DeepSeek Reportedly Plans Its Own AI Chip to Reduce Dependence on NVIDIA and Huawei

DeepSeek reportedly plans to build its own AI chip as the Chinese startup looks to reduce its dependence on NVIDIA and Huawei for running its popular AI models. The chip will focus on inference, which means it will help generate responses from already trained models rather than train new AI systems from the beginning.

Reuters reports that DeepSeek has already started hiring chip design engineers in recent months, though the company has kept the process private instead of posting roles publicly. The move shows that DeepSeek wants more control over the hardware behind its AI services as demand for its models continues to grow.

DeepSeek Wants More Control Over AI Hardware

DeepSeek currently uses AI chips from Huawei and NVIDIA, but relying on outside suppliers creates cost, supply, and performance challenges. By designing its own inference chip, the company can build hardware that fits its own AI workloads more closely and reduce pressure from the wider chip supply chain.

However, the plan comes with major problems. Building an AI chip takes years, and the tape-out process alone can cost a large amount of money before mass production even begins. DeepSeek also has to work within China’s current chip manufacturing limits because U.S. sanctions restrict access to advanced tools and manufacturing equipment.

Manufacturing Limits Remain a Big Challenge

The report says DeepSeek will likely depend on SMIC’s older 7nm process for production, which puts it behind NVIDIA chips made using more advanced TSMC technology. The company also cannot access ASML’s most advanced EUV machines, which makes catching up with global chip leaders even harder.

DeepSeek has also reportedly raised $7 billion in funding, taking its valuation to around $52 billion to $59 billion. That money gives the company more room to pursue custom silicon, but success still depends on design quality, manufacturing access, and long-term execution.

For now, DeepSeek’s chip plan remains at an early stage, but it signals a larger shift in China’s AI market as major companies try to build their own hardware instead of depending fully on NVIDIA or Huawei.