Nvidia ships first H200 chips to China, forecasts zero data-centre revenue
Synopsis
Nvidia has shipped H200 AI chips to China for the first time under a US licence, but the sales are less than 1% of its US$89B quarterly data-centre revenue — and its next-quarter forecast still banks on zero contribution from China.
Key Takeaways
Nvidia made its first H200 data-centre chip shipments to China under a US licensing scheme, ending a multi-quarter export freeze.
The China sales accounted for less than 1 per cent of Nvidia 's US$89 billion data-centre revenue in the second quarter ended July 26, 2026 .
Nvidia 's US$108 billion third-quarter revenue forecast assumes no data-centre computing revenue from China .
Washington approved H200 exports in January 2026 ; Beijing only permitted selected Chinese AI companies to buy limited quantities from last month.
CEO Jensen Huang said real demand exceeds the company's guided 70 per cent revenue growth for fiscal 2028 , with supply — not orders — as the binding constraint.
Huang said hyperscalers account for only about half of demand; enterprises, neo-clouds, and sovereign AI projects make up the rest.
Nvidia has confirmed its first shipments of H200 data-centre processors to China under a new US licensing framework, ending a months-long export lockout — yet the sales contributed less than 1 per cent of the company's US$89 billion data-centre revenue for its second quarter ended July 26, 2026. Despite the symbolic breakthrough, Nvidia's US$108 billion third-quarter revenue forecast assumes zero data-centre computing revenue from China.
From Lockout to Limited Access
The H200 disclosure marks a clear departure from the previous quarter, when Nvidia reported no shipments of Data Centre Hopper products to China at all. Washington approved H200 exports in January 2026, but Beijing initially restricted purchases, actively steering domestic buyers toward home-grown chip alternatives. Only from last month did selected Chinese AI companies gain permission to buy limited quantities, according to the company's disclosures.
Why It Matters
The negligible revenue contribution underscores how geopolitical friction continues to suppress what should be one of Nvidia's largest addressable markets. China was historically a significant buyer of high-end Nvidia accelerators before successive rounds of US export controls — covering the A100, H100, and earlier H200 variants — effectively shut the market. The resumption, even at minimal volumes, signals a cautious diplomatic recalibration rather than a structural reopening.
Global Demand Outlook Remains Bullish
Setting aside the China drag, Nvidia struck a notably confident tone on worldwide demand. The company said it expects revenue to grow approximately 70 per cent in its 2028 financial year, characterising the forecast as supply-constrained rather than demand-constrained. CEO Jensen Huang told analysts on an earnings call webcast on Wednesday that actual demand was 'much greater than 70 per cent', with available supply — not orders — setting the ceiling.
Hyperscalers Are Only Half the Story
Jensen Huang said the company had secured significant manufacturing capacity but still needed 'a lot more.' He noted that hyperscalers represented only about half of the total opportunity, with the remaining demand coming from enterprises, neo-cloud providers, and sovereign AI projects. 'The unconstrained [growth] would be a lot higher,' Huang said, signalling that Nvidia's bottleneck is production, not appetite.
What's Next
Investors and analysts will watch whether Beijing expands the pool of approved Chinese AI buyers beyond the current select group, and whether Washington tightens or loosens the licensing regime in response to evolving trade negotiations. Any meaningful shift in China policy — in either direction — could materially alter Nvidia's revenue trajectory, given the company's own admission that demand there remains structurally large but politically gated.
Point of View
Not a commercial reopening. The sub-1% revenue contribution also exposes a deeper structural shift: China's deliberate push toward domestic alternatives has already eroded the reflexive buying behaviour that once made it a top-tier Nvidia customer. Mainstream coverage focuses on the symbolic 'first shipment' milestone, but the more consequential signal is Jensen Huang's supply-constrained framing — it means Nvidia has so much demand elsewhere that China's political gatekeeping barely registers on the growth curve. The real risk is the inverse: if Washington tightens controls again, Nvidia loses little in the near term, but if Beijing fully reopens, the upside could be enormous and sudden.
NationPress
27 Aug 2026
Frequently Asked Questions
Has Nvidia started selling H200 chips in China?
Nvidia confirmed its first H200 data-centre chip shipments to China under a US government licensing scheme, ending a prolonged export freeze. However, the sales were less than 1 per cent of the company's US$89 billion quarterly data-centre revenue.
Why did Nvidia previously stop selling chips to China?
Successive rounds of US export controls restricted Nvidia from selling its most powerful accelerators — including the A100 , H100 , and earlier H200 variants — to China . Washington approved new H200 exports in January 2026 , but Beijing initially blocked purchases to promote domestic chip alternatives before selectively allowing limited buys last month.
What is Nvidia's revenue forecast for Q3 2026?
Nvidia issued a US$108 billion third-quarter revenue forecast that explicitly assumes no data-centre computing revenue from China . The company characterised its broader 2028 financial year growth target of approximately 70 per cent as supply-constrained.
What did Jensen Huang say about Nvidia's demand outlook?
CEO Jensen Huang said actual demand was 'much greater than 70 per cent' growth, with supply — not orders — limiting the company's guidance. He added that hyperscalers represent only about half the total opportunity, with the rest coming from enterprises, neo-cloud providers, and sovereign AI projects.
Who are Nvidia's biggest customers beyond hyperscalers?
According to Jensen Huang , roughly half of Nvidia 's addressable demand comes from sources other than major cloud providers. These include enterprise customers, neo-cloud operators, and national sovereign AI programmes — a customer mix that diversifies Nvidia 's revenue base beyond the traditional hyperscaler dependency.
