Quick Answer: An AI bubble burst would happen if the companies spending heavily on AI infrastructure cannot earn back their debts. A 2026 report from the Open Markets Institute argues the ingredients for that crisis are already in place: AI spending far outpaces AI revenue, speculative data center debt is being sold to pension funds, and private equity is loading insurers with risky debt. The report does not predict a date, but it says regulators should act now.
The biggest companies in the U.S. are now mostly tech giants, and most of them are pouring money into data centers. Here's the uncomfortable part: they're spending far more on AI infrastructure than they earn from AI products. A new report from the Open Markets Institute argues that this gap is now being filled with borrowed money, and that the risk is spreading well beyond Silicon Valley. If you have a retirement account, a pension, or an insurance policy, this affects you more than you might think.
What the Report Actually Says
The report, written by Matthew Scherer and published in October 2026, is blunt about the scale of the problem. Alphabet, Amazon, Meta, Microsoft, Oracle, and SpaceX are all at the center of what the authors call a multi-trillion-dollar spending spree on data centers and related infrastructure. Training and running generative AI models is expensive, and the revenue from those models hasn't kept pace.
The authors don't claim a crash is certain. They say explicitly that the report's purpose is "not to predict a specific sequence of events or portray an economic crisis as inevitable." Their argument is narrower: the ingredients for a serious crisis are present, and waiting to see what happens is the wrong response. That distinction matters. You can take the risk seriously without buying the most dramatic version of the story.
The core mechanism is simple. Companies that borrow to fund infrastructure need future revenue to repay it. The report says AI giants will need to earn "multiples more" in revenue than they earn today to cover their spending. If that revenue doesn't arrive on schedule, the debt becomes the problem.
Where "Circular Financing" Fits In
The headline uses the phrase "circular financing," which is widely used in coverage of AI deals. The general idea is that money loops between a small group of firms. A chip supplier or cloud provider invests in an AI startup, and that startup then uses the money to buy the investor's chips or cloud capacity. The revenue looks real on the books of both sides, but a large share of it may come from the same pool of capital.
Here's why this matters for a bubble. Circular arrangements make demand look stronger than it is, because the customer's spending is partly funded by the vendor. When growth numbers rely on that loop, a slowdown at one node can hit the others quickly. Most people stop reading at this point, but the subtle risk is that no single firm has to fail for the whole chain to feel the pressure.
One caveat: the Open Markets report focuses more on debt flows and institutional exposure than on the specific deal structures. Treat "circular financing" as a useful lens for asking where revenue comes from, not as a proven measure of how much of today's AI demand is artificial. The honest answer is that the disclosure is too thin to know yet.
Speculative Data Center Debt Is Reaching Pensions
This is the section that should make you pay attention. According to the report, speculative data center debt is being repackaged and sold to pension and retirement funds, university endowments, and other institutional investors. The authors describe this as "eerily reminiscent of the subprime bubble."
The comparison deserves a precise reading. Subprime mortgages were bundled into securities that were then sold widely, so the risk was spread across many buyers who didn't fully understand the underlying loans. The worry here is a similar pattern: infrastructure loans tied to future AI revenue get packaged and sold to institutions that expect steady returns.
In practice, a pension fund buying this kind of paper is betting that data centers will generate cash flow for decades. If AI revenue lags the capital spending, those assets may be worth less than their face value. Your retirement savings could absorb that loss without you ever having chosen to take the risk. That's the core concern.
Here's where it gets interesting for regulators. The report argues that current disclosure rules make it hard to see how much of this debt sits inside a given fund. If you're a plan participant, ask your plan administrator directly what private credit and infrastructure exposures it holds.
Private Equity, Life Insurers, and Contagion
The second channel the report highlights is insurance. Private equity firms have bought life insurers and loaded them with risky debt. Policyholders are the ones exposed, because an insurer that can't meet its obligations doesn't just lose shareholder money. It can fail to pay out to the people who bought policies for security.
This is a classic contagion pattern. Life insurers hold large bond portfolios, so if those portfolios include speculative AI-linked debt, a sell-off can force losses across the insurance sector. The report warns this could "spread well beyond the tech sector," which is the reason the authors frame AI debt as a systemic issue rather than a tech-stock issue.
The counter-argument deserves a fair hearing. Insurers and institutional investors are heavily regulated, and many would say diversified portfolios absorb shocks. That's true for a single bad bond. It's less reassuring when many institutions hold correlated exposure to the same sector at the same time, which is exactly the scenario the report describes.
What Regulators Should Do, and What You Can Do Now
The report's policy recommendations are specific. It argues regulators, Congress, and state policymakers should shine a light on AI industry finances, stop subsidizing AI infrastructure, and better protect workers' retirement savings from outsized risk. It also criticizes the Trump administration and federal financial regulators for loosening guardrails. Those are political claims, and you should weigh them as the authors' position rather than settled fact.
For an individual, the practical steps are more modest. Check whether your pension or endowment discloses private credit exposure. Look at your life insurance provider's financial strength ratings from agencies like AM Best or S&P, and remember that a rating is a lagging indicator. Keep your own savings diversified so you're not relying on one sector's growth.
Most guides stop at "diversify." Here's the more useful version: write down what you'd do if your portfolio fell 30 percent in six months. If the answer is "panic," adjust your allocation now, while markets are calm.
Frequently Asked Questions
What is an AI bubble burst?
An AI bubble burst is a sharp drop in AI-related asset prices and company valuations, usually after investors realize that spending on AI infrastructure won't be repaid by revenue. The Open Markets Institute report argues that heavy debt, speculative data center financing, and interconnected deal structures could make such a drop more damaging. It does not predict a specific timing or date.
How do you protect your money if the AI bubble bursts?
Start by checking how much exposure your retirement plan, pension, or insurance provider has to private credit and data center debt. Diversify across sectors and asset types, and avoid concentrating savings in a few AI-linked stocks. Review your allocation regularly rather than reacting after a drop.
Why could AI data center debt affect pension funds?
Pension and endowment funds buy debt and private credit to earn steady returns. If that debt is tied to data centers whose revenue depends on uncertain AI demand, losses can reduce the funds' assets. Because the risk can be hidden in complex packaged securities, members may not know the exposure exists.
Is circular financing the same as a bubble?
No. Circular financing is a pattern where money cycles between vendors and customers, which can inflate apparent demand. It's a warning sign rather than proof of a bubble. Whether it becomes a serious problem depends on how much real end-customer revenue exists to repay the debt.
The Takeaway
The most useful thing you can do is treat the AI bubble question as a debt question, not a hype question. Ask where the revenue comes from, who holds the debt, and whether your own savings depend on the answer. Pick one thing from your retirement or insurance paperwork and check it this week, then pass this article to someone who's wrestling with how much AI exposure their portfolio already carries.