Should The Government Take Stakes In Frontier Labs? The Best and Worst Idea in AI
It depends on the incentive structure and the wisdom of our leaders.
Happy Sunday and welcome to Investing in AI. Housekeeping notes:
Be sure to check out our AI in NYC podcast for my take, along with my co-hosts, on the AI news each week.
Also check out our first public case study of SLMs crushing frontier models. For this use case it was a 20 point improvement in performance for 1/20th the price.
Today I want to talk about the news that the US government may take a stake in AI frontier labs. I kinda like the idea but, am cautious because of something I learned in high school.
In the 10th grade at Manual High School in Louisville, KY, I had took a world history class and I remember the teacher saying something that surprised me. She said “a monarchy is actually a very efficient form of government, provided you have a wise and just monarch, but that is rarely the case.” It’s true. Unfortunately the lesson of history is that Lord Acton’s statement that “power corrupts” is too true. Limited government, for all its flaws, is a check on power to avoid this very thing.
When I read about the Trump administration’s idea to take stakes in frontier labs, I admit I saw some appeal to it. It could solve many of the problems we worry about in AI, and it could also generate substantial wealth for US taxpayers. Or, the incentives could cause us to lose the AI race, prop up failing companies, and make bad choices.
The great thing about the United States is - its probably the only place in the world where you could walk into a room full of investors, tell them you are going to take down Anthropic and OpenAI, and you might actually get funded. Most investors would still pass but, there are enough investors out there who would love that shot, you would get money if your approach was sound. The constant lure of disruption makes America the entrepreneurial engine of the world. Do we lose that if the government owns stakes in the status quo labs? Here’s my (gemini supported) analysis:
How It Would Actually Work
The mechanics being floated are more nuanced than outright nationalization. Rather than the government writing checks for equity, the framework envisions tech companies trading stakes—potentially to a newly created “Public Wealth Fund”—in exchange for tangible government-provided assets: federally owned land for data centers, accelerated energy permits, access to the power grid at scale, and national security clearances that unlock defense and intelligence contracts.
Think of it less as the government buying stock and more as a structured barter: the state supplies irreplaceable infrastructure inputs; frontier labs supply ownership upside. The American taxpayer becomes, in effect, a passive limited partner in the AI revolution—not directing operations, but sharing in the returns. On paper, it’s elegant. It mirrors how Norway’s sovereign wealth fund turned oil reserves into generational national wealth. The question is whether AI equity behaves more like North Sea oil or like a venture bet in a rapidly shifting market where today’s frontier model is tomorrow’s commodity.
The Case For: Why This Idea Has Real Teeth
Wealth Generation at a National Scale
The most compelling argument is straightforward: the public is already subsidizing the AI buildout; it should share in the upside. The federal government is a massive provider of the inputs that make frontier AI possible—the power infrastructure, research grants flowing from DARPA and the NSF, the broadband and telecommunications backbone, and the defense contracts that provide early revenue to many of these companies. If OpenAI or Anthropic becomes the defining platform of the next economy, and the government helped make that possible, why should 100% of that equity appreciation flow only to private investors?
A well-structured sovereign wealth fund could generate meaningful returns. If even a modest stake in two or three frontier labs appreciates 10x over a decade—a conservative scenario given current trajectories—that fund could finance infrastructure, reduce deficits, or theoretically distribute dividends to citizens in the way Alaska distributes oil revenue. The concept isn’t fringe: Singapore, Norway, and the UAE have used sovereign funds to transform national balance sheets. Why should America be the only superpower that builds the infrastructure for a revolution and then hands the equity entirely to private parties?
Geopolitical Alignment at a Critical Moment
The second argument is harder to dismiss: we are in a race, and the adversaries aren’t playing by free-market rules. China’s AI development is state-directed, state-funded, and state-prioritized. The idea that the U.S. can win a strategic technology competition while operating on pure laissez-faire principles—while Beijing coordinates industrial policy with precision—deserves real scrutiny.
Government equity stakes would create structural alignment between national security objectives and the roadmaps of the labs developing the most powerful AI systems in the world. That alignment has real value. It could mean faster security clearances, better intelligence sharing, and a shared incentive to keep critical model weights and training data out of adversarial hands. In a world where the most important AI breakthroughs are as strategically sensitive as nuclear technology was in 1945, there’s an argument that the government shouldn’t just be a regulator standing on the outside—it should have skin in the game.
Making Taxpayers Whole on Infrastructure
The energy math alone makes a compelling case. Frontier AI training runs are consuming power at a scale that is reshaping utility planning across the American Southwest and Southeast. New data centers require water, land, grid upgrades, and permitting that flows through state and federal agencies. Taxpayers are bearing real costs and real externalities. An equity stake is one of the few mechanisms that creates a direct, proportional financial return for that public contribution—rather than simply hoping that jobs and tax revenue trickle back eventually.
The Case Against: Where This Gets Dangerous Fast
Regulatory Capture on Steroids
Here’s the problem my 10th-grade history teacher would have spotted immediately: give the government a financial stake in the success of a company, and you have fundamentally compromised its ability to regulate that company. This isn’t a hypothetical. It’s the oldest story in political economy.
If the Treasury holds equity in OpenAI:
What happens when the FTC wants to investigate OpenAI for anticompetitive behavior?
What happens when the AI Safety Institute wants to mandate costly evaluations before a new model ships?
What happens when Congress considers privacy legislation that would crimp the data practices these companies depend on?
Every one of those decisions now carries a financial conflict of interest for the federal government. The regulator becomes the investor. The entity tasked with protecting the public from AI risk has a direct financial incentive to let the AI companies move fast and win.
This isn’t speculative—it’s the logic of incentives. And it’s especially dangerous in AI, where the safety questions are genuinely unsettled and the potential downside risks are civilizational in scale.
Picking Winners Kills the Game
The second risk is what I’d call the incumbent moat problem. The thing that makes American AI competitive isn’t OpenAI or Anthropic specifically—it’s the system that could produce the next OpenAI or Anthropic. Government equity stakes in today’s frontier labs would, almost inevitably, create structural barriers to the companies that want to replace them.
A startup trying to build a challenger model doesn’t just face capital disadvantages—it now faces a competitor that has the government as a structural ally, with preferred access to energy permits, federal land, and security contracts. Open-source AI projects, which represent perhaps the most democratizing force in the space, would be competing against a model that has the full weight of state infrastructure behind it. We’d be trading the ecosystem that makes American AI dynamic for the short-term comfort of owning a stake in the current leaders—who may not be the leaders in five years.
The State Capitalism Trap
Finally, there’s the ideological question that shouldn’t be dismissed just because it sounds abstract: this is a meaningful departure from the model that created American technological supremacy. The U.S. didn’t dominate the internet era because the government owned stakes in Netscape, Google, or Amazon. It dominated because capital was free to flow toward the best ideas, including the ideas that destroyed the previous winners.
State capitalism—where government equity stakes distort the competitive landscape in favor of chosen champions—has a mixed record at best. It produces national champions that win politically long after they’ve stopped winning economically. It creates zombie companies propped up by strategic importance rather than actual performance. And it shifts decision-making from markets, which are ruthless about failure, to bureaucracies, which are not.
Summary: A Great Idea That Requires a Wisdom We Rarely Have
My 10th-grade teacher was right about monarchies. A government equity stake in frontier AI could, in theory, be an elegant solution: aligning national interests, generating public wealth, and ensuring taxpayers share in the revolution their infrastructure helped build. The logic is sound. The precedents exist.
But it requires exactly the kind of wise and just stewardship that history tells us is the exception, not the rule. It requires a government capable of holding equity without distorting regulation—of being a passive investor while remaining an aggressive watchdog. It requires bureaucracies that can resist the gravitational pull of protecting their portfolio companies from competitive disruption.
That’s a very high bar. And the downside of getting it wrong isn’t just a bad investment. It’s an AI industry that stops producing the disruptive challengers that keep America ahead—replaced by a small oligopoly of government-backed incumbents, insulated from competition, too big and too connected to fail.
The question worth sitting with is this: Are we confident enough in the wisdom and restraint of the American government to hand it a financial stake in the most powerful technology ever built—and trust it not to use that stake to protect its investment at the expense of everything else?
History suggests we should answer that question very carefully before we call it a beautiful thing.
Thanks for reading.
Governments have a percentage of all future revenue. It's called taxes.