Microsoft recently announced that due to concerns about AI computing power oversupply, it will cancel investments in some data centers in the United States and Europe. This decision directly impacted the stock prices of several cryptocurrency mining companies, including Bitfarms, CleanSpark, and Marathon, with stocks generally falling 4%-12%.
Market analysis indicates that Microsoft's investment reduction intensifies the cryptocurrency mining companies' dependence on AI business. With declining revenues after Bitcoin halving, the entire mining industry is under pressure. Microsoft plans to focus on retrofitting existing data center equipment rather than expanding new facilities, with plans to further slow expansion in the second half of 2025.
Alibaba's Cai Chongxin Warns: Global Data Center Construction Too Fast, Bubble Risk Exists
At the HSBC Global Investment Summit in Hong Kong, Alibaba co-founder Cai Chongxin warned that global data center construction has exceeded initial AI needs, especially with redundant and overlapping data center investments in the United States, potentially creating a bubble risk in AI data centers. Cai Chongxin stated:
Many current data center plans are launched without confirmed customer demand, and this blind expansion carries high risks, urging the market to assess carefully.
Cai Chongxin further pointed out that many companies and investment funds are racing to establish server bases, with signs of blind expansion already emerging. Particularly when data center plans begin fundraising without securing "rack" agreements, he is concerned about the market's future performance. His remarks have drawn market attention to AI bubble risks and put pressure on Taiwan's AI-related stocks, with Inventec's stock declining and companies like Quanta and Wistron showing weakness.
Global Enterprises Increase AI Infrastructure Investment, Yet Encounter Oversupply
Despite Cai Chongxin's warning about rapid AI data center construction, global tech giants continue to increase investments in AI infrastructure. Amazon, Alphabet (Google's parent company), and Meta have respectively committed to investing $100 billion, $75 billion, and up to $65 billion in AI infrastructure this year. However, with over-expectations of future demand, these massive investments might lead to surplus computing resources.
According to Goldman Sachs' latest report, due to lower-than-expected customer demand and supply chain bottlenecks, AI server rack shipment expectations have been significantly reduced. Goldman Sachs has downgraded AI server shipments for 2025 and 2026 to 19,000 and 57,000 units respectively, representing decreases of 38.7% and 13.6% from original estimates. This forecast adjustment significantly impacts Taiwan's AI supply chain, leading to target price reductions for companies like Quanta, Foxconn, and Wistron.
Market Concerns About AI Computing Power Oversupply Intensify
Taiwanese AI server OEMs in the supply chain privately confirm the risk of computing power oversupply. These suppliers note that AI data centers built by large enterprises like Alibaba, Amazon AWS, Microsoft, and Google are not only for internal use but also for leasing to other customers. For small and medium cloud service providers offering computing power rental services, current supply and demand imbalances pose risks.
The emergence of China's open-source AI model DeepSeek earlier this year has further intensified market concerns about potential AI computing power demand decline. Supply chain vendors in Taiwan also point out that the AI server procurement market, previously dominated by large cloud service providers, is now seeing small and medium enterprises purchasing AI servers to train their own AI models, indicating market structure changes.
Goldman Sachs Downgrades Taiwan Stock AI Server Supply Chain Forecast
According to Goldman Sachs' report, AI server rack shipments for 2025 and 2026 will significantly decrease due to unmet customer demand and supply bottlenecks. This led Goldman Sachs to lower target prices for Taiwan's AI supply chain companies, reducing Quanta's target price to 293 and downgrading its rating from "buy" to "neutral", with target prices for 10 other stocks also reduced by 5-13%.
While Goldman Sachs remains optimistic about Quanta's AI server business growth, expecting a 33% revenue increase this year and 57% next year, current market conditions suggest potential future supply-demand imbalances.
