A hedge fund worth $45 billion at its height sold nearly its entire stock portfolio to Citadel Securities last week. As recently as a month ago, the fund was up 439 percent on the year, according to an investor letter from its founder, 24-year-old former OpenAI employee Leopold Aschenbrenner. But its portfolio, aggressively invested in companies tied to the artificial intelligence industry, dropped 67 percent in July.
Hilariously, the fund was called Situational Awareness.
Last week, the tech-heavy Nasdaq saw its second correction of the year, named for a drop of at least 10 percent from its peak. SpaceX has lost the equivalent of the entire value of Tesla since its post-IPO high in June, and it’s still dropping. For whatever reason—the rise of cheaper and more flexible Chinese AI models, the recognition that U.S. AI companies simply aren’t generating enough revenue to justify skyrocketing capital expenditures, the general economic drag from Trump’s tariffs and wars, or the increasingly operatic financial maneuvers to keep the wheels moving—the shine is way off the AI rose for investors.
The problem is that the industry is bound so tightly with the stock market that a change in feeling from AI investors could be all it takes to generate a market-wide crash, as we’re seeing to some degree. In other words, if AI is propping up the economy, who is propping up AI?
The answer, extrapolating from a fascinating new paper about private credit and the life insurance industry, could be the U.S. taxpayer.
Private credit is private equity’s $3 trillion financing arm. They make largely unregulated, relatively high-risk, relatively complex and opaque loans, mostly to their own portfolio companies. Private credit is entangled, maybe more than the rest of Wall Street, in AI mania. For example, in the mid-2010s private equity bought up hundreds of software-as-a-service providers, and its private credit affiliates made thousands of loans to them, only to see these portfolio companies buckle recently as AI replicated their tasks.
Instead of passing through bankruptcy, insolvent life insurers have all their liabilities—in particular the policyholder claims—paid for by state guaranty funds.
A deeper intertwining can be seen with all the loans private credit companies have made to build AI data centers across the country. As delays in construction grow—some 530 local jurisdictions and counting have passed data center restrictions or bans—and as the AI industry seems less of a solid bet, these loans could go into mass default. Higher interest rates, thanks to the disastrous war in Iran keeping inflation high, have also become a private credit stress point. As a result, redemption requests from nervous investors continue to grow.
Yet none of this has stopped the private credit giants, even as share prices have crumbled. One of the biggest, KKR, recently reported record earnings. What explains this disconnect? It could be linked to a separate business line that has given many private equity firms what consulting giant McKinsey has described as “permanent capital.”
Private equity today owns at least $1.5 trillion in assets in life insurance companies. Apollo bought Athene in 2022; KKR got Global Atlantic a couple of years earlier. As explained in a research paper by Andrew Granato, an assistant professor at the University of Texas at Austin, and Pranjal Drall, a Ph.D. candidate at Yale, these life insurers have mounds of available capital from policyholder payments that don’t need to be paid out until the end of their lives. Private equity firms have plowed this capital into risky private credit loans that could weaken the structure of the life insurer. But if the insurer goes insolvent, the private equity firm won’t have to pay; you will.
That’s because instead of passing through bankruptcy, insolvent life insurers have all their liabilities—in particular the policyholder claims—paid for by state guaranty funds. After the fact, those funds impose assessments on surviving life insurers in a particular state. The apparent idea was to protect life insurance beneficiaries, and to pay for it through fees on the industry as a whole. But the choice of policy design means the successful companies pay for the bad behavior of the insolvent ones. And it gets worse in most of the country: In 44 states, life insurers can take a tax credit for that assessment and get paid back 100 percent over a period of time.
In other words, the risks being taken by private credit companies ultimately transfer to state taxpayers. It’s a familiar story of privatizing the profits and socializing the risks. Private equity scented that opportunity a long way off.
“These incentives for insurers to take on this amount of risk have existed for a long time,” said Granato in an interview. “But private equity’s entry into the business marks this really sharp shift in the asset management profiles of their balance sheets.” Private equity is capitalizing on the bailout guarantee scheme by playing very close to the edge, and threatening the entire economy alongside it.
Private equity profits greatly from its relationship with life insurers. Granato and Drall find that after a life insurance company is bought by a private equity firm, it typically cuts its internal investment team in half, and drops that team entirely in one-third of the cases. Instead, the life insurer pays an affiliate of the private equity firm to invest for them. Numerous other fees for accounting, data systems, and other services funnel up to the private equity parent.
Life insurance capital is invariably used to purchase underperforming private credit loans, rather than the boring mix of Treasurys and corporate bonds that usually make up the investments. “In 2024, 49.5 percent of new private equity-owned insurers’ investment was in privately-placed instruments,” Granato and Prall write. For nonaffiliated life insurers, that number was around 14 percent.
The captive life insurer, in other words, is leaking out more money to its private equity owner and taking much higher risk on its balance sheet. No policyholder really knows this is happening and isn’t used to checking the holdings of its life insurer. “The consumers might actually get a better price,” said Drall, noting that higher risk often comes with a higher return, at least initially. “You might imagine that consumers are even happy to buy Athene.”
The threat, of course, is that all these bad private credit loans are being dumped into a part of the system that is not being closely monitored, by customers or regulators.
Insurance products are regulated at the state level by overmatched and often captured insurance commissioners. They don’t have the expertise to understand the complex loans coming into the life insurer, and they don’t really want to know anyway. “What ends up happening is that state regulators are actually competing with each other to loosen life insurance regulation,” Granato said. In addition, private equity firms pay lenient rating agencies to give their sketchy loans a clean bill of health, which is quickly accepted by the insurance commissioner.
Granato and Drall estimate in their paper that private credit loan default rates above 15 percent could generate insolvency. The subscription-based Capitol Forum noted last week that KKR and Apollo’s three life insurers have just a tiny amount of room in its hardest-to-value assets to absorb losses. “There is an expectation, even by the regulators, that the music is going to stop,” one insurance actuary told the publication.
A major life insurer insolvency is uncharted territory. “We’ve been very lucky, I think, that there’s been no major insurer insolvencies that have had to be mediated through the guaranty fund system, because there’s kind of an open question as to if this is even logistically possible to do,” Granato said. He noted that AIG, the insurer that collapsed after the 2008 crisis, was saved by quick action by the Fed and later Congress. “If AIG had been allowed to go under and then we would have had to manage 50 simultaneous AIG insolvencies across every state, in the middle of the financial crisis?” he asked. “You can imagine what a nightmare that would be.”
It’s unclear to Granato and Drall how we even ended up with this state guaranty/tax credit system; the historical literature shows shockingly little debate. But while protecting policyholders from ruin if their insurers made bad investments is certainly a good goal, doing it with after-the-fact assessments that get paid out by state tax credits is just a backdoor bailout. And it isn’t a particularly good disciplinary tool for the industry, as the insolvent insurer is off the hook from paying anything.
“It makes sense if you had a good supervisory regime,” Drall said. But the weak regime means that risk-taking is incentivized, especially if a private equity fund can extract from the insurer along the way and be held harmless afterward.
And when you bring in the fact that private credit is making all these data center loans, this risk-taking is dangerously concentrated. “Based on the fact that private equity firms specialize in this kind of lending, just the whole financial system is more and more aligned on that kind of activity,” said Drall. “And the more you do that, you know, AI better work out to sort of support the ecosystem.”
Granato and Drall offer several reform proposals, mostly around making risks more transparent by reforming the valuation of private credit (including penalizing the companies for their opacity and complexity) and scrutinizing the rating agencies that claim the assets are safe. Internal transactions between private credit and the life insurer should have enhanced review, they argue, and the fees private equity takes out of the insurer should be identified in risk review.
But their big ask is to change how guaranty funds assess the industry after the fact instead of before, and how the tax credits effectively put the risk on taxpayers. An alternative would be assessing all companies before the fact, and assessing their corporate parents—including the private equity firms. “The core problem is having the ex post design in the first place, rather than funding it the way federal deposit insurance does it,” Granato said.
Otherwise, there’s this whole super-important industry that’s precariously tied to the AI bubble, and if the bet fails, we’d all be stuck holding the bag.
