You can’t always get what you want. Well, for China, that may not be the case. Chinese universities and research centers recently got their hands on banned AI Nvidia chips via resellers, Reuters reported April 23. But why would universities in China need these hefty chips in the first place? IT Brew caught up with Neil Sahota, an AI advisor to the United Nations, to find out.
“I’m not surprised,” he told IT Brew. “It’s quite possible that they got them before the ban went into effect, or some resellers tried to skirt the ban afterward. So, there could be a very good reason the universities need them.”
The chips, acquired by 10 Chinese entities, were “embedded in server products made by Super Micro Computer Inc., Dell Technologies Inc., and Taiwan’s Gigabyte Technology Co. Ltd.,” the Reuters article also said.
IT Brew has reached out to the above companies for comment.
Class in session. China has some of the top universities in the AI industry, with Beijing-based Tsinghua University ranking at No. 1 for the best universities for artificial intelligence along with four other Chinese universities ranking in the top 10 worldwide, according to US News & World Report.
This AI tech could very well be used for research, Sahota said, noting that certain processes require “pretty significant computing power.” “If you think about things, like cracking the human immune system or doing genomics analysis—those things require some pretty heavy computing power.”
The US–China chip war means China has been looking everywhere but the US for its chips—including homegrown options and specialized chips for the China market, as IT Brew previously reported.
Top insights for IT pros
From cybersecurity and big data to cloud computing, IT Brew covers the latest trends shaping business tech in our 4x weekly newsletter, virtual events with industry experts, and digital guides.
“The truth…is everyone has ignored hardware and infrastructure, even though this is the backbone of all the stuff that works,” Sahota said.
Sahota, who is also the author of Own the AI Revolution, explains that an entity that has access to these powerful chips may claim some of the chips are faulty in order to keep them for resale.
It could play out something like this, according to Sahota: Jane Doe works for a company in the US. The company has ordered 1,000 chips, but Jane says that 80 of those chips are defective. From there, Jane may “pocket the chips” and fly to China to resell.
“It’s not illegal to sell them inside the Chinese market by Chinese law,” he said. “So, it’s not that difficult to do that. You’re not supposed to do those things. There’s no way to control that.”
Déjà vu. Back in January, Chinese military, state-run AI research centers, and universities were also able to purchase “small batches of Nvidia semiconductors,” that the US had restricted from being exported to China, according to Reuters.
In a military setting, TPUs—or tensor processing units—can “help control 10,000 tanks, whereas the other chips out there, you can only control 1,000. So that’s a big difference in terms of coordination and military capability,” Sahota said.
Ken Brown, the PR director at Nvidia, declined to comment.