Report says DeepSeek is developing its own AI data center chip

In short: Reuters reports that DeepSeek is planning to design its own chip for running AI in data centers.

What happened

DeepSeek, a Chinese startup known for building large language models (AI systems that generate and understand text), is developing its own AI chip, according to Reuters. Reuters cited three people familiar with the plan.

The report says DeepSeek has been working toward this move for about a year. It has been meeting with possible partners in the chip and hardware world, and hiring engineers to help.

DeepSeek’s chip effort is aimed at “inference,” which means running an AI model after it has already been trained. Think of it like using a finished recipe to cook meals, not testing and perfecting the recipe in the first place.

The goal appears to be reducing reliance on both Nvidia and Huawei. Nvidia chips are widely used for AI in the US and Europe, but US export controls have limited what Nvidia can sell into China. In China, Huawei has a large share of the data center AI chip market, and other Chinese companies like Alibaba and Baidu are also working on their own chips.

Why it matters

AI companies are starting to treat chips like a key part of their business, not just something they buy off the shelf. If more companies design their own chips, it could change who controls the supply of the “engines” that power AI services, especially in places where access to advanced chips is restricted.

Source: Arstechnica