Speaking Tuesday at the HSBC Global Investment Summit in Hong Kong, Tsai said he was “astounded” by the AI data center investment figures he was seeing in the U.S., citing the $500 billion that OpenAI, SoftBank, and Oracle say they will spend on the Stargate project announced back in January.

“I don’t think that’s entirely necessary,” the cofounder of the Chinese tech conglomerate said. “I think in a way people are investing ahead of the demand that they’re seeing today… Is that thinking correct or incorrect? You can make a judgement.”

But the real red flag for Tsai is the phenomenon of data center developers building new projects “on spec,” without binding agreements from the large AI companies like Microsoft or Google or Alibaba itself.

“I start to see the beginning of some kind of bubble,” Tsai warned.

The Alibaba chair is, of course, not the first to suggest that a bubble is forming—the idea has been commonplace throughout the AI arms race that began when OpenAI released ChatGPT in late 2022. But bubble predictions have largely come from those who are critical of the industry’s reliance on large language models, on the basis of unresolved problems such as inaccuracy and a lack of profitability.

Tsai is bullish on AI, as he took pains to stress immediately after mentioning his “bubble” fear: “However, having said that, that doesn’t take away from our excitement about AI and the future proliferation of AI and how AI can be applied in every walk of life and every industry.”

Alibaba recently pledged to spend $53 billion on data centers and cloud computing, more than it’s spent on AI infrastructure over the past decade.

Bridgewater Associates founder Ray Dalio also recently warned of a bubble in U.S. AI stocks, comparing the current moment to the run-up to the dotcom bust around the turn of the century. “Where we are in the cycle right now is very similar to where we were between 1998 or 1999,” Dalio told the Financial Times in January. “In other words, there’s a major new technology that certainly will change the world and be successful. But some people are confusing that with the investments being successful.”

However, there may be another aspect to Tsai’s warning. Led by DeepSeek, Chinese AI companies are currently going all in on models that are open-source and lightweight, in some cases even being small enough to run on consumer hardware.

On Monday, Alibaba’s AI team announced a “smarter and lighter” version of its open-source Qwen2.5-VL series of models, claiming significant improvements in areas such as mathematical reasoning and image understanding despite the fact that the version has just 32 billion parameters. At that size, the model is small enough to run on many powerful desktop computers, meaning its use wouldn’t require tapping into any data center for some people.

Advances such as this—and DeepSeek’s low-cost model creation—suggest that Stargate-level horsepower may indeed be overkill (though some would argue that exploding AI uptake may justify that level of investment). So perhaps Tsai’s warning had a hint of a taunt.