research or bearing the impact on the labor market—it makes sense for them to have a seat at the dividend table.

The debate over the ownership of AI companies is not going away. It will escalate. As the technology becomes more ubiquitous and the concentration of computing power more acute, the pressure for some form of redistribution—whether through taxes or public equity—will increase. The question is not whether we will redraw the boundaries between private capital and public interest in the AI era, but whether we will do so deliberately or reactively. For now, we are just making noise.