Nvidia has developed three chips for the Chinese market as it adapts again to US export restrictions. The H20, L20 and L2 are intended to serve demand for artificial intelligence computing while meeting the controls, according to leaked documents and four people familiar with the matter cited by the Financial Times.
A new round of designs for China
This is the second time in just over a year that Nvidia has had to adjust products for Chinese customers in response to US rules. In 2022, the US government blocked sales of the A100 and H100 GPUs to China. Nvidia then developed the A800 and H800 as alternatives that complied with limits on performance.
More recent restrictions also affected those China-focused models. The source reports that the US government introduced stricter export controls last month and that the Biden administration accelerated their implementation. Nvidia’s latest chips are another attempt to keep products available in China within the changed boundaries.
The sequence shows how export restrictions can prompt a company to redesign products rather than leave a market altogether. Each new limit can make previously compliant hardware unsuitable, requiring manufacturers to reassess what they can offer.
Performance within the limits
The H20, L20 and L2 are expected to deliver less performance than Nvidia chips previously sold in China. Still, sources cited by the Financial Times said the new GPUs should remain competitive in the Chinese market. The report does not provide specific performance figures or details about how the three chips differ.
Analysts at chip consultancy SemiAnalysis described Nvidia’s approach as balancing peak performance with power density to comply with the new US rules. In practical terms, the design challenge is to offer useful computing capacity while staying within regulatory constraints that shape what can be sold.
That balance matters to customers seeking hardware for AI development and deployment. Lower performance than earlier products may affect what systems can do, but the report suggests Nvidia expects the chips to remain commercially relevant in China.
Why Nvidia wants to stay in the market
Large orders from China are a growth driver for Nvidia’s sales, giving the company a business reason to continue serving the market. Nvidia has not commented on the reports about the new chips, according to the source.
The company’s position also rests on more than hardware. The article describes Nvidia as combining development tools, software innovation and chips, creating an ecosystem that could be difficult for competitors to match. That broader offering can matter to buyers building AI computing clusters, since chips are only one part of the systems used to train and run models.
The source says Nvidia is the main supplier of almost all new computer clusters used to train and run AI models, and reports that it has so far faced no serious competition in China or elsewhere. Those factors help explain why access to its products remains valuable even when the available versions have reduced performance.
Constraints on Chinese alternatives
Chinese chipmakers face their own challenges. The source says government regulations require them to produce chips only in China, while they lack access to sophisticated chip production facilities from suppliers such as Dutch company ASML.
These production limits complicate efforts to replace Nvidia hardware with domestic alternatives. The result is a market shaped by both US export controls and constraints on local manufacturing: Nvidia must adapt its products to keep selling, while Chinese firms face barriers to making competing chips.
The H20, L20 and L2 therefore represent more than a new product lineup. They reflect an ongoing effort to preserve access to AI computing in China under shifting trade restrictions, with performance, manufacturing capacity and competition all influencing what buyers can obtain.