SMIC warns AI data center boom risks excess capacity

China’s leading chipmaker, Semiconductor Manufacturing International Corp. (SMIC), has warned that rapid investment in AI data centers may lead to underutilization.

SMIC’s co-CEO Zhao Haijun said that companies are rushing to build capacity that could outpace actual demand, with some data centers potentially remaining idle.

Despite a projected US$3 trillion in AI infrastructure spending over the next five years, US export restrictions limit SMIC’s ability to produce the most advanced AI chips, which are mainly made by Nvidia and TSMC.

Zhao also noted a persistent shortage of high-bandwidth memory, a key component for AI computing, which could last for years.

The company compared the infrastructure build-up to constructing high-speed rail stations and highways in a short period, regardless of current usage levels.

Major Chinese AI firms, including Huawei and Cambricon, are increasing chip production to meet rising domestic demand.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

US tech giants are pouring money into AI data centers at a rare pace

SMIC’s co-CEO warns about too much spending on AI data centers. Even so, Alphabet, Amazon, Meta, and Microsoft expect capital expenditures to reach about $650 billion in 2026 1.
That forecast from four companies tops the combined $180 billion in projected capex for 21 other large US firms. The group includes the biggest automakers, defense contractors, and Walmart 1.
Investors have reacted. The four companies have lost over $950 billion in market value since issuing recent earnings results and outlooks that included these spending plans 1.

The AI surge is pushing up phone and PC prices

The high-bandwidth memory (HBM) shortage raised by SMIC lines up with IDC (International Data Corporation, a tech market research firm). IDC says memory makers are shifting capacity from consumer devices to higher-margin AI data center products such as high-bandwidth memory (HBM) and high-capacity DDR5, which tightens supply for phones and PCs 2.
IDC expects smartphone average selling prices to rise 3% to 5% in a moderate downside scenario, or 6% to 8% in a pessimistic downside scenario in 2026. IDC also expects a 2.9% to 5.2% market contraction, depending on how long constraints persist 2.
PCs face similar strain. IDC projects 2026 PC average selling prices rising 4% to 6% in a moderate downside case, or 6% to 8% in a pessimistic downside case 2.
The situation complicates the industry push for “AI PCs,” since many ship with more RAM. IDC says Microsoft’s Copilot+ PCs require at least 16GB, which can raise costs and limit configuration options for PC makers and retailers 2.