Beijing (Reuters): A review by Reuters showed that China’s top universities and state-run research institutes have been using a U.S. computer chip to power their artificial intelligence (AI) technology. However, the U.S. government has now made it harder for the chip to leave the country.
Nvidia (NASDAQ:NVDA), a company that makes chips in the United States, said last week that the U.S. government has told it to stop sending its A100 and H100 chips to China. Advanced Micro Devices Inc. (NASDAQ:AMD), a local company, also said that its advanced AI chip MI250 can’t be sent to China because it needs a new licence.
Also Related: Analysis: The public is angry about the BOJ’s dovish isolation before a change in leadership.
The news was a big step up in the U.S.’s campaign to slow down China’s technological progress, which is happening as tensions rise over what will happen to Taiwan, where chips for Nvidia and almost every other major chip company are made.
China thinks of Taiwan as a rebel province, and it hasn’t ruled out using force to take control of the democratically run island. In response to the restrictions, China said that they were a useless attempt to block a rival’s access to technology.
Reuters looked at more than a dozen public government tenders from the past two years and found that some of China’s most important research institutes have a high demand and need for Nvidia’s A100 chips.
Also Related: Analysis: Turkey’s currency nightmares keep coming back.
One of the tenders showed that China’s top university, Tsinghua University, spent more than $400,000 last October on two Nvidia AI supercomputers, each of which was powered by four A100 chips.
The Institute of Computing Technology, which is part of the top research group Chinese Academy of Sciences (CAS), bought about $250,000 worth of A100 chips in the same month.
In July of this year, the school of artificial intelligence at a CAS university spent around $200,000 on high-tech equipment, such as a server powered in part by A100 chips.
Also Related: Analysis: Italy’s “drugged” building boom runs into a wall.
Jinan University in Guangdong, China, spent more than $93,000 on an Nvidia AI supercomputer in November. Just last month, its school of intelligent systems science and engineering spent almost $100,000 on eight A100 chips.
The tenders showed that less well-known institutes and universities, such as those in Shandong, Henan, and Chongqing, which were helped by local and provincial governments, also bought A100 chips.
When asked what the A100 export limit would mean for their projects, none of the research departments replied.
Nvidia didn’t answer when asked for a comment. Last Wednesday, the company said that it had sold $400 million worth of the affected chips in China this quarter. This money could be lost if customers don’t buy other Nvidia products. It also said that it would try to get out of some of the new rules.
Also Related: Analysis-Payments businesses brace for M&A after value freefall.
REPLACEMENTS
Without chips from companies like Nvidia and AMD, Chinese organisations may not be able to do advanced computing tasks like image and speech recognition in a cost-effective way.
Consumer apps like smartphones that can answer questions and label photos often use image recognition and natural language processing. They can also be used by the military to do things like search satellite images for weapons or bases and filter digital communications to get information.
Experts said that there aren’t many Chinese chipmakers that could easily replace advanced Nvidia and AMD chips. Instead, they said, buyers could use a number of lower-end chips to get the same amount of processing power.
Also Related: Analysis-Musk investigates the boundaries of governance by using children as guinea pigs.
Reuters couldn’t find any Chinese government bids for the other two restricted chips, the H100 from Nvidia and the MI250 from AMD.
But some of the tenders showed plans to buy chips from U.S. tech company Intel Corp (NASDAQ:INTC) and less advanced products from Nvidia. This showed how dependent China is on a wide range of U.S. chip technology.
In May, a tender showed that the Chinese Academy of Surveying and Mapping, which is part of the Ministry of Natural Resources and does research, was thinking about getting an Nvidia AI supercomputer to help it make better three-dimensional maps from geographic data.
“The proposed NVIDIA DGX A100 server will have 8 A100 chips with 40GB memory,” the tender said. “This will greatly improve the data-carrying capacity and speed of computing, shorten the time it takes to do scientific research, and get better and faster results from scientific research.”
The National University of Defense and Technology (NUDT), which calls itself a “military university” and says it is “under the direct leadership of the Central Military Commission,” China’s top military body, is also one of the buyers of A100 chips.
Also Related: The CEO of Enel is waiting to hear from Lukoil about the sale of its Russian unit.
Since 2015, the NUDT, which is home to one of the most powerful supercomputers in the world, Tianhe-2, has been on a U.S. blacklist because of national security concerns. This means that the university can’t get the Intel processors it needs to run its supercomputers.
In a May bid, the institute said it wanted to buy 24 Nvidia graphics processing units that could be used for AI. Last month, the tender was put out again, which showed that NUDT still hadn’t found the right deal or supplier.
NUDT didn’t answer when asked for a comment.
