During a keynote address at the HSBC Global Investment Summit in Hong Kong, Alibaba Group Holding Ltd. Chairman Joe Tsai expressed serious concerns about what he perceives as an emerging bubble in AI data center investments, particularly in the United States. His remarks come amid a wave of massive infrastructure commitments from major technology companies.
Tsai highlighted the worrying trend of data center projects securing funding without confirmed uptake agreements, noting that numerous entities are raising substantial capital for speculative builds. This comes at a time when industry giants like OpenAI, SoftBank, Oracle, and other American technology firms have announced plans for AI infrastructure projects totaling approximately $500 billion. Additionally, Meta has unveiled $200 billion in data center
initiatives, while Apple has also declared new AI-related investments in the United States.
The Alibaba chairman questioned the necessity of such extensive investments, suggesting that current spending levels may be outpacing actual market demand. His comments coincided with significant market movements, as Alibaba’s Hong Kong-listed shares declined by 4%, while Goldman’s China Data Center basket experienced an 8% overnight drop.
The situation has been further complicated by the recent launch of DeepSeek, a Chinese alternative to ChatGPT that Goldman analyst Rich Privorotsky describes as “40-50x more efficient than other large language models.” This development has prompted discussions about the possibility of achieving more with less infrastructure investment.
Market observers have noted potential signs of a slowdown in the data center expansion race. TD Cowen analyst Michael Elias recently caused market uncertainty by reporting that Microsoft had begun canceling data center leases, though the company subsequently denied these claims.
Goldman Sachs analysts have revised their projections for AI server volumes downward, citing various factors including product transitions and supply-demand uncertainties. They now forecast AI training server revenues to grow at 24% and 58% year-over-year to $150 billion and $237 billion in 2025-2026, respectively, reduced from previous estimates of $179 billion and $248 billion.
The impact of these developments has extended beyond individual companies, affecting broader market segments. Chinese technology stocks have experienced a sharp decline, falling from a three-year high to near correction territory in just five trading sessions. Saxo Markets chief investment strategist Charu Chanana noted that Alibaba’s warnings about the AI data center bubble have contributed to market pressure, suggesting potential short-term challenges for the AI sector.
Tsai expressed particular amazement at the scale of investment figures being discussed in the United States, emphasizing his belief that such extensive capital deployment might not be entirely justified by current market conditions. The emerging concerns about an AI data center bubble could potentially impact major infrastructure
initiatives, including President Trump’s Stargate AI infrastructure project.
The situation reflects a growing debate within the technology sector about the appropriate pace and scale of AI infrastructure development, particularly as more efficient models like DeepSeek demonstrate the potential for achieving similar capabilities with reduced resource requirements. As the industry continues to evolve, the balance between infrastructure expansion and actual market demand remains a critical consideration for investors and industry leaders alike.