Key facts
- Chinese AI startup DeepSeek is planning to develop its own silicon chips.
- The move is driven by US export controls and a desire to reduce reliance on Nvidia and Huawei.
- DeepSeek is focusing on chips for AI inference in data centers.
- Huawei currently dominates China's data center chip market.
- Other major Chinese tech firms, including Alibaba and Baidu, are also pursuing custom chip development.
Chinese AI startup DeepSeek is reportedly planning to develop its own silicon chips, a move driven by U.S. export controls and a desire to reduce reliance on foreign technology. The company has been working on this initiative for about a year, meeting with potential hardware partners and hiring engineers.
The focus of DeepSeek's chip development is on hardware for AI inference within data centers, rather than for training models. This strategic shift aims to lessen dependence on both U.S. chipmaker Nvidia, whose access to China is restricted by export bans, and domestic giant Huawei, which currently holds approximately half of China's data center chip market.
DeepSeek is not alone in this endeavor; other major Chinese technology firms such as Alibaba and Baidu are also reportedly exploring custom chip designs. This trend mirrors similar efforts by U.S.-based AI companies. OpenAI, for instance, recently announced its first custom inference chip, Jalapeño, developed in partnership with Broadcom. Anthropic is also said to be exploring custom chip designs.
These moves by AI companies to develop their own silicon are partly aimed at reducing reliance on Nvidia. Additionally, companies like OpenAI seek greater control over their technology stack, similar to Apple's integrated approach. Entering the silicon and data center level also offers a competitive advantage in a market where compute resources are expected to remain constrained as companies scale their AI models and services.