Nvidia is expected to launch AI chips tailored to China, responding quickly to recent US restrictions on sales of high-end chips in the Chinese market. (Photo I-Hwa Cheng / AFP)
- Nvidia is expected to launch AI chips tailored to China, reacting quickly to recent US restrictions.
- HGX H20, L20 PCIe and L2 PCIe are names of Nvidia chips.
- An official announcement from Nvidia could come as early as next week.
When the US Commerce Department first imposed restrictions on companies in October 2022, stopping them from shipping advanced chips and chip-making equipment to China, it affected Nvidia’s A100 and H100 chips. But it didn’t take long before Nvidia introduced the A800, a scaled-down version of the A100, as a workaround for export restrictions. Subsequently, in March 2023, Nvidia introduced the H800 as a replacement for the banned H100 chips, aligning its performance with criteria set by the Department of Commerce.

Anyone who has read Romeo and Juliet knows that you can’t ban your way to success. Nvidia (Juliet) finds a way to see its Chinese Romeo.
For the US, the goal is simple: curb China’s ability to produce cutting-edge weapons chips and other defense technology. So, in an expected move, the US enacted new regulations in October of this year, immediately restricting exports of both the A800 and H800 chips tailored to China. Nvidia’s initial strategy of an extra 30 days to fulfill multiple orders has become obsolete.
But Nvidia did determined to maintain its supply to China. After all, China’s share of Nvidia’s data center revenue is around 20-25%, which has increased 171% compared to the previous year last quarter, generating more than USD 10 billion. The demand for Nvidia’s chip is too strong to ignore. So much so that the price of Nvidia’s AI chip on the underground market in China last June was high 20,000 US dollarsdouble the retail price.
China’s top internet players alone – Alibaba, Baidu, ByteDance and Tencent – spent a whopping US$1 billion to buy around 100,000 Nvidia A800 processors in August. That means Nvidia won’t give up on Chinese demand easily, so it decided to work on three newer chips to get around the latest US export ban.
“When the US lifted the updated AI restrictions, we thought the US had locked every imaginable loophole. To our surprise, Nvidia still found a way to deliver high-performance GPUs to China with its upcoming H20, L20 and L2 GPUs,” chip industry newsletter SemiAnalysis explained, conveying the news to the public. In the newsletter, SemiAnalysis also shared that Nvidia has product samples for these newer GPUs, which will go into mass production in the next month.

Detailed specifications include FLOPS figures, NVLink bandwidth, power consumption, memory bandwidth, memory capacity, die size and more. Simple specifications are on this table. Source: SemiAnalysis
“Once again, Nvidia demonstrates its mastery of the supply chain,” wrote Dylan Patel, principal analyst at SemiAnalysis. But Nvidia saw its most significant stock drop in months as the Biden administration stepped up its efforts in October. The company also warned that its product development and customer supply capabilities could be at risk due to the recent restrictions.
Under new US export control regulations, Nvidia cannot ship to its main consumer graphics cards for games, RTX 4090, in China, the world’s largest semiconductor market. While Nvidia said it does not foresee an immediate financial impact from the recent restrictions, reports point to the possibility of billions of dollars worth of orders being canceled from Chinese technology companies.
Nvidia’s latest AI chips for China and possible implications
According to Patel, one of the China-specific GPUs is over 20% faster than the H100 in LLM inference and more similar to the new GPU that Nvidia launches early next year than the H100. Spec details on Nvidia’s new GPUs show that AI chip giant “It perfectly crosses the line of peak performance and performance density with these new chips to pass the new US regulations,” he added.
According to the report of Reuters citing a note from Wells Fargo analyst Aaron Rakers, all three of Nvidia’s reported chips appear to be below the absolute limit of computing power, although one appears to be in the gray area and will require a license.
Rakers believes that while the introduction of these three new GPUs is positive, “we would expect investors to wonder if (Nvidia) is being a little too aggressive in its efforts to circumvent US restrictions and could ultimately just result in further moves ( US government) going forward,” Rakers wrote, noting that Nvidia gets about a quarter of its data center chip revenue from China.