Nvidia considers increasing H200 chip output as Chinese demand surges

Nvidia has told Chinese clients it is evaluating adding production capacity for its powerful H200 AI chips after orders exceeded its current output level, according to two sources briefed on the matter. 

The move comes after US President Donald Trump said on Tuesday the US government would allow Nvidia to export H200 processors, its second-fastest AI chips, to China and collect a 25 per cent fee on such sales. 

Demand for the chip from Chinese companies is so strong that Nvidia is leaning toward adding new capacity, one of the sources said. They declined to be named as the discussions are private. 

"We are managing our supply chain to ensure that licensed sales of the H200 to authorized customers in China will have no impact on our ability to supply customers in the United States," an Nvidia spokesperson said in a statement to Reuters after the story was published. 

Major Chinese companies including Alibaba and ByteDance have already reached out to Nvidia this week about purchasing the H200 and are keen to place large orders, Reuters reported on Wednesday. 

However, uncertainties remain, as the Chinese government has yet to greenlight any purchase of the H200. Chinese officials convened emergency meetings on Wednesday to discuss the matter and will decide whether to allow it to be shipped into China, said one of the two sources and a third source. 

Very limited quantities of H200 chips are currently in production, Reuters reported on Wednesday, as the US AI chip leader is focused on producing its most advanced Blackwell and upcoming Rubin lines. 

China promoting own AI industry 

Supply of H200 chips has been a major concern for Chinese clients and they have reached out to Nvidia seeking clarity on this, sources said. 

As part of the briefing provided by Nvidia, the company has also given them guidance on current supply levels, said one of the first two people, without providing a specific number. 

The H200 went into mass deployment last year and is the fastest AI chip in Nvidia's previous Hopper generation. The chip is manufactured by TSMC using the Taiwanese firm's 4nm manufacturing process technology. 

TSMC and China's Ministry of Industry and Information Technology (MIIT) did not immediately reply to requests for comment.

Chinese companies' strong demand for the H200 stems from the fact that it is easily the most powerful chip they can currently access. 

It is about six times more powerful than the H20, a downgraded chip from Nvidia tailored for the Chinese market that was released in late 2023. 

Trump's decision on the H200 comes at a time when China is pushing to promote its own domestic AI chip industry. As domestic chip companies have yet to produce products that match the H200, there have been concerns that allowing the H200 into China could stymie the industry. 

"Its (H200) compute performance is approximately 2-3 times that of the most advanced domestically produced accelerators," said Nori Chiou, investment director at White Oak Capital Partners. 

"I'm already observing many CSPs (Cloud Service Providers) and enterprise customers aggressively placing large orders and lobbying the government to relax restrictions on a conditional basis," he said, adding Chinese AI demand exceeds the capacity of local production. 

During the emergency meetings, there was a proposal to require each H200 purchase to be bundled with a certain ratio of domestic chips, one of the first two sources and a third source said. 

For Nvidia, adding new capacity is also challenging at a time when it is not only transitioning to Rubin but also competing with companies including Alphabet's Google for limited advanced chipmaking capacity from TSMC.