DeepSeek Reportedly Developing Own AI Inference Chip

APPSO · wechat · 2026-07-07

According to Reuters sources, DeepSeek is developing its own AI chip for inference scenarios to decrease dependence on Nvidia. Initiated roughly a year ago, the project is still in its early phases. DeepSeek has started contacting chip design, wafer foundry, and memory manufacturers, while privately ramping up the recruitment of chip design engineers. As the cost bottleneck for large models shifts from training to inference, OpenAI released its first inference chip, Jalapeno, in collaboration with Broadcom last month, and Anthropic is also considering custom silicon. For top AI labs, controlling compute hardware has emerged as the next major battleground.

Related event: DeepSeek Reportedly Developing Own AI Inference Chip to Reduce External Reliance(16 posts)→

Original post →

More from Companies & People

Companies & People channel →