Nvidia CEO Says China Demand for H200 Chips Is 'Very High' as Export Licenses Near Completion

Nvidia's $50 billion China opportunity is inching closer to reality. CEO Jensen Huang revealed Tuesday at CES 2026 that demand from Chinese customers for the company's H200 artificial intelligence chips is "very high," and that the chipmaker has already restarted its supply chain in anticipation of imminent regulatory clearance.

Speaking during a Q&A session with journalists following his blockbuster CES keynote, Huang confirmed that Nvidia is working out the final details of export licenses with U.S. government officials—the last hurdle before what could become one of the most consequential chip sales in semiconductor history.

The $50 Billion Market Reopens

The H200 represents Nvidia's first high-performance AI chip approved for export to China since the Trump administration tightened restrictions in 2022. Previous chips sold to China were deliberately hobbled to comply with export controls, limiting their usefulness for the most demanding AI applications.

The H200 is different. Based on Nvidia's powerful Hopper architecture, it pairs the company's flagship H100 GPU with 141 gigabytes of cutting-edge HBM3e memory and dramatically higher memory bandwidth. The result is a chip purpose-built for training and running the large language models that power generative AI applications—exactly what China's technology giants need to compete globally.

"We're not expecting any press releases, or any large declarations. It's just going to be purchase orders."

— Jensen Huang, Nvidia CEO, on how the company will know China's import approval is confirmed

Huang noted that while the U.S. has signaled approval for H200 exports, China must also approve imports of the chips—a formality that could face political complications given the broader technology competition between the two nations. Nvidia expects to learn of China's decision through customer purchase orders rather than official government announcements.

The Deal Terms

President Trump announced the export approval framework in December, with several notable conditions designed to ensure U.S. national security interests are protected while allowing American companies to capture China's surging AI spending:

• 25% revenue surcharge: A quarter of all H200 sales to China will go directly to the U.S. government as a tax
• Approved customers only: Sales limited to U.S. companies operating in China, preventing direct sales to Chinese firms
• Use restrictions: Chips cannot be resold or transferred to entities on government blacklists
• Monitoring requirements: Nvidia must report on chip deployment and usage to U.S. authorities

Even with the 25% tax, China's major cloud providers—including Alibaba, Tencent, and ByteDance—have reportedly initiated purchase discussions. The chips' performance advantages make them essential for companies seeking to train competitive AI models, regardless of the premium pricing.

What's at Stake for Nvidia

Before export restrictions took effect, China represented roughly 20% of Nvidia's data center revenue—translating to tens of billions of dollars annually at today's run rates. The company has been effectively shut out of this market for over two years, with competitors like Huawei attempting to fill the gap with domestically produced alternatives.

Huang has previously estimated the addressable Chinese market at $50 billion annually—revenue that is currently not included in Nvidia's financial forecasts. Even capturing a portion of this demand would meaningfully accelerate the company's already torrid growth trajectory.

The timing is particularly significant because China's AI spending is accelerating. Chinese technology companies, facing both competitive pressure and government mandates to achieve AI self-sufficiency, are racing to deploy infrastructure capable of training large foundation models. Without access to high-performance Nvidia chips, their efforts have been constrained.

Competitive Implications

H200 exports to China could reshape the competitive landscape in multiple ways:

• Huawei pressure: Chinese chip champion Huawei has developed AI accelerators to substitute for restricted Nvidia products, but technical limitations have hampered adoption. H200 availability could slow Huawei's momentum in the AI chip market.
• Cloud competition: U.S. hyperscalers operating in China—including AWS, Microsoft Azure, and Google Cloud—could gain advantages over local competitors by offering superior AI computing infrastructure.
• AI model parity: Access to cutting-edge training hardware could help Chinese AI labs close the gap with Western counterparts on model capabilities.

The geopolitical implications are complex. Allowing H200 sales generates revenue for U.S. companies and tax receipts for the government while maintaining America's technological lead. But it also equips potential adversaries with powerful tools for AI development, raising questions that will likely be debated throughout the year.

Investor Response

Nvidia shares have rallied approximately 15% since the export approval was first signaled in December, reflecting investor enthusiasm about the China revenue opportunity. The stock touched fresh all-time highs during CES week as Huang unveiled a parade of new products and partnerships.

However, some analysts caution that execution risk remains. The 25% tax will pressure margins on China sales, and political developments—in either Washington or Beijing—could complicate the business. The complexity of compliance requirements may also slow the pace of chip deployments.

For Nvidia's financial outlook, the key question is how quickly Chinese customers can absorb supply. Huang indicated the company has already restarted its China-focused supply chain, suggesting shipments could begin within weeks once final approvals are secured. Revenue recognition would follow shortly thereafter.

The Broader Trade Picture

The H200 export decision fits into a larger pattern of the Trump administration using trade policy as a negotiating lever rather than a blunt prohibition. By allowing sales with conditions, the U.S. captures economic value from the transaction while maintaining some control over technology deployment.

This approach represents a departure from the more restrictive Biden-era policies that focused on denying China access to advanced semiconductors entirely. The shift reflects a calculation that absolute export bans are difficult to enforce, encourage foreign competitors to fill the vacuum, and deprive American companies of significant revenue opportunities.

Whether this balanced approach proves durable will depend on geopolitical developments. Any escalation in U.S.-China tensions—whether over Taiwan, trade imbalances, or technology competition—could prompt a reassessment of export permissions. Nvidia's China business remains subject to political risk that is difficult to hedge.

For now, Huang's message at CES was unmistakably optimistic. After years of watching China's AI market develop without Nvidia's participation, the company appears poised to reclaim its position as the dominant supplier of AI computing infrastructure worldwide—including in the world's second-largest economy.