Chinese firms plan bigger AI chip budgets for local suppliers
A Bloomberg Intelligence survey found Chinese executives plan to shift more AI accelerator spending to domestic chips as U.S. tensions reshape supply plans.
By Maya Lindqvist · Senior Technology Correspondent
3 min read
Chinese companies are planning to spend a larger share of their AI accelerator budgets on homegrown chips, according to a Bloomberg Intelligence survey released Tuesday. The shift matters for Nvidia because it points to a narrowing path in one of the world’s most important AI infrastructure markets.
Bloomberg Intelligence said executives in China expect to direct 46% of their accelerator budgets to domestic products over the next 12 months. That compares with 30% at present, based on responses from 60 executives at Chinese software, finance, manufacturing and retail companies.
The survey also found that 80% of executives said their infrastructure spending has exceeded budget this year. Bloomberg Intelligence said the overruns are mostly tied to the cost of AI-related projects.
Domestic suppliers stand to gain
Bloomberg Intelligence said China’s push to replace foreign AI semiconductors with local alternatives is advancing. Its report identified Huawei Technologies Co. and Hygon Information Technology Co. among the domestic chipmakers likely to benefit from that trend.
The survey also found that accelerators from Hygon and Cambricon Technologies Corp. were under review by a broad set of respondents. Bloomberg Intelligence said major AI infrastructure builders and suppliers including Tencent Holdings Ltd., Alibaba Group Holding Ltd. and Huawei are well placed as spending shifts toward Chinese-made systems.
Nvidia’s chips remain widely used, according to the survey, but Bloomberg Intelligence said the company’s share in China is expected to decline. The report tied that outlook to reduced availability of Nvidia’s H20 chips and to Beijing’s guidance urging technology companies not to use them.
Nvidia, based in Santa Clara, California, has been caught between strong Chinese demand for AI hardware and U.S.-China technology restrictions. Bloomberg Intelligence said those pressures are helping redirect China’s AI buildout toward domestic alternatives.
Beijing is funding a larger AI buildout
Bloomberg News has reported that China is preparing to spend about 2 trillion yuan, or $294 billion, on data centers over five years. The government-led effort is intended to expand the use of AI across areas including health care and city management, according to Bloomberg News.
Bloomberg News also reported that at least 80% of core technologies for the plan, including chips, are expected to come from domestic suppliers. That target aligns with Beijing’s broader effort to reduce reliance on American technology, according to Bloomberg Intelligence.
The report said the next constraint for Chinese AI companies may come from memory rather than raw computing capacity. Bloomberg Intelligence said shortages of high-bandwidth memory chips, which move data rapidly inside AI systems, are becoming a key bottleneck.
That shortage could limit growth for Chinese AI firms such as Semiconductor Manufacturing International Corp., according to the report. Bloomberg Intelligence said ChangXin Memory Technologies Inc. is positioned to benefit as demand shifts toward memory supply.
The survey shows Chinese companies are still spending heavily on AI infrastructure, even as costs rise. It also shows that the money is increasingly expected to flow to domestic suppliers, a change that could reshape competition in AI hardware inside China.
This story draws on original reporting from Fortune.