The H200 sits in a peculiar limbo—approved but not yet flowing
In a rare gesture of commercial openness, Washington has cleared ten of China's most powerful technology companies to purchase NVIDIA's advanced H200 processors — a chip that had long been withheld over fears of military application. Yet the authorization has not translated into transaction: Beijing, wary of hidden conditions and strategic dependency, has quietly counseled its firms to hesitate. The episode reveals how deeply the question of semiconductor access has become entangled with sovereignty, trust, and the slow-moving logic of geopolitical rivalry.
- Washington's December 2025 relaxation of chip export rules opened a door that had been firmly shut — allowing up to 75,000 H200 units per company for ten Chinese tech giants including Alibaba, Tencent, and ByteDance.
- Despite the authorization, not a single chip has changed hands, as Chinese government guidance has caused several approved buyers to pull back from their purchase intentions.
- Beijing's reluctance centers on suspected hidden vulnerabilities in the chips and an unacceptable condition requiring processors to transit American soil — a dependency China views as a strategic liability.
- NVIDIA CEO Jensen Huang traveled to Beijing alongside President Trump for talks with Xi Jinping, signaling how urgently the chip impasse now registers at the highest levels of both corporate and state power.
- China's hesitation is inseparable from its long-term push to build domestic semiconductor capacity and reduce reliance on Western technology — making even an approved purchase feel like a concession.
- The H200 remains in a peculiar limbo: legally purchasable, strategically desired, yet politically frozen — a precise emblem of the partial, mistrustful opening that defines US-China tech relations today.
The US Department of Commerce has authorized ten major Chinese technology companies — including Alibaba, Tencent, ByteDance, JD.com, Lenovo, and Foxconn — to purchase NVIDIA's H200 processors, among the most capable AI chips currently available. Each firm may acquire up to 75,000 units, either directly from NVIDIA or through intermediaries. The move follows a December 2025 relaxation of export restrictions that had previously kept the H200 off-limits due to concerns over its potential military applications.
The H200 is a significant generational leap — second only to NVIDIA's newer B200 in performance, and far more powerful than the H20 model already permitted in the Chinese market. For companies racing to build competitive AI infrastructure, the gap between these generations is consequential.
Yet the authorization has not produced sales. Several of the approved companies have reportedly stepped back after receiving guidance from Chinese authorities, who harbor reservations about conditions attached to the purchases — particularly a requirement that chips pass through American territory, which Beijing views as an unacceptable point of dependency. Suspicions about potential hidden vulnerabilities in the hardware have added to the reluctance.
This caution fits a larger pattern. China has been systematically investing in domestic chip manufacturing for years, driven by both economic strategy and national security imperatives. State support for local semiconductor firms has grown, and Beijing actively steers its companies toward homegrown alternatives.
NVIDIA's Jensen Huang recently traveled to Beijing alongside President Trump for talks with President Xi Jinping — a measure of how central the chip question has become to both corporate and diplomatic agendas. Still, no resolution has emerged. The H200 sits approved but undelivered, a symbol of a relationship defined by partial openings and persistent mutual suspicion.
The United States Department of Commerce has cleared the way for ten major Chinese technology companies to purchase NVIDIA's H200 processors, among the most sophisticated artificial intelligence chips the company manufactures. The authorization marks a significant shift in American export policy, yet the move has done little to ease the underlying tensions between Washington and Beijing over semiconductor access and technological sovereignty.
The approved buyers include household names in Chinese tech: Alibaba, Tencent, ByteDance, JD.com, Lenovo, and Foxconn, among others. Each company is permitted to acquire up to 75,000 H200 units, whether directly from NVIDIA or through intermediaries. As of now, no chips have actually been delivered. The authorization came after the United States relaxed its restrictions on advanced chip exports to China in December 2025, a partial reversal of the blanket prohibitions that had kept the H200 off-limits due to concerns that the technology could strengthen Chinese military capabilities.
The H200 represents a significant leap forward in AI processing power. It ranks second only to NVIDIA's newer B200 model in performance, and it vastly outpaces the H20, which had already been cleared for the Chinese market the previous year. For companies hungry to build competitive AI systems, the difference between these generations matters enormously.
Yet despite the American authorization, something has stalled the momentum. Several of the ten approved companies have reportedly pulled back from their purchase intentions after receiving guidance from Chinese government authorities. The hesitation appears rooted in Beijing's own reservations about the technology and the conditions attached to it. The Chinese government has not publicly detailed what specific concerns or changes to the sales terms prompted this caution, but the suspicion lingers that the chips may contain hidden vulnerabilities or that the requirement for the processors to pass through American territory—ostensibly so the US government can collect its revenue share—creates an unacceptable dependency.
China's wariness sits within a broader strategic pivot. For years, Beijing has been systematically investing in domestic semiconductor manufacturing and design, explicitly aiming to reduce reliance on Western technology. State support for local chip makers has intensified, and the government actively encourages Chinese companies to use domestically produced alternatives whenever possible. This push is not merely economic; it is a matter of national security and technological independence.
Jensen Huang, NVIDIA's chief executive, has been vocal about the damage these restrictions inflict on his company's position in the Chinese market. He recently traveled to Beijing alongside President Donald Trump to participate in talks with Chinese President Xi Jinping, a sign of how central this issue has become to both corporate and diplomatic interests. Yet even that high-level engagement has not resolved the fundamental question: will China actually permit its companies to move forward with these purchases?
For now, the answer remains unclear. The American authorization is in place, the companies are technically permitted to buy, but Beijing has not given the final green light that would allow its firms to proceed. The H200 sits in a peculiar limbo—approved for sale but not yet flowing into Chinese hands. It is a snapshot of the current state of US-China technology relations: a partial opening that neither side fully trusts, and a reminder that export licenses and government approvals are only the beginning of a much longer negotiation.
Notable Quotes
NVIDIA CEO Jensen Huang has warned that US restrictions are seriously damaging the company's position in the Chinese market— Jensen Huang, NVIDIA CEO
The Hearth Conversation Another angle on the story
Why would China hesitate if the US has already approved the sale? Isn't that what they wanted?
The approval is only half the story. China's government is worried about what comes with these chips—whether there are backdoors built in, whether the requirement to route them through American territory gives Washington leverage, or whether the conditions have changed in ways that weren't disclosed.
So they don't trust NVIDIA or the US government?
It's not quite that simple. They don't trust the arrangement itself. If they buy these chips and something goes wrong—if there's a vulnerability, or if the US uses the supply chain as a pressure point later—they've made themselves dependent on American goodwill. That's the real risk.
But China is pushing its own chip makers. Couldn't they just use those instead?
They're trying to, but they're not there yet. Domestic chips can't match what NVIDIA makes. So China is caught between needing the technology now and wanting to be independent later. The H200 is powerful, but accepting it means accepting American terms.
What does NVIDIA get out of this?
Access to the world's largest AI market. Without China, NVIDIA's growth is capped. But right now, even with approval, the chips aren't moving. That's the frustration Huang is trying to solve with these high-level visits.