Warnings from Michael Burry, Ray Dalio and Temasek CIO Rohit Sipahimalani converge on a risk: debt is playing a growing role in the AI buildout, making the market more exposed if expected returns fail to arrive or financing becomes less available. But their comments are risk warnings, not a shared forecast that markets will crash in 2027.
1
17
20
24
What each investor warned
Michael Burry: market “denial” and fragile financing
Burry compared the market’s mood with the periods before the 2000 dot-com crash and the 2008 financial crisis. He called it a “denial” phase and said that, in those past episodes, the phase lasted six to nine months. That is his historical comparison—not a reliable countdown to a repeat crash.
7
His concerns also extend beyond share prices. Burry has highlighted risks in private-credit investments tied to AI infrastructure and assets held by private-equity-owned insurers, including the difficulty of assessing less-liquid credit. The risk he describes is that financing problems could spread if AI spending or revenue weakens.
13
Ray Dalio: debt and rates could test the AI boom
Dalio called AI a “classic bubble,” pointing to heavy borrowing to fund investment. He warned that higher interest rates and pressure to turn wealth into cash could bring the market closer to a breaking point.
17
Separately, Dalio said he sees the possibility of a U.S. debt crisis within three years. That is a warning about national debt, distinct from his concerns about AI financing.
18
Temasek: an AI-trade unwind is a risk, not an immediate call
Temasek CIO Rohit Sipahimalani said an unwind of the AI narrative was the biggest risk to markets. He identified possible triggers including safety concerns, regulation, or end users failing to see sufficient returns as 2027 progresses.
20
Sipahimalani also said Temasek did not see an unwind as imminent, while allowing that markets could face turbulence in 2027. His remarks express caution about a possible reversal, not a prediction that one is certain.
24
Why the borrowing figures matter—and what they don’t prove
AFP reported that tech-sector borrowing went from next to nothing in 2024 to about $500 billion in the first nine months of 2026, as major technology companies financed AI-related chips, servers and data centers. The report also attributed a $1.2 trillion 2027 estimate to Goldman Sachs.
32
That larger figure needs qualification: AFP describes it as a borrowing estimate, while another report describes $1.2 trillion as projected AI-infrastructure investment. Without a consistent definition, it should not be presented as a confirmed forecast of new debt issuance.
32
37
The underlying concern is conditional: when an infrastructure buildout relies more on borrowing, higher rates or weaker-than-expected revenue can make projects harder to finance and returns harder to justify. Dalio’s remarks focus on debt and rates; Burry’s include the risks of interconnected financing; and Sipahimalani’s point is whether users ultimately see enough value in AI.
1
13
17
20
What to watch in 2027
The key test is whether AI-related revenue and returns can keep pace with the cost of the buildout. Sipahimalani specifically cited inadequate returns for end users as a possible reason the AI narrative could weaken. That makes company disclosures on spending, financing and realized returns useful signals to watch—but not proof, either way, that a broad market downturn is coming.
20
Taken together, the three warnings describe different paths to a possible market shock: Burry emphasizes complacency and credit-market vulnerabilities, Dalio focuses on debt and interest rates, and Temasek highlights the risk that enthusiasm could fade if the business case disappoints. They identify vulnerabilities; they do not establish the timing or certainty of a crash.