Chinese technology executives are increasingly warning that the AI trade is becoming top-heavy, with huge spending on chips, cloud capacity and data centers outrunning the commercial applications needed to justify it.
Chinese Tech Executives Warn on AI Spending
The caution, voiced at the APEC Business Leaders Summit in Shenzhen, matters because it goes to the core of the current AI investment cycle: capital expenditure is still rising sharply, but many companies are not yet seeing a matching lift in revenue or profit. That gap is exactly what turns a technology wave into a valuation risk for investors.
Baidu founder Robin Li said the industry has concentrated too much on the first three layers of the AI stack — GPUs, computing infrastructure and large language models — while the real economic value must eventually come from end-user applications. He warned that three years of exponential AI investment had not translated into meaningful improvement in earnings for many companies, and that if practical use cases do not emerge, fears of an AI bubble could become reality.
The message is significant because it came from executives who are still deeply committed to the technology, not from skeptics outside the sector. Li did not argue that AI is overhyped in principle; he argued that the market is pricing too much of the story into infrastructure and model-building before the application layer has matured. That is a more dangerous warning for investors, because it implies the bubble risk is not about AI disappearing, but about returns arriving too slowly to support the money being spent.
Xiaomi founder Lei Jun echoed the concern, saying the sector has poured enormous resources into AI over the past four years while only a limited number of applications have become profitable. He pointed instead to AI’s potential in physical-world manufacturing and robotics, including factory automation and humanoid robots. Xiaomi has already been testing an AI robot in a plant, underscoring how China’s tech champions are looking for the next leg of monetization beyond consumer software and internet services.
Unitree Robotics founder Wang Xingxing said the robotics industry still faces major hurdles in generalization, safety, reliability and mass production, even if current technical breakthroughs bring the field closer to a tipping point. His comments highlight the same economic issue from another angle: robotics may create a new wave of AI spending, but the path from demonstration to scalable profit remains uncertain.
The investment implications are straightforward. If AI demand is shifting from model training to deployment, winners will be the companies that can prove software and automation tools actually lower costs or raise output. Losers are the suppliers dependent on ever-rising infrastructure capex without clear evidence of monetization, as well as investors paying premium multiples for growth that has yet to show up in cash flow.
That concern is already visible in market behavior. Nvidia, the emblematic AI hardware play, has seen its shares fall sharply in recent sessions, while Microsoft has also retreated from recent highs. The broader AI complex has become more sensitive to any sign that customers are delaying infrastructure spending or reining in budgets, a risk that major chip suppliers have already flagged in regulatory filings. For the market, that means the debate is shifting from whether AI is transformative to whether the economics of the boom are sustainable.
There is still a bull case. TCL chairman Li Dongsheng said AI integration in manufacturing had already delivered about 2.5 billion yuan in cumulative economic benefit, showing that some industrial users are extracting real value. CATL sees opportunity in the energy demands of data centers and electrified transport, and BYD argues that China’s manufacturing advantage will increasingly depend on embedding automation across global supply chains.
But the bearish narrative is hard to ignore: the biggest costs in AI are being booked now, while the payoff may arrive later and unevenly. If adoption stays confined to flashy demos and infrastructure buildout, the sector risks repeating a familiar cycle in which enthusiasm outruns earnings. For investors, the critical question into next year is not whether AI matters, but which parts of the value chain can turn scale into profit before patience runs out.
| Entity | Gains | Losses |
|---|---|---|
| AI application providers | ▲Monetization if real use cases expand | ▼Margin pressure if adoption lags |
| Chip and cloud suppliers | ▲Near-term capex demand | ▼Risk if customers cut spending |
| Industrial automation leaders | ▲New factory and robotics demand | ▼Integration costs and execution risk |
| AI equity investors | ▲Upside from successful commercialization | ▼Valuation reset if profits stay weak |


