SMIC CEO Warns: AI Spending Spree Risks Bubble and Data Center Oversupply

Zhao Haijun, CEO of SMIC, has issued a warning. He believes that the huge spending on AI chips could lead to an oversupply in production capacity and leave many data centers unused. He pointed out that companies are trying to build enough capacity for ten years in just two years, without fully planning for the future operational needs of these centers, according to a Bloomberg report. Moody's estimates suggest that investments in AI infrastructure will go beyond $3 trillion over the next five years, with massive amounts of money being poured into data centers. Tech giants like Amazon, Microsoft, and Meta are expected to spend a whopping $650 billion in capital expenditures in 2026 alone. Zhao also mentioned that Chinese companies such as Alibaba and ByteDance are finding it tough to secure local alternatives to Nvidia chips due to U.S. export restrictions. There's also a severe shortage of High Bandwidth Memory (HBM) components, which are crucial for advanced computing, and this shortage is expected to last for years. Despite these challenges, local companies like Huawei are working hard to quickly boost silicon production to meet national needs. This is a proactive strategy aimed at preparing the infrastructure for a decade of digital transformation, even with the short-term economic risks tied to the viability of these investments. SMIC's warning highlights a paradox in the global AI race: while spending on infrastructure and chips is increasing at an unprecedented pace, the question of how these massive investments will actually be used remains open. This could potentially lead to a short-term tech bubble amidst a costly technological arms race.