English
In a dramatic reversal of the multi-year US export control regime, Washington authorized the sale of NVIDIA's H200 GPUs to China in late 2025, signaling a fundamental recalibration of semiconductor strategy. The decision, confirmed by NVIDIA CEO Jensen Huang at GTC in March 2026, ended a 10-month supply freeze and introduced a novel mechanism where a portion of China sales revenue is directed to the US government. This article analyzes the strategic calculus behind the pivot—balancing domestic industry pressure against national security concerns—and examines how China is leveraging this opening to accelerate its domestic AI chip ecosystem. The shift reshapes global supply chains, competitive dynamics for AI leadership, and the future of US-China technological competition through 2026 and beyond.
Background: From Export Controls to Strategic Recalibration
The US began tightening semiconductor export controls in October 2022 under the Biden administration, targeting advanced AI chips like NVIDIA's A100 and H100. These restrictions expanded through 2023 and 2024, eventually encompassing the H200—a Hopper-architecture GPU featuring 141 GB of HBM3e memory and 4.8 TB/s bandwidth, delivering nearly double the inference performance of the H100. By early 2025, a complete freeze on H200 sales to China was in effect, leaving only the downgraded H20 chip available for the Chinese market.
However, the US-China chip war entered a new phase when President Trump announced on December 8, 2025, that NVIDIA could resume selling H200 chips to China in exchange for 25% of revenue flowing to the US government. This policy reversal, formalized through Commerce Department licenses granted in late February 2026, represents the most significant recalibration of technology export strategy since restrictions began. The Mitsui Global Strategic Studies Institute (MGSSI) published a detailed analysis in March 2026 documenting this major policy shift, noting that the competition has moved beyond initial trade measures into a 'Chip-for-Compliance' era.
The Strategic Calculus Behind the Pivot
Domestic Industry Pressure and Economic Imperatives
NVIDIA's fiscal 2026 revenue reached a record $215.9 billion, up 65% year-over-year, with data center revenue hitting $62.3 billion in Q4 alone. The China market represents a critical growth vector: Chinese firms have ordered over 2 million H200 units for 2026 delivery, representing a potential $54 billion revenue opportunity. However, NVIDIA only held approximately 700,000 chips in inventory at the time of the announcement, creating significant supply constraints.
McKinsey's March 2026 report highlights that AI-related trade—particularly semiconductors and data-center equipment—now accounts for one-third of global trade growth, with roughly 37% annual growth in 2024–25, far outpacing the global trade average of 6.5%. This economic reality forced Washington to reconsider the cost of complete decoupling. The semiconductor export controls had created unintended consequences, including lost market share for US firms and accelerated Chinese domestic innovation.
The 25% Revenue-Sharing Mechanism
The novel revenue-sharing framework functions as a sovereign tax on corporate earnings from China sales. Under the terms, NVIDIA must remit 25% of H200 revenue to the US government, creating permanent margin compression but allowing the company to access a market that would otherwise be lost entirely. This follows a similar 15% arrangement struck earlier in 2025 with NVIDIA and AMD for less advanced chips. The mechanism effectively turns chip exports into a revenue-generating tool for US government coffers while maintaining some control over technology flow.
China's Response: Accelerating Domestic AI Chip Ecosystem
Beijing has responded to the US pivot with a dual strategy: leveraging access to H200 chips while aggressively pushing domestic alternatives. Chinese authorities have issued informal boycott directives urging domestic firms to prioritize homegrown silicon, particularly Huawei's Ascend 910C and Cambricon's Siyuan series. The results have been striking: in Q1 2026, China's domestic AI chip industry reached a historic inflection point, with multiple companies turning profitable for the first time.
Cambricon posted a net profit of ¥10.13 billion (up 185%) on ¥28.85 billion revenue (+159.6%), achieving positive operating cash flow. Hygon Information generated ¥40.34 billion in revenue (+68.1%), with ¥14.14 billion from AI accelerators alone. Moore Threads crossed into profitability for the first time since its 2020 founding, with 150%+ revenue growth. Huawei is on track to ship 600,000–650,000 AI chips in 2026, while Cambricon targets 500,000 units of its Siyuan series.
The Chinese AI chip landscape has matured rapidly, with domestic chips now offering x86/CUDA-compatible architectures and real customer adoption across cloud, finance, telecom, and automotive sectors. JPMorgan projects that combined domestic shipments could break the million-unit barrier by 2026, fundamentally challenging NVIDIA's monopoly in China's data centers.
Geopolitical and Supply Chain Implications
The policy reversal has created a complex web of winners and losers. For NVIDIA, the H200 licenses provide a $5–$8 billion annual revenue opportunity, but the 25% fee and supply constraints limit upside. The company's next-generation Blackwell and Rubin chips remain excluded from China exports, ensuring that Chinese enterprises must build larger, less efficient data centers using older technology.
For China, the H200 access provides a temporary bridge while domestic alternatives scale. However, the US-China technology competition has entered a new phase where both sides weaponize chip exports for geopolitical leverage. The Trump administration's January 2026 Section 232 action imposed 25% tariffs on advanced computing chips while exempting chips imported to support US domestic manufacturing capacity, creating a complex tariff regime that intersects with export licensing requirements.
Global supply chains are being reshaped as a result. TSMC has been approached by NVIDIA for H200 production ramp discussions, while Asian hubs—Taiwan, South Korea, and parts of Southeast Asia—continue to supply markets worldwide. McKinsey's analysis notes that semiconductor trade growth is concentrated in these regions, with the US increasingly reliant on Asian manufacturing even as it seeks to onshore critical capabilities.
Expert Perspectives
"The agentic AI inflection point has arrived," NVIDIA CEO Jensen Huang stated during the Q4 FY2026 earnings call, highlighting the company's unveiling of the Rubin platform which promises up to 10x reduction in inference token cost. "The China market is critical to this growth trajectory, but we must navigate the regulatory framework carefully."
Analysts at the Mitsui Global Strategic Studies Institute note that the competition has moved beyond initial trade measures. "The US-China AI semiconductor rivalry has entered a new stage where export controls are being recalibrated to balance commercial competitiveness against national security," the MGSSI report states. "China's push for minimum self-sufficiency—enough to withstand sanctions—may be achievable by 2027."
Critics warn that the H200 approval hands China access to hardware integral to modern military applications. The Commerce Department launched a review in December 2025, and key uncertainties remain about implementation of the 25% revenue structure and its impact on US competitiveness. The April 2026 Trump-Xi summit is expected to be a pivotal moment for further negotiations.
FAQ
What is the NVIDIA H200 GPU?
The NVIDIA H200 is a high-performance AI accelerator built on the Hopper architecture, featuring 141 GB of HBM3e memory with 4.8 TB/s bandwidth. It delivers nearly double the inference performance of the H100 and is designed for data center AI training and inference workloads.
Why did the US approve H200 exports to China?
The US approved limited H200 exports to China in late 2025/early 2026 to balance domestic industry pressure against national security concerns. The decision includes a 25% revenue-sharing mechanism that directs a portion of China sales to the US government, turning chip exports into a revenue-generating tool while maintaining some control over technology flow.
What are the conditions of the H200 export licenses?
The licenses cover a limited quantity of H200 chips with a 25% federal revenue fee on exports. The chips are approved for commercial customers only, with a mandatory US security review step before re-export to China. NVIDIA's next-generation Blackwell and Rubin chips remain excluded from China exports.
How is China responding to the H200 approval?
China is pursuing a dual strategy: leveraging access to H200 chips while aggressively pushing domestic alternatives. Beijing has issued informal boycott directives for domestic firms to prioritize homegrown silicon like Huawei's Ascend 910C and Cambricon's Siyuan series. Chinese domestic AI chip companies reached profitability for the first time in Q1 2026.
What is the impact on global semiconductor supply chains?
The policy reversal reshapes global supply chains by creating a bifurcated market where China receives older-generation chips while advanced technology remains restricted. AI-related semiconductor trade now accounts for one-third of global trade growth, with Asian manufacturing hubs playing a central role. The April 2026 Trump-Xi summit is expected to be a pivotal moment for further negotiations.
Conclusion and Future Outlook
The Great Chip Pivot of 2025–2026 marks a defining moment in US-China technological competition. Washington's decision to approve H200 exports represents a pragmatic recalibration that acknowledges the economic realities of AI-driven trade growth while attempting to maintain strategic advantage through revenue-sharing and technology tiering. For China, the opening provides breathing room but also accelerates the domestic push for self-sufficiency.
Looking ahead to 2027 and beyond, several key trends will shape the landscape: the ramp of NVIDIA's Rubin platform, China's progress toward million-unit domestic chip production, and the evolution of the 'Chip-for-Compliance' framework. The April 2026 Trump-Xi summit will be a critical test of whether this calibrated approach can stabilize US-China technology relations or merely delay a more comprehensive decoupling. What is clear is that semiconductors have become the central battleground of 21st-century geopolitics, and the decisions made in 2026 will reverberate for years to come.
Sources
- Mitsui & Co. Global Strategic Studies Institute, 'The US-China AI Semiconductor Rivalry Enters a New Stage,' March 2026
- McKinsey Global Institute, 'Geopolitics and the Geometry of Global Trade: 2026 Update,' March 2026
- NVIDIA Corporation, Q4 FY2026 Earnings Release, February 2026
- US Commerce Department, Export Licensing Determinations, February 2026
- White House Fact Sheet, 'President Donald J. Trump Takes Action on Certain Advanced Computing Chips,' January 14, 2026
- Congressional Research Service, 'U.S. Export Controls Targeting China: Advanced Semiconductors,' Report R48642
Follow Discussion