Chinese AI startup DeepSeek is developing its own artificial intelligence chip for inference workloads, a move that could reduce its reliance on Nvidia and Huawei hardware as US export controls tighten, according to three people familiar with the matter cited by Reuters.
The project remains at an early stage. DeepSeek has begun reaching out to external partners in chip design, foundry, and memory, the sources said. The effort started roughly a year ago, and the company has quietly stepped up hiring of chip-design engineers in recent months without posting roles on public platforms.
Strategic shift and industry trend
The push marks a departure for a company that has historically prioritized model breakthroughs over commercial expansion. DeepSeek, known for its low profile despite becoming China's AI standard-bearer, did not respond to a request for comment. Despite that low profile, DeepSeek has become a central figure in China's AI sector, and developers can access Deepseek AI Training to get hands-on with its models.
The move follows a broader pattern. OpenAI unveiled its own custom inference chip last month, and Anthropic has been weighing a similar step, Reuters previously reported. For DeepSeek, the incentive is particularly sharp. US export controls bar Chinese firms from buying Nvidia's most advanced processors, and Beijing has urged its tech champions to build domestic alternatives.
Export controls and hardware constraints
DeepSeek has used both Nvidia and Huawei chips in the past but has increasingly leaned on Huawei's Ascend line. An in-house chip would give it greater control and potentially lower costs, especially as inference demand soars. However, significant hurdles remain. Designing a competitive AI chip typically takes years and heavy investment. Manufacturing is also constrained because separate US curbs block Chinese companies from accessing the most advanced overseas foundries and high-bandwidth memory, a critical component for inference chips.
News of the effort nudged Nvidia shares down about 2% in premarket trading. The chip push coincides with DeepSeek's first embrace of outside capital, with the company poised to raise $7 billion in a maiden funding round, Reuters reported last month.
Why this matters for IT and Development
For IT and development teams, the shift toward custom inference chips signals a need to stay current with hardware-aware AI deployment. As more AI labs design their own silicon, infrastructure choices and optimization strategies will change. Courses on AI for IT & Development can help professionals adapt to these evolving hardware backends without getting locked into a single vendor's stack.
Your membership also unlocks:
