Just weeks after Washington lifted a short-lived ban on selling AI chips to China, Nvidia has locked in a massive order—300,000 high-end H20 chipsets from Taiwan Semiconductor Manufacturing Company. This move signals not just a comeback in sales, but a surge in Chinese demand that could reshape the balance in the global AI race.
The chips in question? Among the most powerful currently manufactured anywhere in the world. And they’re heading straight to China’s booming artificial intelligence sector, a development many U.S. security analysts see as a double-edged sword.
A Rapid U-Turn in U.S. Policy Leaves Heads Spinning
Back in April, the Biden administration—building on Trump-era restrictions—shut the door on Nvidia’s most advanced chip sales to China. The H20 chips, powerful and versatile, were labeled a national security risk.
That door has swung wide open again. In July, under mounting pressure from industry and geopolitical allies alike, the White House quietly gave Nvidia the green light to resume exports. CEO Jensen Huang confirmed the news himself on July 14, saying U.S. officials had assured the company it could sell H20s again in China.
It’s a major reversal, and one that’s already being felt across both sides of the Pacific.
China’s AI Industry Is Eating Up Chips Like Candy
Demand in China has exploded since the launch of DeepSeek, a homegrown AI model unveiled in January. Built on limited resources, DeepSeek still managed to rival early iterations of ChatGPT—and that sent alarm bells ringing in Washington.
But instead of waiting for domestic manufacturing to catch up, Chinese firms are racing to grab every H20 they can get their hands on. Nvidia’s latest order from TSMC reflects that desperation—and the growing urgency among Chinese developers to scale fast.
-
Chinese AI labs have shifted into overdrive, especially in Beijing and Wuhan, where at least two major institutions have opened new branches aimed at developing alternative AI frameworks.
-
TSMC, the Taiwan-based foundry producing these chips, is operating near capacity as the orders pour in—not just from Nvidia, but from rivals racing to fill similar demand.
-
Analysts at TrendForce estimate that demand for high-performance AI chips in China could rise 45% this year alone, fueled largely by the greenlighting of Nvidia’s exports.
U.S. Officials Say the Chips Can Stay… For Now
The White House hasn’t given a clear explanation for the policy reversal, and that’s led to a storm of speculation.
Is it a temporary concession to business interests? Or a calculated risk to monitor China’s AI progress more closely?
Alexandra Mousavizadeh, head of the Global AI Index, sees two competing schools of thought. “Either you wall off the chips and hope your own industry surges ahead, or you keep everything flowing, betting that the collaboration accelerates innovation for everyone,” she told Newsweek.
The administration has declined to say how long the current licenses will remain active, but insiders suggest it’s a matter of months, not years.
That uncertainty has manufacturers scrambling to frontload orders.
China’s Messaging: We Don’t Want Dominance, Just a Fair Shot
While Washington debates its next move, Beijing is already positioning itself as the reasonable player.
A spokesperson for the Chinese Embassy, speaking to Newsweek, emphasized that China does not seek AI supremacy. Instead, it wants fairness, inclusion, and access—insisting that cooperation, not confrontation, is the way forward.
It’s a common refrain, but the tone has shifted slightly in recent months. Officials have begun referring to AI development as “human-centered” and “benevolent”—language likely aimed at soothing Western anxieties without ceding an inch of ambition.
Meanwhile, backroom talks between U.S. and Chinese tech regulators have reportedly resumed, though no formal agreements have emerged.
The Race Is On, and the U.S. Might Be Fueling It
The irony? U.S. chips are helping China build systems designed to eventually outperform U.S. models. That’s exactly what some security experts feared.
In May, Georgetown University’s Center for Security and Emerging Technology flagged China’s strategy: while Western developers focus on large, data-hungry statistical models, Chinese researchers are targeting smaller, more adaptive AI frameworks.
These alternative models, if successful, could leapfrog existing architectures. That’s the bet. And with Nvidia’s hardware back on the table, it’s now a bet with a lot more chips in play—literally.
Nvidia Walks a Fine Line—And It’s Working
For Nvidia, this is a massive business win. H20s are already the most in-demand AI chips in the market, and China’s appetite is only getting stronger.
Jensen Huang has tried to thread the needle. In his July remarks, he argued that U.S.-based models, tools, and technologies are still the best foundation for civil AI. His pitch? Let everyone build on the American tech stack.
That message resonates in some circles—and irritates others. Lawmakers on both sides of the aisle have questioned whether these sales ultimately undermine U.S. interests.
But for now, the money talks louder.
