WASHINGTON — The Pentagon could pick up technology and talent at bargain prices if soaring investments in artificial intelligence finally come crashing down in 2026, experts told Breaking Defense.
That doesn’t mean everyone’s expecting an AI-induced economic apocalypse. Indeed, there was lively disagreement amongst the three experts interviewed for this forecast about how bad any coming bust might be. But all three agreed that a return to reality was inevitable for at least the most heavily hyped companies.
“All the evidence suggests the bubble is real,” said retired Lt. Gen. Jack Shanahan, the former Air Force three-star who stood up the Pentagon’s Joint AI Center (JAIC, since reorganized, twice). “AI input spending is such a high ratio of current GDP, with an output that is still nowhere close to what’s desired or needed[:] return on investment, total productivity growth, etc.”
[This article is one of many in a series in which Breaking Defense reporters look back on the most significant (and entertaining) news stories of 2025 and look forward to what 2026 may hold.]
Shanahan said he’s especially worried about “circular funding,” where one tech company invests in another that then invests or pays that money right back — a scenario uncomfortably like the Leslie Nielson “spot me a twenty?” bit in the original Naked Gun.
“Everyone is becoming so intertwined and interdependent that you have to wonder if a seemingly minor symptom in one isolated part of the AI ecosystem generates widespread ripple effects everywhere, not only within the tech companies, but across the rest of society,” he told Breaking Defense.
However, he went on, “for the DoD, there could be new opportunities to partner with all kinds of tech companies that might suffer from the ripple effects of a bubble. If their commercial opportunities start to evaporate and they risk going bust, having a reliable source of funding from the government could entice a lot of CEOs who might otherwise think twice.”
Shanahan’s successor as Joint AI director, retired Marine Lt. Gen. Michael Groen, was more optimistic about the fundamental economics of AI.
“As we saw with Nvidia and DeepSeek, spooking the herd leads to outsized reactions in the markets,” Groen told Breaking Defense. He doesn’t even think the word “bubble” is a fair description of the AI sector today, at least “as of yet.”
“AI infrastructure will remain dynamic, meaning not every model will succeed over time,” he said. “That does not represent a bubble so much as acknowledging the dynamic environment of change and growth.”
That said, “a shakeout among AI vendors may yet flare up with winners and losers,” he cautioned — if only because, he said, “competition is really only beginning in the marketplace.” This isn’t so much a bubble bursting, in his view, but the normal “fits and starts” of a novel technology as economic realities reassert themselves.
Groen’s preferred analogy was the boom-and-bust in telecoms and dot-coms in the 1990s, which led to lasting technological and economic gains overall, even if many companies didn’t make it.
“It is like the construction of the internet,” he said. “Individual companies misjudged the moment, but enough companies stayed the course, were successful, and changed history. We find ourselves in that moment again.”
Our third expert, CNAS scholar Joshua Wallin, effectively split the difference. The former Pentagon fellow didn’t predict disaster, but he does expect 2026 will at least cull the AI herd. “Several companies seem to be offering similar services, especially focused on productivity enhancement in the commercial sector,” he said. “If anything, I would expect to see a winnowing of those down.”
The danger comes, he said, if a few key failures set off a chain reaction. “If there is damage to leading companies, it will cascade because most of the sector can’t afford to train their own foundation models,” he said. (A lot of companies’ AI products rely on or even merely repackage the mega-models of a few key players like OpenAI). “If you’re a business that uses a service provided by a major player and you lose access to that,” he said, “your performance will suffer.”
Wallin didn’t see a lot of cheap tech waiting to be picked up in the wreckage, he said, because AI software and hardware rapidly become obsolete without regular upgrades. Instead, he argued, the real treasures worth salvaging are human.
“The most valuable asset may be talent,” Wallin said. “If there are big layoffs of scientists and engineers in the wake of a bubble bursting, that talent could be valuable for defense players that are sorely in need of expertise.”
However bad the damage, whether disastrous crush or modest correction, the experts agreed that the Pentagon needs to position itself carefully to make the best of it. That includes having the fast-twitch flexibility to hire orphaned AI talent before it’s snapped up elsewhere and contract with ailing AI companies before they die.
Shanahan sees grounds for optimism here in Defense Secretary Pete Hegseth’s push to loosen up the notoriously sclerotic procurement bureaucracy.
“The Secretary has paved the way — at least in terms of guidance, policies, and memos — for accelerating acquisition and adoption of AI and commercial software across the Department,” he said. “[Now] the DoD must show that it is a reliable, credible, serious partner who can put companies on contract rapidly and show the ability to scale way beyond pilot projects and demos” — not just for a few favored companies, he added, but for the tech sector writ large.
Bridging the divide between the Pentagon and Silicon Valley will cultural change on both sides, Groen said.
“The greatest danger might be the failure to … leverage both warfighting experience and technological proficiency,” he said. “Today, I observe ‘tech-bros’ who know little about the real challenges of warfighting, and military ‘experts’ who do not understand tech application.”
But getting those warfighters and technologists on the same page would assure America kept its edge over rivals (such as China), he said: “The US has the best of each and can readily win by better teaming.”