Donald Trump’s America First approach to governing has produced plenty of unusual developments. But taking stakes in potential national champions in AI may be among the most curious.
Late last month, the Commerce Department announced $874 million in proposed funding for seven companies developing the memory, packaging, photonics, and materials needed to build faster AI systems. In return, Washington will receive a minority equity stake in each company.
The investments span several key parts of the AI supply chain. The Commerce Department is giving GlobalFoundries $300 million to bring co-packaged optics, which place light-based connections alongside AI processors, to market two or three years sooner. Kepler is slated to receive $245 million to develop a new kind of AI memory, while five smaller companies would share the remaining funding.
The deals attracted public attention, but they are not a one-off. In May, the Commerce Department offered more than $2 billion to nine quantum computing and manufacturing companies on the same terms. Last year, the government paid $8.9 billion for roughly 10% of Intel. Together, the investments amount to an industrial strategy suggesting that the Trump administration does not merely want to subsidize strategic industries. It wants to own part of them.
But is this the nationalization of AI, public-sector venture capital, or something else?
Chris Miller, author of Chip War, says last year’s Intel deal is “fairly sui generis” because Intel is both strategically vital and financially troubled, as well as the only American manufacturer of high-end chips. Allowing the company to fail would have been disastrous.
The smaller investments announced last month look more like a public-sector version of venture capital. “If taxpayer dollars will be invested in these companies, some of which will work, some of which will not, taxpayers deserve to be compensated in the upside of the ones that do work,” he says.
Washington is acting more like Silicon Valley in part because of the competitive threat posed by China. U.S. venture capital firms tend to favor software companies that can grow quickly and generate returns within a few years. They are often more reluctant to fund expensive factories, new materials, and hardware businesses that may take a decade to mature.
China’s state-guided capital, by contrast, is far more patient. It has helped the country’s national champions establish strong positions in batteries, electric vehicles, and drones. The Commerce Department is trying to close that gap without abandoning the market altogether, preserving the best of the market system without giving China a guaranteed advantage in capital intensive industries.
But ownership does not necessarily equal power, argues Todd Tucker, director of industrial policy and trade at the Roosevelt Institute. Equity stakes can make sense when markets fail or when democratic control is the goal. But the Trump administration “has generally kneecapped some of its own ability to control the decisions of the companies it invests in, by pledging for instance that it would not vote its shares against the wishes of management,” says Tucker.
He contrasts the latest chip deals with Washington’s golden share in U.S. Steel, which gives the government veto rights over more than 10 categories of corporate decisions. The chip investments, he argues, do little more than nationalize part of the companies’ balance sheets without granting the government meaningful decision-making power.
Matt Stoller, director of research at the American Economic Liberties Project, notes that the U.S. has historically funded strategically useful companies, typically through grants, loans, procurement, and research spending. But he is wary that this new approach appears to have few visible rules. “This is a new and untested form of regulation that is very embedded in the hands of the president,” Stoller says. While it could be beneficial, “there really aren’t any guardrails about what they’re doing.”
That matters when the government is simultaneously an investor, customer, and regulator. Traditional regulation can be infuriatingly slow, Stoller says, but it generally includes formal proposals, public comments, and judicial review. The fear is that this new model replaces that process with presidential cajoling and threats from the author of The Art of the Deal.
Stoller suggests that Washington’s pressure on major chip buyers to use Intel’s factories, thereby supporting another company the government has backed, could ultimately matter more than its ownership stake.
“There’s nothing inherently problematic about the government taking equity stakes,” he says. “The issue is the basic lack of clarity about what they’re doing, why they’re doing it, and who’s benefiting.”
Miller raises similar concerns. “We should want a clear ring-fencing between what is the government’s responsibility and what are companies’ responsibilities—and those should be as distinct and transparent as possible,” he says.
Without a coherent strategy, Stoller worries the arrangement could become deeply confusing for the government, the companies involved, and U.S. taxpayers. “I just think it’s a mix of reasonable, incoherent, and dangerous.”