Microsoft and Alibaba Warn of AI Server Oversupply
Microsoft and Alibaba have issued warnings regarding an oversupply of AI servers and announced plans to reduce investments in certain data centers. This decision has raised concerns in the market about a potential bubble in the AI industry and has put pressure on the stock prices of cryptocurrency mining companies.
Background Information
(Background: OpenAI recently launched its strongest image generation model: precise creation of infographics, multimodal input, and realistic quality, integrated into GPT-4o.)
Recently, Microsoft announced that it would cancel investments in some data centers in the United States and Europe due to concerns over an oversupply of AI computing power. This decision directly impacted the stock prices of several cryptocurrency mining companies, including Bitfarms, CleanSpark, and Marathon, which saw declines ranging from 4% to 12%.
Market analysis indicates that Microsoft’s reduction in investment exacerbates cryptocurrency mining companies’ reliance on AI businesses. As Bitcoin halving leads to decreased profits, it has put pressure on the entire mining industry. In response, Microsoft plans to focus on upgrading existing data center equipment rather than expanding new facilities, with further expansion plans expected to slow down in the second half of 2025.
Alibaba’s Joe Tsai Warns: Rapid Global Data Center Construction Poses Bubble Risks
At the HSBC Global Investment Summit in Hong Kong, Alibaba co-founder Joe Tsai warned that the pace of global data center construction has outstripped the initial demand for artificial intelligence (AI), particularly noting instances of redundancy and overlap in investments across multiple data centers in the United States. This could potentially lead to a bubble risk in AI data center construction. Tsai stated:
“Currently, many data center projects are initiated without confirmed customer demand, presenting a very high risk of blind expansion, and I urge the market to assess cautiously.”
Tsai further pointed out that many enterprises and investment funds are competing to establish server bases, but this wave of expansion has begun to reveal signs of reckless growth. Particularly concerning is that some data centers are raising funds before securing “go-live” agreements, which makes him apprehensive about the future performance of the market. Tsai’s remarks have drawn attention to the risks of an AI bubble and have put pressure on the performance of AI-related stocks in the Taiwanese market, especially with regard to Inventec, which has close ties to Alibaba. Other Taiwanese companies like Quanta and Wistron have also shown signs of weakness.
Global Giants Invest in AI Infrastructure Amid Oversupply Concerns
Despite Tsai’s warnings about the rapid construction of AI data centers, global tech giants continue to increase their investments in AI infrastructure. Amazon, Alphabet (Google’s parent company), and Meta have each committed to investing $100 billion, $75 billion, and up to $65 billion, respectively, to build AI infrastructure this year. However, with overly optimistic expectations for future demand, these large-scale investments may lead to an oversupply of computing resources.
According to a recent report from Goldman Sachs, expectations for AI server rack shipments have been significantly downgraded due to customer demand falling short and supply chain bottlenecks. Goldman Sachs has revised its AI server shipment forecasts for 2025 and 2026 down to 19,000 and 57,000 units, respectively, representing declines of 38.7% and 13.6% from previous estimates. This adjustment has had a considerable impact on the Taiwanese AI-related supply chain, leading to downward revisions of target prices for various Taiwanese companies, including Quanta, Hon Hai, and Wistron.
Concerns About AI Computing Power Oversupply Intensify
Sources within Taiwan’s AI server manufacturing industry have privately indicated that there is indeed a risk of oversupply of computing power. They noted that data centers built by major companies such as Alibaba, Amazon AWS, Microsoft, and Google are not only for their own use but are also rented out to other clients. If small to medium-sized cloud service providers are merely offering computing power rental services, there is a genuine risk of an imbalance in supply and demand.
Earlier this year, the emergence of the open-source AI model DeepSeek in Mainland China has further heightened concerns about a potential decline in demand for AI computing power. Taiwanese supply chain providers have also noted that the AI server procurement market, previously dominated by large cloud service providers, has begun to show signs of small and medium-sized enterprises purchasing AI servers to train their own AI models, indicating a shift in market structure.
Goldman Sachs Lowers Forecasts for Taiwanese AI Server Supply Chain
According to a report from Goldman Sachs, AI server rack shipments for 2025 and 2026 are expected to decline significantly due to customer demand not meeting expectations and supply chain bottlenecks. This has led Goldman Sachs to lower target prices for Taiwanese AI supply chain companies, with Quanta’s target price reduced to NT$293 and its rating downgraded from “Buy” to “Neutral.” The target prices for ten other stocks have also been cut by 5-13%.
While Goldman Sachs remains optimistic about Quanta’s growth momentum in the AI server business, projecting a 33% increase in revenue this year and a 57% increase next year, the current market conditions suggest a potential imbalance between supply and demand in the future.
