Baidu's Robin Li Warns AI Bubble Fears Could Come True Without Profit
Baidu CEO Robin Li says most companies see no profit from AI and bubble fears could become real, as Q3 shows AI cloud up 94% but ads down 22%.
- By
- Tom Whitfield
- Filed
- Channel
- AI & Compute
- Read
- 3 min read
Baidu's AI cloud infrastructure revenue jumped 94% year on year to RMB8.2 billion in the third quarter, yet the same report showed online marketing revenue down 22% and total revenue down 1% to RMB30.8 billion. Against that backdrop, Chief Executive Robin Li said on Nov. 17 that heavy spending on artificial intelligence has not yet paid off for most companies that use it, and warned that fears of a bubble could come true if that does not change.
Li spoke on a panel at the APEC CEO Summit in Shenzhen, the business gathering held before this week's Asia-Pacific Economic Cooperation leaders' meeting. Asked about the global landscape for AI, he called the situation "urgent" and said its expansion into practical use "has arguably fallen short of expectations," according to AFP.
What did Li say about profits?
The core of his argument was a blunt question about returns. "Can we say that companies are seeing higher revenues or increased profits after adopting AI?" Li asked. "For most, the answer is no. If we fail to address this, the concern that AI might be a bubble could well become a reality."
The numbers behind his own company illustrate the split. Baidu's Q3 results, released Nov. 16, show where money is flowing in and where it is flowing out:
- Total revenue: RMB30.8 billion, down 1%
- AI cloud infrastructure revenue: RMB8.2 billion, up 94%
- Online marketing revenue: down 22%
In plain words, the infrastructure side of AI is growing fast at Baidu, but the traditional advertising business that funds it is shrinking at pace.
Where must value be captured?
Li described AI as four layers: chips, cloud infrastructure, large models and applications. Competition is fiercest at the foundation layers, he said, but the value ultimately has to be captured by applications at the top, Reuters reported.
That framework explains his urgency. Chips, data centers and model training absorb enormous capital, yet they only pay back if businesses buy and deploy software that earns money. If the application layer never catches up, the spending underneath becomes hard to justify.
He pointed to Baidu's new Token Exchange as part of the answer. The exchange connects AI token providers with business and individual customers, and Li said it should ease computing shortages. He also said he remains optimistic that AI is an "inevitable trend" despite the near-term gap between investment and returns. The takeaway: Li is not calling AI a bubble, but he is warning that the industry is on a clock to prove otherwise.
What else did Baidu set out?
Five days before the Shenzhen panel, at Baidu World in Beijing on Nov. 12, Li set out three principles for what he calls the "intelligent organization." The principles cover job boundaries, tacit knowledge and closed-loop processes — a managerial blueprint for how companies should restructure work around AI rather than simply bolt it on.
The timing matters. Baidu is both a large buyer of AI compute and a seller of AI cloud services, so its leadership is exposed to both sides of the profitability question Li raised. A company that sells the infrastructure while its own core ad business contracts has direct incentive to push customers toward applications that demonstrably earn money.
How did other executives respond?
Not everyone in Shenzhen shared Li's caution. Karan Bhatia, Google's global head of government affairs and public policy, took a more upbeat line on a separate panel. "We're entering into an AI-driven economy," Bhatia said.
The summit itself drew a notable U.S. business presence. Google and Boeing are part of the American business delegation to the two-day gathering, while Nvidia's Jensen Huang was unlikely to attend, the South China Morning Post reported last week, citing the organiser and a source.
The split between Li's warning and Bhatia's confidence frames the commercial question for the next several quarters: whether application-layer revenue grows fast enough to justify the capital already committed to chips and cloud capacity.
Source: mlq.ai
More from Tom Whitfield
Show full bio
Staff writer covering companies, deals and strategy in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.
1,016 articles
Related articles
baidu-ai-cloud-infrastructure-revenue-rises-94-as-search-ads-slide
Baidu AI Cloud Infrastructure Revenue Rises 94% as Search Ads Slide
bain-ai-must-find-6-trillion-a-year-to-pay-for-datacentres-581649a4
Bain: AI Must Find $6 Trillion a Year by 2031 to Pay for Datacentres
tsmc-posts-record-q3-profit-lifts-revenue-outlook-9dda902e
TSMC Posts Record Q3 Profit, Lifts 2026 Revenue Outlook
alphabet-s-128b-quarter-still-wipes-billions-off-its-market-value
Alphabet's $128B Quarter Still Wipes Billions Off Its Market Value


