Home Latest Insights | News Baidu CEO Warns AI Boom Could Become a Bubble as Companies Struggle to Turn Spending Into Profit

Baidu CEO Warns AI Boom Could Become a Bubble as Companies Struggle to Turn Spending Into Profit

Baidu CEO Warns AI Boom Could Become a Bubble as Companies Struggle to Turn Spending Into Profit

Baidu Chief Executive Robin Li has warned that the artificial intelligence boom risks becoming a financial bubble if the enormous sums being invested in the technology fail to generate meaningful revenue and profit growth for businesses.

Speaking on Tuesday at the Asia-Pacific Economic Cooperation (APEC) CEO Summit in Shenzhen, Li said most companies had yet to see significant financial returns from their AI investments, raising questions about the sustainability of an industry built on heavy spending on semiconductors, data centers, cloud infrastructure and advanced models.

“If we fail to address this, the concern that AI might be a bubble could well become a reality,” Li said.

His warning cuts to the major challenge facing the global AI industry: the gap between the money being committed to developing the technology and the economic value companies are currently extracting from it. Technology groups are expanding computing capacity and racing to build more powerful models, but those investments will ultimately need to generate paying customers, higher productivity or lower operating costs to justify their scale.

Li divided the AI industry into four layers: chips, cloud infrastructure, large models, and applications. He said competition was currently most intense at the foundational levels, where companies are fighting for access to computing resources and seeking to develop more capable models. However, he argued that the greatest economic value must eventually be captured at the application layer.

His assertion raises a difficult question for companies leading the AI infrastructure boom. Chipmakers and cloud providers can earn revenue from selling computing capacity, but their customers must still find profitable ways to use it. If businesses cannot generate sufficient returns from AI applications, demand for infrastructure could eventually fall short of the expectations underpinning current investment plans.

The industry therefore faces two different measures of success. One is technological progress, measured by computing power, model capabilities and the volume of AI services being delivered. The other is commercial performance, measured by revenue, margins, customer retention and the productivity gains businesses can demonstrate. Progress in the first does not guarantee success in the second.

Li’s comments come as companies across the technology sector continue to commit substantial resources to AI development. The infrastructure required to train and operate advanced models is expensive, while businesses adopting the technology must also pay for computing capacity, software integration and the expertise needed to deploy it.

For AI developers, growing usage can result in a misleading picture of commercial success if the cost of serving customers rises alongside demand. A model may attract millions of users and process large volumes of requests without generating enough revenue to cover the computing and development costs involved.

The application layer is where that imbalance will become clearer. Businesses need AI products that solve specific problems, improve efficiency, or create services for which customers are willing to pay. Without those outcomes, demand for powerful models risks remaining concentrated among technology companies and organizations experimenting with the technology rather than a broad base of profitable users.

Li’s four-layer framework also highlights the uneven distribution of value across the AI supply chain. Companies supplying chips and cloud infrastructure can monetize the construction of AI capacity, while model developers face the cost of maintaining and improving their systems. Application providers, meanwhile, must prove that customers will pay enough for AI-enabled services to support the underlying expenses.

Li said Baidu had launched a product called “Token Exchange”, which connects AI token providers directly with enterprise and individual customers to help address computing shortages.

AI tokens are units used to process and generate information through AI models. They provide a way to measure model usage and are central to the cost of running AI applications, particularly as businesses integrate the technology into software and automated workflows.

Baidu’s exchange is intended to connect customers with providers of the computing capacity required to process AI requests. The initiative places the company in a market where access to computing resources has become an important constraint on AI adoption.

However, expanding access to tokens does not automatically resolve the industry’s profitability challenge. More available computing capacity can encourage companies to experiment with AI and increase usage, but the economic case still depends on what those applications deliver relative to their operating costs.

Li remained optimistic about the long-term direction of the technology, describing AI as an “inevitable trend”. His warning was directed at the industry’s investment and monetization gap, not the future relevance of AI itself.

That is considered crucial for investors and technology companies. A technology can transform industries over time while individual companies still overinvest, misjudge demand, or struggle to earn adequate returns. The expansion of AI does not guarantee that every infrastructure provider, model developer, or application business will benefit equally.

No posts to display

Post Comment

Please enter your comment!
Please enter your name here