HONG KONG/SHENZHEN, China, June 20 (Reuters) – Psst! Where can a Chinese buyer buy high-end Nvidia? (NVDA.O) AI chips in the wake of US sanctions?
Visiting the famous Huaqiangbei electronics area in the southern Chinese city of Shenzhen is a good bet, particularly the SEG Plaza skyscraper, whose first 10 floors are lined with shops selling everything from camera parts to even drones. The chips are not advertised but asking discreetly works.
They are not cheap. Two vendors there, speaking to Reuters in person on condition of anonymity, said they could provide small quantities of A100 artificial intelligence chips made by the US chip designer, priced at $20,000 a piece, double the usual price.
While buying or selling high-end US chips is not illegal in China, US export restrictions have created a de facto underground market with sellers willing to avoid scrutiny by US authorities. or Chinese.
President Joe Biden’s administration in September ordered Nvidia stop exporting its two most advanced chips, the A100 and newly developed H100, to mainland China and Hong Kong, as part of efforts to hamper China’s AI and supercomputing development amid escalating political tensions and commercial. That was then followed up with a series of semiconductor-related export controls.
But, as AI grows around the world following the runaway success of OpenAI’s ChatGPT, demand for high-end chips has skyrocketed, particularly for Nvidia’s microprocessors, which are widely regarded as the best at handling of machine learning tasks.
“We’re talking to two vendors now to get some,” said Ivan Lau, co-founder of Hong Kong’s Pantheon Lab, who is trying to buy 2-4 new A100 cards to run the startup’s latest AI models.
Those vendors, who bought the chips from outside the US, were quoting HK$150,000 ($19,150) per card, he said, adding: “We were told directly that there will be no warranty or support.”
Reuters spoke to 10 suppliers in Hong Kong and mainland China who described that they can easily purchase small numbers of A100s. Their information highlighted both the strong demand in China for the chips and the relative ease with which Washington sanctions can be circumvented for small-batch transactions.
Reuters was unable to estimate the overall volumes of Nvidia A100 and H100 chips flowing into China or to what extent the transactions taking place will meet demand.
The buyers are typically app developers, startups, researchers or gamers, said the sellers, who declined to be identified because the imports contravene US trade restrictions. One seller said the buyers also included local authorities. Chinese.
Nvidia said in a statement to Reuters that it did not allow exports of the A100 or H100 to China, instead providing reduced-capacity substitutes that comply with US law.
“If we receive information that a customer is breaching their agreement with us and exporting restricted products in violation of the law, we will take immediate and appropriate action,” the statement said.
A spokesperson for the US Commerce Department said in a statement to Reuters that the export control measures have had a “substantial impact” on the availability of high-end chips in China.
The statement also said that reports of parties seeking to obtain these chips through illicit means were “not a surprise,” adding that “reports of violations are being investigated.”
China’s State Council Information Office and China’s Ministry of Industry did not respond to requests for comment.
Nvidia said in September that $400 million in sales could be lost during its third quarter if Chinese companies decided not to buy alternative Nvidia products.
Your new design tailored to China slower variants – the A800 and H800 – developed to cushion that impact are now being bought by big Chinese tech firms like Tencent Holdings (0700.HK) and alibaba (9988.HK)who have deep pockets to buy large quantities.
ONLINE OFFERS
Chinese suppliers said they procured the chips primarily in two ways: by snatching up excess stock that hits the market after Nvidia ships large quantities to big US companies, or by importing through locally incorporated companies in places like India, Taiwan and Singapore.
This means that the amounts they can secure are small, far from what is needed to build a large, sophisticated AI language model from scratch.
A model similar to OpenAI’s GPT would require more than 30,000 Nvidia A100 cards, according to research firm TrendForce. But a handful can run complex machine learning tasks and improve existing AI models.
According to an electronics procurement website that listed about 40 A100 sellers, most were located in the Huaqiangbei electronics area. But A100 listings can also be found on Alibaba. (9988.HK) Taobao’s e-commerce site, on Xiaohongshu, which is similar to Instagram, as well as Douyin, the Chinese version of TikTok.
Alibaba, Xiaohongshu and Douyin owner ByteDance did not respond to requests for comment.
Some of the vendors warned that fraud had become common with refurbished chips posing as A100.
Nvidia’s more advanced H100 chips, only on the market since March, seem much harder to come by.
Vinci Chow, an economics professor at the Chinese University of Hong Kong, whose department has procured four A100 cards from local vendors for research purposes, said he had been told packs of eight H100 chips could be purchased. But only one of the 10 vendors Reuters spoke to said it could purchase H100s.
The United States probably won’t be too bothered by small chip transactions, said Charlie Chai, a Shanghai-based analyst at 86Research.
“Only if/when China poses a greater threat after significant progress will we see stricter enforcement,” he said.
He added that the premiums currently commanded by Chinese suppliers for the A100 and H100 chips could collapse in the future, as many of the Chinese AI startups that were driving purchases would eventually pull out of the market.
($1 = 7.8307 Hong Kong dollars)
Reporting by Josh Ye in Hong Kong, David Kirton in Shenzhen, and Chen Lin in Singapore; Additional reporting by Fanny Potkin in Singapore and David Shepardson in Washington; Edited by Brenda Goh and Edwina Gibbs
Our standards: The Thomson Reuters Trust Principles.
Discover more from PressNewsAgency
Subscribe to get the latest posts sent to your email.