We’ve spent the past year documenting the quiet crisis in federal research funding—watching as basic science budgets flatline while grant success rates plummet. In our lab, like labs across the country, we’ve felt the squeeze.
So when we started hearing about hundreds of billions flowing into AI infrastructure, framed as vital to national security, we got curious. Not cynical—curious. After all, the Space Race began with similar rhetoric about Soviet threats, and it ultimately delivered transformative investments in science, technology, and infrastructure that benefited society broadly.
Maybe this AI competition with China could do the same? Maybe the rising tide of defense-justified spending would lift all boats, including research?
We decided to follow the money. What we found doesn’t look much like the Space Race. It looks more like something else entirely.
In June, four Silicon Valley executives raised their right hands and took an oath to defend the Constitution. Palantir’s CTO, Meta’s CTO, and two OpenAI executives were commissioned as Army Reserve Lieutenant Colonels—a rank that typically takes nearly two decades to earn. No basic training. No physical fitness tests. Just instant military leadership for tech executives whose companies stand to profit enormously from an AI race.
“America wins when we unite the dynamism of American innovation with the military’s vital missions,” Palantir’s CTO Shyam Sankar declared. Sam Altman, the CEO of OpenAI, warns that “authoritarian governments are willing to spend enormous amounts of money to catch up and ultimately overtake us.” The narrative is clear: in the AI arms race, America must win at any cost.
But what if the war rhetoric itself is the real threat?
The Circular Money Machine
Follow the money, and a pattern emerges. In October, OpenAI announced deals with five major chipmakers worth hundreds of billions of dollars. Broadcom’s stock jumped nearly 10% on the announcement. AMD shares rose 3.7%. Microsoft, OpenAI’s primary backer, has pledged $80 billion in AI infrastructure spending for fiscal 2025. Nvidia, whose market cap hovers near $4.6 trillion, receives a significant portion of its revenue from OpenAI.
Here’s what makes this dangerous: each player’s success depends on the others. OpenAI needs the chips to build its technology. The chipmakers need OpenAI’s massive orders to justify their stock valuations. Microsoft needs OpenAI to succeed to justify its own $80 billion investment. The cloud providers need both. Remove any one piece, and the whole structure wobbles.
And all of them need the narrative of existential competition with China to keep the money flowing. The rhetoric of an AI arms race isn’t just metaphor; it’s a business model. Every new warning about Chinese AI dominance—every claim that we must spend whatever it takes or lose—inflates the same few balance sheets. The “national security” framing transforms what might otherwise look like speculative excess into patriotic necessity.
The result? Just seven companies—all heavily invested in AI—now account for 60% of the S&P 500’s gains. As Morgan Stanley notes, these same companies plus about 34 related AI firms are responsible for 75% of overall market returns, 80% of earnings growth, and a staggering 90% of capital spending growth since late 2022.
Think about what that means: nearly the entire stock market’s gains come from a handful of companies all betting on the same technology, all dependent on each other’s success, all relying on the same “beat China” narrative to justify ever-larger investments. Meanwhile, youth unemployment sits at 10.5%. Most Americans aren’t seeing benefits from this supposed technological revolution.
The Bubble
OpenAI projects needing at least $1 trillion in investment over the next decade to build the data centers necessary for its ambitions. Against this trillion-dollar need, the company expects to generate just $13 billion in revenue this year. That’s a 77:1 ratio of required investment to actual revenue.
The last time the U.S. economy depended this heavily on a single narrative (”the internet will change everything”) it crashed. This time, the narrative isn’t about freedom or connection. It’s about survival.
Former Biden administration economists Jared Bernstein and Ryan Cummings don’t mince words. In their October 2025 New York Times analysis titled “Warning: Our Stock Market Is Looking Like a Bubble,” they point out that Nvidia’s shares trade at 55 times earnings—nearly double the historical norm. Critically, “the share of the economy devoted to AI investment is nearly a third greater than the share of the economy devoted to internet-related investments” at the peak of the dot-com bubble.
We know how that ended.
“Should lackluster AI performance or sluggish adoption cause investors to doubt these lofty profit expectations, this probably-a-bubble will pop,” Bernstein and Cummings warn. Even Sam Altman has hedged. After years prophesying artificial general intelligence would arrive during Trump’s term, he suddenly declared in August that “AGI is not a useful concept.”
That’s a red flag. When the people selling the vision start backing away from it, pay attention.
Yet the “national security” framing makes questioning this overinvestment seem unpatriotic. Congress proposes $32 billion in discretionary spending for “AI innovation projects.” States offer massive incentives. The race is on! And anyone suggesting we slow down gets accused of handing victory to China.
The Trap
Here’s where war rhetoric becomes dangerous: it’s leading us into a three-pronged economic trap.
First, massive opportunity costs. Utilities are planning to spend $1.1 trillion over the next five years primarily to support AI data centers. We’re rebuilding America’s power grid for a handful of servers, not the people who pay the electric bill. What is more, these data centers employ almost no humans, so it will likely have minimal impact on employment.
Second, unprecedented concentration. Seven companies drive 60% of market gains while holding defense contracts, wearing military uniforms, and shaping procurement decisions. Palantir’s CTO is now literally a military officer advising on technology purchases from companies like Palantir. The conflicts of interest are more than normalized and systemic, at this point they are brazen. Circular investments and hinge points that rely on poorly diversified firms threaten to undermine the stability of these stock market gains. Like the subprime mortgage crisis, once one of these companies falters, the others might go with it like a house of cards.
Third, normalized risk. Markets have become “inured” to daily chaos. Century-high tariffs? Record attacks on federal institutions? No problem. AI stocks keep increasing, so everything must be fine. We’ve learned to treat chaos as background noise. As long as AI stocks rise, no one questions it.
That’s the quiet power of the war metaphor: it suspends disbelief. If this is a war, then waste is an acceptable strategy and questioning is treason.
Inverted Logic
Here’s the final irony: the war framing itself undermines what America actually needs for AI safety and economic security.
Thomas Friedman and former Microsoft research head Craig Mundie argued that “China and America don’t know it yet, but the artificial intelligence revolution is going to drive them closer together, not farther apart.” Not because of warm feelings, but because technical reality demands it.
Their reasoning is simple but profound: AI will be embedded in everything. Your hip replacement. Your pacemaker. Your car’s navigation system. Would you let a Chinese-made “smart hip” be sewn into you without assurance that China and America had agreed on common ethical standards?
Without trust frameworks, we face “digital autarky”—walled-off ecosystems where innovation dies. But there’s a darker scenario: AI in the hands of bad actors. “Without the United States and China agreeing on a trust architecture,” Friedman warns, AI will produce “super-empowered thieves, scam artists, hackers, drug dealers, terrorists and misinformation warriors” who will “destabilize both America and China, long before these two superpower nations get around to fighting a war with each other.”
The war rhetoric is preparing us for the wrong battle, while the real threats go unaddressed. When AI development is framed as zero-sum competition, cooperation becomes politically impossible—even when it’s technically necessary. You can’t simultaneously declare AI an existential race against China and then propose joint safety standards with them. The war framing forecloses the very diplomatic and technical collaboration needed to prevent the actual risks Friedman and Mundie identify: AI-powered crime, manipulation, and instability that threaten both nations equally. We’re spending hundreds of billions to “win” a competition while creating the conditions that make us both more vulnerable to the real dangers. The framework designed to mobilize resources has become the obstacle to deploying them effectively.
Who Asked for This War?
The public certainly isn’t clamoring for trillion-dollar AI infrastructure. Only 17% of Americans believe AI will have a positive impact over the next 20 years, according to Pew Research. Meanwhile, 51% are more concerned than excited about increased AI use in daily life. Even among scientists using AI tools, concerns about AI generating false information jumped from 51% in 2024 to 64% in 2025. The pattern is clear: the more people learn about how AI works, the less they trust it.
Yet it’s being marketed as an urgent matter of national security. This framing has been remarkably effective—so effective that when AI fails to deliver meaningful value, the prevailing assumption is that the implementation was flawed, not that the technology itself might be overpromised.
Meanwhile, the companies pushing war rhetoric have already won. They’ve secured military commissions for their executives, locked in circular investment deals worth hundreds of billions, and captured 60% of market gains while the broader economy struggles. They’ve built an economic structure where seven companies control most market gains, where trillion-dollar infrastructure investments rest on revenue projections that would make dot-com investors blush, and where the same people profiting from AI contracts wear military uniforms and advise on procurement. Question any of this, and you’re accused of surrendering to China.
There’s a historical precedent for this kind of spending. America won the Cold War by baiting the Soviet Union into an arms race it couldn’t afford. Reagan’s Strategic Defense Initiative—”Star Wars”—forced massive Soviet defense expenditures chasing a threat that was part technology, part theater. The USSR spent itself into collapse trying to keep pace.
The cruel irony: we’re now playing the Soviet role in our own staged arms race. Except this time, there’s no external adversary engineering our overextension. We’re doing it to ourselves. The “threat” is real enough to justify the spending, vague enough to justify any spending, and profitable enough that the people warning about it are the same ones cashing the checks.
We spent the Soviets into defeat by making them chase an existential threat. Tech executives have done the same to us—but they wear our uniforms, sit in our hearing rooms, and drape themselves in the flag while doing it.
The real question isn’t whether we’ll lose to China in a race for AI dominance. It’s whether we’ll economically overextend ourselves—tying up massive amounts of money that could fund healthcare, housing, or infrastructure in potentially unproductive technologies, creating dangerous concentrations of wealth and bankrupting everything else in the name of “winning.”
After all, the Soviets thought they were winning too. Right up until they couldn’t afford to anymore.
History rhymes
We’ve seen this story before. We’ve just never been on this side of it.
Back in our lab, we’re still writing grants, still fighting for funding, still making the case that basic research matters. But it’s hard not to notice the contrast: a medical researcher needs years of peer review to secure a $500,000 NIH grant. A tech executive can secure a military commission and billions in infrastructure commitments by declaring AI a national security priority.
The Space Race was supposed to be our model—existential competition that forced investment in science and technology, delivering benefits that outlasted the Cold War itself. Instead, we’re watching a different playbook: one where the threat rhetoric enriches a handful of companies, the infrastructure serves private interests, and the actual science—the kind that doesn’t promise immediate returns or military applications—gets crowded out.
Maybe we’ll look back and realize this was all necessary, that the investments were justified, that we really were racing China and had no choice. But from where we sit—watching federal research budgets stagnate while AI infrastructure gets blank checks, watching tech executives commission themselves into the military while scientists compete for scraps—it’s hard not to wonder if we’re financing the wrong war.
The Soviets spent themselves into collapse chasing Star Wars. We’re spending ourselves into concentration chasing something that might be equally illusory, but with one crucial difference: the money’s going to the people who are selling us the threat.
History may not repeat, but it rhymes. And right now, the rhyme scheme is sounding awfully familiar.