Baidu chief executive Robin Li warned on Tuesday, November 17, that the artificial-intelligence investment boom could turn into a bubble unless businesses find ways to earn meaningful profits from the technology. Speaking at the APEC CEO Summit in Shenzhen, Li said heavy spending on AI had not yet translated into substantial revenue or profit improvements for most companies, Reuters reported.
A warning from a major Chinese AI developer
Li's intervention is notable because Baidu is itself investing in AI models, cloud infrastructure and applications. He did not say AI had no future. Rather, he argued that the economics of adoption must eventually justify the spending. His remarks distinguished the long-term potential of the technology from the financial performance of companies currently building it.
Li described four layers of the industry: semiconductors, cloud infrastructure, large models and applications. Competition has concentrated on the foundation layers, but he argued that value ultimately needs to be captured in products and services used by customers. The point is relevant to businesses buying expensive computing capacity: demand for chips and model training alone does not guarantee that customers will pay enough for applications to produce sustainable returns.
Baidu's proposed marketplace
At the summit, Li discussed Baidu's Token Exchange, which he said connects AI token providers with enterprise and individual customers to ease computing shortages. In AI services, tokens are units of text or other information processed by models and commonly used to measure usage and billing. A marketplace could help connect supply and demand, although Reuters did not provide independently verified figures on its volume, profitability or customer adoption.
Li nevertheless called AI an inevitable trend, signalling that his warning concerned commercial discipline rather than a rejection of the technology. It is possible for a technology to spread widely while some suppliers fail to earn back their investments; those outcomes are not contradictory.
Why the timing matters
The remarks came as global markets were weighing the cost of AI expansion, with technology shares under pressure ahead of Nvidia earnings, according to a separate Reuters market report on November 17. Investors have been assessing whether spending on data centres, advanced processors and infrastructure can generate lasting cash flows. The market backdrop helps explain why executives are increasingly discussing revenue and profits rather than model capabilities alone.
The United States has also tightened controls on advanced AI chips and overseas computing access, adding another source of uncertainty for Chinese developers and data-centre operators in several countries. That regulatory context could affect supply and costs, but Li's warning about business returns is a distinct question from export restrictions.
What would establish a bubble?
A warning is not a measurement showing that the whole industry is already in a bubble. Evidence would require comparing actual and expected revenue, capital spending, utilisation, margins, financing costs and customer retention across companies. Some AI applications may become profitable even if expensive infrastructure projects disappoint investors. Conversely, rapid user adoption does not guarantee attractive economics when computing costs remain high.
For businesses and policymakers, the practical issue is whether AI improves productivity or creates services customers value enough to purchase. For investors, the question is how those benefits are distributed among chipmakers, cloud providers, model developers and application companies. Li's argument is that the last group must become a meaningful source of economic value.
Bottom line: The Baidu CEO warned that investment without profits could validate bubble concerns, while reaffirming his belief in AI's long-term importance. His comments are an assessment of industry risks, not proof that a collapse has begun.




