The US wants China to become addicted to Nvidia chips. His problem is that his greatest rival is not for the work

The administration led by Donald Trump has realized that her predecessor and herself have made a very serious mistake. SANCTION PACKAGES that have deployed during the last three years To prevent China from doing with the most advanced semiconductor manufacturing equipment, and also with avant -garde chips for artificial intelligence (AI), have forced the Government of Xi Jinping to become independent of foreign technologies.

The Chinese industry of the integrated circuits has advanced a lot during the last five years, and will surely continue to do so. It is very likely that in 2026 Chinese manufacturers have their own extreme ultraviolet lithography equipment (UVE). And currently Huawei, Moore Threads and other Chinese companies have GPU for some scenarios They compete with the Nvidia and AMD chips. The independence of the Chinese semiconductor industry is underway.

The best output given this juncture for the US is none other than to deliver to China in a controlled way advanced chips for AI, but less powerful than the most capable that design Nvidia, AMD or brains. In this way this Asian country may relax a bit its ambition for development and independence. This is exactly what the Trump administration is doing by allowing Nvidia to give back to its Chinese clients Your GPU for IA H20as Chris Miller holdsthe author of ‘The chips war’in his Newsletter.

China is getting out of Nvidia

The future of Nvidia in China is largely in the hands of the administration of the cyberspace of China, known as CAC for its English denomination (Cyberspace Administration of China). This institution is the main Internet regulatory body in China and is thoroughly investigating NVIDIA H20 GPU Because it suspects that this chip could incorporate a back door of difficult location by Chinese experts. If so, the possibility of China to use this GPU could be possible.

David reber Jr., Nvidia Security Director, published last week An article in the blog of this company entitled “There are no rear doors in the Nvidia chips. There are no deactivation switches. There are no spy software” in an obvious attempt to defend the company’s reputation and appease the growing distrust to which he faces in China. There is a lot at stake. If the final resolution of the CAC turns out to be unfavorable with all probability NVIDIA will lose a good part of the income currently obtained in this country.

The Chinese government is urging Chinese companies that are dedicated to the development of large models of the use in their Chinese chip servers

However, this is not all. Whatever the CAC verdict is a fact that the Chinese government He is urging Chinese companies that are dedicated to the development of large models of AI to use in their servers integrated circuits of Chinese origin. In this scenario their best bazas are now Huawei and Moore Threads. The first of these companies has lists its own GPU for iathe chips ascend AI, for more than five years. During this period of time it has been refining them and increasing their abilities with the purpose of matching or even overcoming the performance of the chips A100 and H100 of Nvidia.

His most ambitious proposal right now is the chip Ascend 910dwhich seeks to overcome the performance of the H100 GPU. However, this Chinese company has also recently presented its chip Ascend 920a solution that is clearly destined to occupy in the Chinese market The H20 GPU space. This proposal will enter large -scale production during the second half of 2025 using 6 NM integration technology that have presumably developed elbow with Huawei elbow and SMIC.

On the other hand, Moore Threads He has developed several GPU for AI applications that, on paper, rivaize some of the advanced solutions that have placed in the Nvidia, AMD or Huawei market. The MTT S4000 and MTT S3000 cards are its most interesting proposals right now, although, curiously, in its porpholio the MTT S80 card, a proposal for games and content creation that, according to Moore Threads itself, has a 14.4 TFLOPS calculation capacity also appears in Floating Coma operations of simple precision. It doesn’t impress, but it’s not bad at all.

Image | Gage Skidmore | Wikipedia

More information | Financial Times

In Xataka | Ten Chinese companies in Chips and IA have allied with a common goal: to put an end to the domain of Nvidia

Leave your vote

Leave a Comment

GIPHY App Key not set. Please check settings

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.