The fears of the CEO of NVIDIA are confirmed: China prepares an avalanche of chips for artificial intelligence

Jensen Huang, the founder and CEO of NVIDIA, made it clear in a statement he made at Computex at the end of May: “China is devoting massive resources to start-ups specializing in GPU development. Don’t underestimate them.” This warning was addressed to the US government in a clear attempt to warn it about the consequences of sanctions that seek to stop China’s technological development.

The Administration led by Joe Biden has expressly prohibited NVIDIA from selling its most advanced artificial intelligence chips, the A100 and H100 models, to Chinese companies . The problem is that for Jensen Huang’s company the huge Chinese market is essential, which is why it has cut the benefits of these GPUs in order to meet the requirements imposed by the US government. The result is the A800 and H800 chips , which are nothing more than simplified revisions of the original GPUs.

Be that as it may, it is evident that in the medium term what interests China is to have its own processors for artificial intelligence applications, and, in particular, for deep learning . This is just what Jensen Huang has predicted. And what, as we are about to see, the Xi Jinping government is orchestrating . Even so, meanwhile, China is buying through official channels the chips that the US-led alliance allows it to acquire and in parallel markets the CPUs and GPUs that the sanctions seek to put out of its reach.

Chinese companies prepare maelstrom of chips for artificial intelligence

The statement in the first paragraph of this article is not the only one that Jensen Huang has made in recent weeks with the purpose of describing the strategy of this gigantic Asian country. This executive also assures that “if China cannot buy the chips for artificial intelligence from the US, it will simply manufacture them itself.” It’s just what you’re doing. In fact, the Asian media confirm that the Chinese machine is advancing relentlessly with the aim of developing its own GPUs for artificial intelligence.

At this juncture, there are dozens of Chinese companies that are engaged in the development of their own hardware for this use scenario. MetaX, Biren Technology, Moore Threads, Innosilicon, Zhaoxin, Iluvatar CoreX, DenglinAI or Vast AI Tech are some of the most important. The interesting thing is that not all of them are following the same path. Moore Threads and Biren Technology are two of the companies that are taking their gaming GPUs as a starting point with the aim of refining them to be competitive when running artificial intelligence processes.

Other companies, such as DenglinAI, Vast AI Tech or MetaX, have taken a different tack: they are fine-tuning their own GPUs designed specifically for artificial intelligence or deep learning applications from scratch. Even a giant like Alibaba has plunged into this industry, taking advantage of the financial backing of the Chinese government through its subsidiary T-Head. Of course, its strategy, curiously, involves fine-tuning dedicated chips whose architecture does not expressly adhere to that of a GPU. China needs to have its own chips for artificial intelligence, and extreme diversification is its best bet to get them.