Nvidia has permission to ship some H200 AI chips to China. It has yet to turn that permission into significant sales.
In its latest quarterly filing, the company said H200 shipments under the U.S. licensing program accounted for less than 1% of data center revenue in the quarter ended July 26. It had shipped only a fraction of the amount permitted by its licenses.
That gap matters for Nvidia stock as Chinese President Xi Jinping begins a U.S. visit scheduled for September 23–25. A meeting between Xi and President Donald Trump may bring chip trade back into focus, but there is no announced agreement that would expand Nvidia’s sales.
Why Aren’t the Chips Selling?
A U.S. export license addresses only one side of the transaction. Nvidia said Chinese government restrictions have prevented it from selling all the H200 products covered by its licenses.
There is a cost problem, too. Under the current process, licensed chips must pass through the United States for inspection before reaching customers. Nvidia says that makes them subject to a 25% U.S. import tariff, which it has been unable to pass on to buyers. The company recorded a $400 million charge in the first half of its fiscal year tied to excess H200 inventory and purchase obligations as demand weakened.
What Would Change the Stock Story?
Nvidia’s broader AI business is growing without a China rebound. Data center revenue reached $89 billion in its latest quarter, up 117% from a year earlier. Its outlook for the following quarter excluded China data center computing revenue because of geopolitical uncertainty.
Investors already have other ways to follow the AI chip trade: AMD recently crossed $1 trillion in market value, while Nvidia’s growth is also driving a much wider AI infrastructure financing boom.