Americas

  • United States

Nvidia eyes Q2 production of delayed AI chip for China

News Analysis
Jan 09, 20243 mins
CPUs and ProcessorsData CenterGenerative AI

November's planned launch of H20, the most powerful of three Nvidia GPUs designed to comply with new US export rules, is now pushed to a limited release in Q2.

nvidia santa clara headquarters
Credit: Nvidia

US chipmaker Nvidia will finally in the second quarter of 2024 begin mass production on an artificial intelligence (AI)-focused chip designed for China, a launch that already has been delayed for several months, according to a published report.

The chipmaker had set back its planned November production timeline for the H20 chip – the most powerful of several designed for Chinese manufacturers to comply with recently set US export rules – due to issues that server manufacturers were having integrating the chip.

That production will now begin in Q2 but in a limited quality, with major customers having first dibs on the H20 GPUs, according to a report by Reuters published Tuesday. H20 along with L20 and L2 are all processors Nvidia is designing to meet new US export restrictions set by the Biden administration in October that limit on-chip performance and density for processors exported to China.

This new release plan for H20 is slightly later than some early projections of when the delayed production might begin, which was first speculated to be pushed to February or March, Reuters previously reported.

A Nvidia spokesperson would not comment on the production timeline but did say that the company is “engaged with the US government and, following its clear guidelines, [plans] to offer compliant product solutions to customers worldwide.”

Changing rules and processors

The rules were set by the US Department of Commerce and Bureau of Industry Security, the latter of which measures the total processing power and performance density data and sets export laws. The restrictions were part of a raft of measures that seek to stop Beijing from receiving cutting-edge US technologies to strengthen its military.

The rules made Nvidia’s latest data center-grade GPUs and many of its high-end retail cards, such as its advanced A800 and H800 AI chips, ineligible for export, spurring the development of modified processors to secure Nvidia’s market position in China.

However, even the forthcoming modified chips don’t rule out a new change in regulation that could ban even H20 and its companion technology for export to China, industry watchers say. In fact, the A800 and H800 themselves were introduced as alternatives for Chinese customers in November 2022 about a month after the U.S. first restricted exports of advanced microchips and equipment to China.

This worry that “another round of restrictions might even make this H20 not compliance-ready in a few months” could even cause some Chinese OEMs to balk at adopting the new chips, noted Guarav Gupta, vice president analyst, emerging technologies and trends, with Gartner Group.

Still, as “there are limited options for China to procure local chips with similar AI capability,” the news that H20 is finally rolling toward production should be positively received – for now, he said.

“That combined with Nvidia’s ecosystem still makes it attractive for the time – [until] China develops its own capabilities,” Gupta said.