The indiscretion

Nvidia stops production of its chips for China

With new uncertainties accompanying the war in Iran, the company prefers to allocate resources to development for American AI players

by Biagio Simonetta

5 March 2026

2' min read

Translated by AI

Versione italiana

2' min read

Translated by AI

Versione italiana

The long and difficult relationship between Nvidia and China now has a new chapter. Jensen Huang's company has decided to stop the production of H200 chips destined for the Chinese market and to reallocate the production capacity it has within TSMC to the new Vera Rubin architecture.

The decision seems to have a rather logical design: the uncertainty hovering around relations between Washington and Beijing, also in the light of the new war in Iran. The fear is that new restrictions on advanced semiconductors could limit China's access to the most powerful technologies for artificial intelligence.

The move also signals that Nvidia does not expect significant sales of the H200 in China in the short term. And perhaps there is also a technological factor here. The processor, in fact, belongs to the company's previous generation of AI chips. And it was designed to comply with the limits imposed by US regulations on the export of advanced semiconductors. And today, with Chinese chip companies pushing hard on the accelerator to produce increasingly high-performance semiconductors, Nvidia's H200 could be considered an outdated technology even in China.

On the other hand, there is clearly a need to develop the new Vera Rubin generation as soon as possible, which is eagerly awaited by American players such as OpenAI, Google and the other artificial intelligence biggies.

In short, Nvidia's discourse seems to be twofold: optimise resources on the new technology and reduce risks in an indecipherable market like China.

It is worth mentioning that Nvidia had lobbied both Washington and Beijing to get the go-ahead for the sale of the H200 in China in recent months. And after US President Donald Trump had indicated in December that sales could be authorised, the company had increased production in anticipation of strong demand. The Huang-led company, according to FT, was expecting orders for more than one million units from Chinese customers. Then, at the beginning of January, Huang himself had declared that demand was 'very high' and that production was proceeding at full speed, before ending up in a spiral of new bans (later resolved) that created new uncertainties in Nvidia's plans.

It should be added that China is now also considering limits on H200 purchases in order to protect its semiconductor industry and favour the use of domestic chips in artificial intelligence companies.

To date, Nvidia seems to have already produced around 250,000 H200 chips. And should both Washington and Beijing confirm new order limits, the stock already available should be sufficient to cover the demand authorised so far.

A meeting between Chinese President Xi Jinping and Donald Trump is scheduled for the end of March. Some industry analysts speculate that the summit could pave the way for an agreement on chip export controls. But with the war in Iran breaking out in recent days, the scenario is very uncertain.

Biagio SimonettaRedattore
- @biagiosimonetta
- LinkedIn
Lingue parlate: italiano, inglese
Argomenti: Finanza, Fintech, Silicon Valley, Tecnologia
Scheda autore
Trust project