The warning came from a buyer, not a skeptic, and that is what made it land. Temasek, the Singapore state investment firm, has spent years pouring money into artificial-intelligence companies, including the private model developers OpenAI and Anthropic. On Wednesday, the firm’s chief investment officer walked onto a stage in Singapore and told an audience of investors that the AI trade, the very trade his own fund has ridden upward, is now the biggest risk to global markets.
“The unwinding of the AI trade is the biggest risk,” Rohit Sipahimalani said at the Milken Institute Asia Summit. “We don’t see that as imminent. But, will you have bumps in 2027, possibly yes.”
A cautious investor warning about a frothy market is ordinary. What made Sipahimalani’s remarks travel was the firm delivering them. Temasek has not pulled back from AI. It has kept adding to its holdings, and Sipahimalani said the fund remains bullish on the sector over the long term. The change he described is subtler: moving money out of private, illiquid AI positions and into publicly traded ones that can be sold in an afternoon.
Temasek currently holds roughly half of its AI exposure in listed securities. Sipahimalani said he wants to push that share to 70% or 75%. The reason is simple. “AI is such a fast changing environment, things could change quite easily and you have to be able to pivot,” he said. A stake in a private startup cannot be dumped quickly when sentiment turns; a position in Nvidia or a listed cloud provider can.
The triggers he named for a reversal are the same ones investors have spent the past two years debating and mostly dismissing. Tighter regulation driven by AI safety concerns. Signs that corporate customers are not earning enough from their spending on the technology to justify the cost. A collapse in the assumption, now baked into share prices across the sector, that the hundreds of billions of dollars flowing into data centers and chips will eventually pay off.
Sipahimalani’s second risk for 2027 was inflation, and he tied the two together. Higher energy prices, supply-chain disruptions, or geopolitical conflict could keep price pressures elevated, he argued, pushing bond yields to a level that finally cracks the equity market’s run. “Inflation is a risk, together with what is happening in the rates environment,” he said. “There probably is a risk of some breaking point in the equity market at some point.” Together, he said, the AI trade and inflation are “two key risks I would see for 2027.”
The assessment lands at a moment when the AI rally looks both powerful and fragile. U.S. markets have priced in years of AI-driven growth in a very short stretch. Investment banks have published forecasts for trillions of dollars in data-center spending by 2030. Chipmakers, cloud providers, and power companies have all been bid up on the premise that the spending wave will continue. Any credible sign that customers are pausing their purchases would force a repricing that reaches far beyond the technology sector.
That is exactly the scenario Sipahimalani is positioning Temasek to survive. A fund that holds most of its AI exposure in liquid public assets can cut and move in a week. A fund locked into private rounds cannot. His message to the rest of the market was not a call to exit AI. It was a call to stop pretending the trade is permanent.
The speech also carries a quieter point about the fund’s own discipline. Temasek has been an early and aggressive AI investor, and a public correction would hit its portfolio too. By shifting toward listed securities now, the firm is admitting that the positions it can exit quickly are worth more to it than the ones it cannot. Analysts who follow the firm read the move as prudence dressed up as strategy, a hedge taken while the market is still rising rather than after the decline begins.
The framing is unusual for a fund of Temasek’s kind. Sovereign-style investors are typically slow to name a trade they are heavily exposed to as a top risk, because doing so invites the question of why they are still in it. Sipahimalani’s answer is that Temasek is in AI for the long run, and that long-run conviction is compatible with short-run wariness. The difference between a bubble and a boom, in his account, is not whether valuations are rich but whether an investor can exit when the exit is needed. That is a position built on liquidity, and it explains the arithmetic behind the shift into listed assets.
For everyone else, the lesson is the same one that circulates at the end of every boom: the people closest to the trade are often the first to start measuring the distance to the door. Sipahimalani did not predict a crash. He predicted bumps, and then spent the rest of his remarks explaining how his firm plans to be somewhere else when they arrive.
