DeepSeek Plans to Develop Its Own Chips to Sidestep US Export Controls
Translated by a model; the original is in Russian.
DeepSeek, the Chinese company behind a number of well-known language models, has announced plans to develop its own chips. According to Ars Technica, the decision is tied to US export controls that make it harder for Chinese companies to access cutting-edge American technology.
Reducing dependence
The initiative's main goal is to reduce DeepSeek's reliance on Nvidia chips, which remain subject to US export controls. At the same time, the company is seeking to lessen its dependence on products from the Chinese manufacturer Huawei as well.
By available accounts, the project is in its early stages, and details of its technical implementation have not yet been disclosed.
The context of the restrictions
For several years now, US export controls on shipments of advanced semiconductors to China have remained one of the key factors shaping the country's development of artificial intelligence. Companies are being forced to look for alternative ways of securing the computing power they need to train and run large language models.
DeepSeek previously drew wide attention with the release of models that demonstrated high efficiency at comparatively low training costs. Developing its own chips could be the company's next step in its pursuit of technological independence.
It is still unclear whether DeepSeek plans to design the chips on its own or in partnership with other players in the market, or what timeline it has set for carrying out the initiative.
Source: Ars Technica
Comments
Sign in, to comment.