$4.5 Billion Charge Highlights Cost of Chip Curbs on Nvidia
Key Facts
Nvidia said in its annual filing that U.S. restrictions on chip exports to China cost the company a $4.5 billion charge in the first quarter of its fiscal 2026. The company tied the amount to new licensing requirements it was notified of in April 2025 for its H20 chip, and to excess inventory and purchase obligations. The cost illustrates the impact of the restrictions on the company’s business in China, where it says it cannot supply data centers with a competitive product approved by both the U.S. and Chinese governments. Nvidia said its exclusion from the market benefited competitors in building ecosystems of developers and customers. The initial U.S. restrictions cited by the company date to August 2022, when they targeted China’s semiconductor and supercomputing industries.
Nvidia said the U.S. government notified it in April 2025 that a license would be required to export H20 chips to China. The company said the requirement resulted in a $4.5 billion charge in the first quarter of its fiscal 2026, tied to excess inventory and purchase obligations. In August 2025, the government granted licenses allowing certain H20 products to be shipped to some customers in China. Nvidia generated about $60 million in H20 revenue under those licenses during that month. The difference between the charge and the revenue reflects the limited activity permitted by the licenses, according to figures in the company’s filing.
For its H200 program, Nvidia said it obtained a license in February 2026 allowing small quantities to be shipped to selected customers in China. The company added that it had generated no revenue under the program as of the filing date. This is despite the license being granted, placing the program at a different stage from licensed H20 shipments, which generated revenue in August 2025. The filing did not give a date for when H200 revenue might begin or the volume of shipments that could follow. Shipments remain subject to the license terms granted to the selected customers.
Nvidia said current U.S. and Chinese restrictions prevent it from offering data centers in China a competitive product approved by both governments. It added that its absence from the market gave competitors an opportunity to build relationships with developers and customers. The company warned that the prospect of additional export controls had negatively affected demand for its products and could continue to do so in the future. It said those controls could benefit competitors offering alternatives less exposed to restrictions. The company thus links regulatory restrictions to demand for its products and competitors’ position in the Chinese market.
The U.S. restrictions targeting China’s semiconductor and supercomputing industries began in August 2022, according to Nvidia’s annual filing. The measures included licensing requirements and restrictions on exports of chips and related technologies. Since then, the company said, licensing requirements have affected different products, including H20 and then H200. The figures in the filing show a mixed trajectory: a $4.5 billion charge following H20 requirements, then about $60 million in revenue under specific licenses, compared with no revenue recorded under the H200 program as of the filing date. The company also said it currently cannot offer Chinese data centers a competitive product approved by both sides.