Chinese Companies Shift Nearly Half of AI Computing Budget to Domestic Chips, HBM Emerging as New Bottleneck Under $2 Trillion Data Center Plan

According to ChainNews, a survey report released by Bloomberg Industry Research indicates that due to geopolitical and policy factors, Chinese companies are accelerating the shift of their computing power budget to domestic resources. A survey of 60 Chinese executives in the software, financial, and manufacturing industries revealed that in the next 12 months, the proportion of domestic AI accelerators in enterprise budgets will increase from the current 30% to 46%. The high cost of AI projects has put a strain on corporate finances, with 80% of the executives surveyed stating that this year's infrastructure spending has exceeded the budget.

Nvidia's China-exclusive H20 chip is facing more stringent acquisition restrictions, prompting local computing power service providers such as Tencent, Alibaba, Huawei, as well as Phytium Information and Cambricon to accelerate filling the market gap. At the policy level, the Chinese government also plays a role in this trend, with plans to invest around 2 trillion yuan over the next five years to build national data centers and setting a target for the domestic supply of core technologies such as chips to reach at least 80%.

However, computing power autonomy is also facing new supply chain bottlenecks. The core challenge for Chinese AI companies has now shifted from mere computational performance to the acquisition of High Bandwidth Memory (HBM). Due to a global memory chip shortage limiting foundry growth, local memory manufacturers like Sunway Storage are seizing the opportunity to fill the supply gap.