Nvidia faces significant challenges following the White House’s decision to impose export restrictions on its H20 GPU, primarily intended for the Chinese market. This regulation is an extension of previous policies aimed at controlling the accessibility of AI chips and advanced AI technologies.
Export Controls and Regulatory Framework
The US Bureau of Industry and Security (BIS) introduced the Framework for Artificial Intelligence Diffusion on January 15, 2025, which establishes export administration regulations for advanced semiconductors. These controls specifically limit the export, re-export, and transfer of high-capacity computing integrated circuits to selected destinations and users. The BIS aims to mitigate risks associated with state and non-state actors potentially accessing advanced AI models.
Implications for Nvidia
Nvidia’s Vice President for Government Affairs, Ned Finkle, criticized the export regulations, connecting them to efforts by the Biden Administration that he claims undermine American technological supremacy. In a blog post, he denounced the extensive regulations as an overreach that could restrict how American semiconductor products are designed and marketed globally. He expressed concern that these controls might harm competition and fail to enhance national security.
Market Impact
The new restrictions are set to start on April 15, with Nvidia confirming that it must secure licenses to export its H20 chip to China. During a Q4 2025 earnings call, Nvidia’s CFO Colette Kress indicated that the company’s datacentre revenue from China remains significantly lower than pre-regulation levels. She suggested that, without regulatory changes, shipments to China would likely continue at current reduced levels, highlighting the fierce competition within the Chinese datacentre solutions market.
Performance of the H20 Chip and Competitors
The H20 chip, while less powerful than Nvidia’s other AI accelerators, had reportedly sold a million units to Chinese customers in 2024. However, competition is rising, particularly from Huawei, which has been increasing production of its own AI chip, the Ascend, though it managed to ship only 200,000 units the previous year. Industry experts, like Antonia Hmaidi from the Mercator Institute for China Studies, point out flaws within China’s semiconductor sector and note that Huawei’s software is still trailing behind Nvidia’s offerings, affecting developer adoption rates.
Nvidia’s Future Plans
Despite these setbacks, Nvidia is pursuing ambitious growth. The company announced plans to invest in up to $500 billion worth of AI infrastructure in the US over the next four years, establishing partnerships with notable manufacturers such as TSMC, Foxconn, Wistron, Amkor, and SPIL. Production of Nvidia’s Blackwell chips has commenced at TSMC’s facilities in Phoenix, Arizona, while supercomputer manufacturing sites are being developed in Texas, with hopes to scale operations within the next 12 to 15 months.