The tug-of-war over advanced AI chips between the U.S. and China has taken another turn. For months, Chinese tech companies, facing tightening American export controls, scrambled to acquire processors powerful enough to fuel their AI ambitions.
Nvidia’s H20 chip, a processor specifically designed to comply with U.S. rules while still offering substantial AI capabilities, became a focal point in this scramble, with firms reportedly placing orders worth at least $16 billion in early 2025 alone. But just as Washington appeared ready to clamp down further, the Trump administration abruptly hit pause.
This policy adjustment, reported by NPR, halts planned restrictions on the H20 chip that were reportedly finalized and ready for implementation. The timing is notable, coming directly after Nvidia CEO Jensen Huang attended a high-priced dinner event at Mar-a-Lago the prior week.
Citing two sources familiar with the matter, NPR reported the administration’s change of heart followed a promise from Nvidia regarding new investments in U.S.-based AI data centers. Neither the White House, the Commerce Department, nor Nvidia offered comment on the situation.
Industry Warnings Meet Washington’s Security Apparatus
This unexpected leniency follows months of lobbying by major U.S. tech players concerned about the repercussions of strict export controls.
without effectively stopping China’s AI development. Microsoft’s Brad Smith cautioned that denying U.S. tech could simply push global customers to alternatives and potentially speed up China’s own semiconductor efforts.
If American technology companies are unable to meet demand, businesses and researchers around the world wouldturn to other suppliers instead—whether in Europe, Asia, or elsewhere, Smith argued in a February blog post. He also questioned the feasibility of containment.
Nvidia itself had previously warned that aggressive U.S. policies might unintentionally accelerate China’s path to chip self-reliance. These arguments highlight a growing tension between corporate interests and Washington’s national security aims, which view access to advanced AI chips as a strategic vulnerability.
A Chip Born from Restrictions
The H20 wouldn’t exist without prior U.S. actions. Washington first banned Nvidia’s top-tier A100 and H100 chips from China in October 2022. Nvidia adapted, creating the less powerful A800 and H800 to fit within the rules, only for those to be banned in late 2023. The H20 represents a further downgrade, using the GH100 silicon but with significantly reduced performance density (around 2.9 TFLOPs/die size versus the H100’s 19.4, according to early 2024 reports).
Reports from early 2024 suggested its FP32 performance lagged behind Huawei’s Ascend 910B, though its interconnect speed might offer advantages in large clusters. Despite the modifications, the H20 remained desirable for AI inference tasks and running models like DeepSeek R1. Its reported price point of $12,000-$15,000 earlier in 2024 reflected its position as a premium, yet compliant, option for the Chinese market.
Political Winds and Policy Questions
The administration’s pause on H20 curbs contrasts with bipartisan calls for even stricter controls. Concerns amplified after Chinese firm DeepSeek demonstrated potent AI capabilities, allegedly using H20 chips. Sens. Elizabeth Warren and Josh Hawley, along with Reps. Raja Krishnamoorthi and John Moolenaar, had urged action.
Krishnamoorthi reacted strongly to the policy halt, stating, “Export controls work, and we don’t have time to waste… Every day America fails to restrict the export of chips designed to circumvent existing controls is a day that our adversaries have to build up stockpiles to defeat us” . Adding complexity, the idea of restricting the H20 wasn’t new; reports indicate it was considered under the previous Biden administration.
Previous Commerce Secretary Gina Raimondo had also previously acknowledged (in late 2023) the dual-use nature of AI chips, suggesting some sales to China for commercial purposes could be permissible within limits. Meanwhile, the agency responsible for these rules, the Bureau of Industry and Security, has reportedly faced internal challenges linked to federal cuts and staff departures under the Trump administration, raising questions about enforcement capacity. For his part, Jensen Huang previously stated Nvidia’s commitment to working with the U.S. government to ensure compliance.