• NVIDIA must obtain a U.S. government license and pay a 15% revenue share for sales of certain new AI chips, including the H20, to China.
  • The unprecedented agreement, driven by recent changes in U.S. export policy, allows NVIDIA to resume sales to its major market but under strict conditions.
  • The move stabilizes NVIDIA's near-term China revenue but raises significant questions about the long-term integrity of U.S. export controls and national security.

NVIDIA's efforts to navigate stringent U.S. export controls have resulted in a novel and contentious arrangement: the chipmaker can now sell its advanced artificial intelligence processors to Chinese clients, but only if it first secures a U.S. government license and agrees to hand over 15% of the revenue from those sales. The policy, which applies to new chips like the H20 designed for the Chinese market, marks a dramatic shift from an outright ban to a pay-for-access model.

The agreement, driven by a recent reversal in U.S. export control policy, effectively ends a period of uncertainty that had halted shipments of NVIDIA's most powerful chips to a region that represents a critical source of demand. The company's specialized AI chips, for which China's domestic industry still lacks a direct performance equivalent, are vital for Chinese tech giants like Tencent to develop advanced AI systems. According to people familiar with the matter, the licensing process is now underway, though actual shipments remain constrained until official approvals are granted.

This transactional approach to export regulation is highly unusual and has immediately sparked debate in policy circles. While it provides a temporary revenue stream for the U.S. government and stabilizes NVIDIA's financial outlook in the near term, national security experts have expressed deep skepticism. The core concern is that the arrangement prioritizes government revenue and commercial interests over the original national security objective of the controls, which was to prevent China's military and AI sectors from accessing cutting-edge Western technology.

“What institutional investors are really focused on is regulatory stability,” a source close to the discussions said, speaking on the condition of anonymity. “This deal provides a path forward for sales, but the conditions remain entirely subject to evolving government licensing and geopolitical objectives.”

The policy does not exist in a vacuum. Rival chipmaker AMD is subject to a parallel arrangement for its MI308 chips destined for China, indicating a broader, sector-wide shift in how the U.S. is managing its technological competition with Beijing. The deal has drawn comparisons to negotiation tactics seen in other sectors, such as steel, suggesting a move toward more transactional trade and export regulation.

For NVIDIA, the immediate calculus is clear: regain access to a multi-billion dollar market. The 15% revenue share, while a significant cost, is likely preferable to a total loss of those sales. However, the long-term implications are far murkier. Analysts warn that such a precedent could degrade the credibility of U.S. technology safeguards, embolden lobbying for further exemptions, and ultimately increase strategic risks by continuing to fuel China's technological advancement. The chip war is not over; its rules have simply become more complex and commercially driven.