July 15, 2025 — Nvidia, the world’s leading AI chip designer, is on the cusp of resuming sales of its advanced H20 AI processor to Chinese customers, following new indications from the U.S. government that export licenses for certain chips will soon be granted. The news arrives after months of heightened restrictions, casting new light on both the competitive dynamics in the world’s second-largest economy and the larger strategic technology rivalry between the United States and China.

U.S. Export Controls: A Shifting Landscape

The origins of the current impasse date back to April 2025, following actions by the Trump administration to further restrict U.S.-designed advanced AI hardware from being exported to China. Under these new rules, Nvidia’s flagship H20 GPU—designed specifically for large-scale artificial intelligence workloads—was placed on a list requiring export licenses for all shipments to China or companies ultimately controlled from the country. This followed previous controls on Nvidia’s H100 and A100 chips, which were considered critical to powering AI training and inference at scale.

Nvidia disclosed in April that the licensing requirement had left billions of dollars’ worth of H20 inventory in limbo, initially estimating a $5.5 billion charge related to unsold units, though the final write-down was around $4.5 billion as of their latest quarterly report. The situation underscored the precariousness of doing business in an environment where policy decisions can instantly change market access for high-tech firms.

Potential Resumption — With Limits Unclear

During a visit to China, Nvidia CEO Jensen Huang revealed that the company has initiated applications to resume H20 AI chip sales. Aspirations for prompt licensing have been buoyed by signals from the U.S. government that approvals are forthcoming, though permanent restrictions remain and the licensing requirement has no set expiration.

What remains murky are the details: Will licenses be limited by strict quotas? How deeply will U.S. officials scrutinize customers and end uses? There is concern that chips may be re-exported or diverted to Chinese military or surveillance applications, something that U.S. lawmakers have sought to avoid. While Huang asserts that the risk of military use is minimal—arguing that Chinese entities would not depend on a supply that could be disrupted at any moment—Washington’s posture remains cautious, and compliance burdens for Nvidia may increase over time.

Nvidia’s Strategic Shift: Launching the RTX PRO

In parallel with regulatory hurdles, Nvidia quickly moved to introduce a new product designed specifically for the Chinese market — the NVIDIA RTX PRO GPU. Announced in mid-2025, the RTX PRO was engineered to comply fully with U.S. technical limitations, utilizing standard GDDR7 memory instead of the high-bandwidth memory (HBM) found in top-tier AI chips and eschewing advanced CoWoS packaging technologies.

The result is a more affordable chip (approximately half the price of the H20), but also one that is considerably less powerful. Market analysts characterize the RTX PRO as a significant step down from the H20 and far behind the H100, which remains the gold standard for advanced AI workloads globally. This strategic adaptation highlights Nvidia’s determination to maintain its presence in China, even as its most cutting-edge technologies are largely out of reach for Chinese firms.

Competitive Erosion: Huawei and Local Champions Surge

The drawn-out licensing process and limitations on high-end chips have already dented Nvidia’s formidable market share in China. CEO Jensen Huang recently acknowledged that Nvidia’s share in the region has fallen from 95% to around 50%, with domestic titans like Huawei aggressively filling the void. Huawei’s Ascend AI processors, developed independently of American technology, have seen surging adoption by Chinese tech firms and cloud providers, supported in part by extensive government policies aimed at achieving semiconductor self-sufficiency.

According to a June 2025 report by market intelligence firm TrendForce, shipments of Huawei Ascend chips in China grew by over 80% year-over-year in the first half of 2025, and the company now commands more than a third of China’s advanced AI chip market. Other regional upstarts, including Alibaba’s T-Head and Baidu’s Kunlun, are also investing heavily to catch up to international standards, although the performance gap remains wide for now.

Financial Stakes and Geopolitical Risks

China contributed $17 billion in revenue to Nvidia’s bottom line in its most recent fiscal year ending January 26, 2025, representing roughly 13% of global sales. Although Nvidia’s soaring stock price has given it a market cap exceeding $4 trillion—briefly making it the world’s most valuable listed company in June 2025—the situation underscores how geopolitical risk remains a major threat to future growth.

The Biden administration’s ongoing support for Trump-era export controls, despite fierce lobbying from U.S. chipmakers, reflects a growing consensus in Washington that technology access should be limited as a strategic lever. Meanwhile, Beijing has ramped up spending on its semiconductor sector, pledging more than $50 billion in new investments through 2026 and expanding policies to favor purchases of locally produced chips in critical applications.

Broader Industry Implications

As the AI and semiconductor industries become increasingly enmeshed in U.S.-China rivalry, other chipmakers are similarly navigating uncertain terrain. AMD, for instance, has developed China-specific variants of its AI hardware, and Intel has emphasized compliance with evolving U.S. regulations.

Analysts expect that, even with resumed H20 deliveries, Chinese firms will accelerate efforts to wean themselves off foreign suppliers and build a robust domestic AI chip ecosystem. Combined with aggressive competition from local champions, Nvidia’s ability to maintain revenue streams from China may be permanently diminished.

Outlook: Stuck Between Opportunity and Oversight

For Nvidia, the most likely scenario over the coming year is one of partial engagement: selective product sales enabled by licenses, continued development of downgraded “compliant” chips, and unrelenting pressure from both competitors and policymakers. With the next U.S. presidential election on the horizon and global supply chains still recovering from years of trade disruptions, artificial intelligence hardware will remain a focal point of broader economic and strategic dialogue.

Ultimately, while the recent green light to resume H20 shipments signals a modest easing of tensions, the larger risks — export quotas, intensified competition, and evolving regulatory burdens — will persist. Nvidia, and the semiconductor industry at large, must continue to innovate not just in technology but in navigating the complex worlds of geopolitics and cross-border regulation.