Analysis: China Bets on Chip Clusters to Survive U.S. Sanctions
Listen to the full version
Facing U.S. sanctions blocking access to advanced semiconductors, Chinese tech companies are increasingly relying on supernode technology that can integrate clusters of lower-performance chips into servers to achieve similar computing capacities. But systematic complexity, high costs and insufficient demand cloud their large-scale application.
Since mid-2025, many Chinese artificial intelligence (AI) chip designers, server manufacturers and cloud services providers have rolled out their own supernode products amid a surge in demand for training large language models (LLMs) that have been applied in a wide range of industries from education and health care to finance and marketing.
Unlock exclusive discounts with a Caixin group subscription — ideal for teams and organizations.
Subscribe to both Caixin Global and The Wall Street Journal — for the price of one.
- RELATED
- DIGEST HUB
Digest Hub
Back
- Chinese tech firms use supernode technology to cluster lower-performance chips, offsetting U.S. semiconductor sanctions, but face high costs and complex engineering challenges.
- Examples include Huawei’s CloudMatrix 384 (384 chips), Alibaba’s Panjiu (128 chips), and Baidu’s announced 256-chip system, mostly targeting large state enterprises.
- Industry experts question the demand and efficiency of massive supernodes, suggesting most needs are met with smaller clusters (around 64 chips) and emphasizing reliability over scale.
- Huawei Technologies Co. Ltd.
- In June, Huawei Technologies Co. Ltd. launched the CloudMatrix 384, an AI computing system. This system uses 384 Ascend 910C chips and offers nearly double the computing power of Nvidia's GB200 NVL72 system. However, each Ascend 910C chip provides only one-third the performance of Nvidia's advanced Blackwell architecture chips.
- Nvidia Corp.
- Nvidia Corp. is a leading company in advanced semiconductors, but Chinese companies are developing alternative technologies due to US sanctions. Huawei's CloudMatrix 384 system, for example, delivers nearly double the computing power of Nvidia's GB200 NVL72 system. While Nvidia dominates with a vertically integrated ecosystem, Chinese firms are focusing on system-level connections to overcome restrictions.
- Alibaba Group Holding Ltd.
- In September, Alibaba Group Holding Ltd. introduced the Panjiu AI server. This server is powered by a cluster of 128 self-developed chips, demonstrating Alibaba's strategy to overcome U.S. sanctions on advanced semiconductors by integrating numerous lower-performance chips into a powerful computing system.
- Baidu Inc.
- Baidu Inc. introduced its own supernode products in November, joining other Chinese tech firms in this endeavor. Additionally, they have plans to launch a computing system capable of housing 256 chips in the first half of 2026. This move reflects their strategy to overcome U.S. sanctions on advanced semiconductors by prioritizing system-level connections.
- IEIT Systems Co. Ltd.
- IEIT Systems Co. Ltd. (中科可控信息产业有限公司) is an IT infrastructure provider. They offer servers capable of an inference cost of approximately 1 yuan (14 U.S. cents) per 1 million tokens. The company is dealing with the broader market's need for substantially lower costs for the mass adoption of large models.
- Since mid-2025：
- Many Chinese AI chip designers, server manufacturers, and cloud services providers rolled out their own supernode products amid a surge in demand for training large language models.
- MOST POPULAR
Sign in with Google
Sign in with Facebook
Sign in with 财新