TECH NEWS – After surprising the Americans with artificial intelligence, the Chinese company is now trying to catch up with hardware.

DeepSeek has experienced significant growth in recent months, initially as a side project, and has since shaken up the international AI market with the emergence of its Large Language Model (LLM). It has been able to compete with OpenAI GPT o1 while being limited in terms of computing power.

According to DigiTimes, DeepSeek wants to develop its own AI chips, so after many other mainstream companies, the Chinese want to develop the technology in-house. The publication doesn’t say much about the project, except that the company has started hiring heavily and that the project leaders include semiconductor experts. That’s not so easy, because developing an AI chip requires an extensive supply chain and its processes, which is not so easy for Chinese companies due to international sanctions.

Currently, DeepSeek can only compete with companies like SMIC, but they are lagging behind global chip trends. We now hear that DeepSeek has about 10,000 Nvidia H800 AI GPUs (that would be Nvidia’s hardware for the Chinese market), and on top of that, they have 10,000 of the more expensive H100 chip. Although the U.S. has imposed a number of sanctions on China and its companies, this has not deterred DeepSeek, which may be hiding more secrets behind the scenes. The company also runs inference workloads on Huawei’s Ascend AI chips, giving it a truly diverse arsenal.

DeepSeek’s in-house chip concept is certainly questionable, given that the company hasn’t evolved as much as its competitors (such as OpenAI), but it’s good to see the company exploring the possibility as it increases diversity in the AI market. The key question is whether DeepSeek will make it to the implementation stage…

A lot could happen in the next 1-2 years or so.

Source: WCCFTech, DigiTimes