DeepSeek is getting into the silicon business. The Chinese AI startup — the one building large language models that compete with OpenAI and Anthropic — is working on its own chips, according to Reuters.
Three people familiar with the matter say DeepSeek has been working on this for about a year. They’ve been meeting with potential hardware partners and hiring engineers for the project. The focus is on data center chips for inference, not training.
The goal is straightforward: reduce reliance on both Huawei and Nvidia. US export controls have already blocked Nvidia from selling its high-end chips in China. Huawei now controls about half of the data center chip market there. DeepSeek doesn’t want to depend on either one.
They’re not alone. Chinese tech giants Alibaba and Baidu are making similar moves. But it’s not just a China story. OpenAI and Broadcom just announced Jalapeño — OpenAI’s first custom chip for inference — a couple weeks ago. Anthropic has been exploring custom chip design too, though nothing public yet.
For OpenAI, it’s partly about reducing Nvidia dependence. But it’s also about having Apple-like control over the full tech stack. And in a market where data center access is constrained and multiple companies are competing for compute, owning your own silicon is a serious advantage.
DeepSeek hasn’t confirmed the chip plans publicly. But the writing’s on the wall — if you’re an AI company and you don’t control your hardware, you’re at someone else’s mercy.
