AI Policy Watch

Washington’s New Chip Ban Slams Nvidia, Reshapes Global AI Race

• By John K. Waters
• 04/16/2025

The United States government just made life a lot harder for Nvidia—and by extension, the entire AI hardware ecosystem. In a fresh escalation of tech tensions with China, the Trump administration has imposed sweeping export restrictions on high-performance chips, and Nvidia's AI-optimized H20 processor is squarely in the crosshairs.

The move, confirmed by Nvidia in a filing to the SEC, requires the company to obtain a license to ship the H20—or any chip with similar bandwidth capabilities—to China, Hong Kong, and Macao. That license doesn't just apply to Chinese buyers. It also extends to any company with headquarters or ownership ties to China, as well as to entities in U.S.-designated "D5" countries with national security concerns. The goal, according to U.S. officials, is to prevent advanced American chips from powering Chinese supercomputers with potential military applications.

The impact? Nvidia expects to eat a $5.5 billion charge this quarter alone—an eye-watering sum that covers inventory, purchase commitments, and associated reserves. And that's just the beginning.

The directive, which Nvidia was told would remain in place "indefinitely," effectively kneecaps its workaround strategy. The H20 is one of three export-compliant GPUs Nvidia developed in late 2023 and 2024, alongside the L20 and L2, as a workaround to U.S. government restrictions on selling powerful AI chips like the A100 and H100 to China. That plan is now dead on arrival.

The financial hit sent Nvidia stock tumbling 6.3% in after-hours trading Wednesday. AMD, which also took collateral damage from the new rules (its MI308 chip is similarly affected), dropped nearly 6% as well. Investors wiped out more than $148 billion in Nvidia's market cap in a matter of hours.

"The U.S. export restrictions on Nvidia's H20 chips highlight the growing geopolitical uncertainty enveloping the tech and semiconductor sectors, particularly under Trump-era-style policy reversals," Michael Ashley Schulman, chief investment officer at Running Point Capital, told Reuters. "This unpredictability rattles businesses and investment markets, as evidenced by Nvidia's selloff this morning and broader pressure across chip stocks."

It's not just Nvidia and AMD feeling the chill. AI-adjacent chipmakers like Broadcom and Micron, as well as semiconductor design company Arm, saw their shares drop between 2.5% and 4.6%, suggesting the market is recalibrating for a new era of semiconductor geopolitics—one in which Washington has a much heavier hand on the throttle.

What’s at stake here isn’t just sales figures or quarterly projections. Nvidia made roughly 13% of its revenue from China last year—down from 21% the year before—but those billions helped bankroll its meteoric rise as the hardware backbone of the global AI boom. Cutting off access to that market doesn’t just hurt margins; it slows innovation and deepens the rift between East and West in the race to dominate artificial general intelligence.

But as one door closes, Nvidia is scrambling to open another—this time, on U.S. soil.

In a parallel push to secure its supply chain and scale up for the long-term AI arms race, Nvidia recently announced it will manufacture its AI supercomputers in the United States for the first time. The company is collaborating with TSMC, Foxconn, Wistron, Amkor, and SPIL to build a domestic manufacturing footprint across Arizona and Texas. Blackwell chips are already in production at TSMC facilities in Phoenix, while supercomputer assembly is ramping up in Houston and Dallas. Mass production is expected within the next 12 to 15 months.

By leveraging its own technologies—including NVIDIA Omniverse for digital twins and Isaac GR00T for robotics automation—Nvidia is effectively building AI factories to create the infrastructure that will power even more AI factories. CEO Jensen Huang framed the move as a historic pivot: "The engines of the world’s AI infrastructure are being built in the United States for the first time."

The company expects to generate up to $500 billion worth of U.S.-made AI infrastructure over the next four years, which would create hundreds of thousands of jobs and embedding resilience into its core operations.

This dual narrative—getting squeezed abroad while scaling up at home—underscores a new phase of the global AI race. In the geopolitics of code and silicon, no one stays untouchable for long. But Nvidia is betting that with the right mix of domestic muscle and manufacturing ambition, it won’t just weather this storm—it might power right through it.