US Secures 15% Cut on AI Chip Sales to China

U.S. ties AI chip exports to revenue sharing in China sales

The AI chip competition just took a new turn. According to the Financial Times, Nvidia and AMD have agreed to hand the U.S. government 15% of revenue from sales of high-end AI chips in China in return for export licenses. Nvidia’s H20 and AMD’s MI308 are named specifically in the report.

This arrangement follows a string of policy moves: the Trump administration first restricted certain inference chips in April, then paused parts of that ban after Nvidia pledged up to $500 billion in U.S. data-center investment. Nvidia later redesigned the H20 for China and resumed plans to sell there when licenses began issuing.

Nvidia told reporters it follows U.S. rules and hopes export controls will let American companies compete globally. Commerce Secretary Howard Lutnick linked Nvidia’s shift to broader trade talks with China over rare-earth elements, essential for components such as EV batteries.

Not everyone welcomes the decision. National security experts and former officials have urged the government to reconsider, arguing that easing access to advanced inference chips could weaken strategic leverage. The new revenue-sharing model blurs the lines between tariffs, export controls, and geopolitical bargaining.

Why this matters

The deal creates several practical and strategic implications for companies, governments, and customers in China and elsewhere:

• Pricing and margins: A 15% revenue share directly reduces seller margins or forces higher prices for Chinese buyers.
• Policy precedent: Revenue-sharing as a condition for export licenses sets a new lever for foreign policy beyond conventional tariffs or bans.
• Compliance complexity: Companies must track China-specific revenue, license scopes, and reporting obligations across product lines.
• Supply-chain negotiation: Tying exports to rare-earth talks signals that hardware access and component sourcing are now tangled in the same diplomatic threads.

What organizations should do next

Whether you are a chipmaker, a cloud provider, or a government policymaker, this development calls for rapid, data-driven planning. Practical steps include:

• Run revenue-impact models to understand margin shifts and customer price sensitivity.
• Map supply chains and rare-earth dependencies to identify leverage points and single-source risks.
• Automate compliance and revenue tracking for country-specific license conditions and audits.
• Stress-test business cases under alternative policy scenarios, including escalations or broader export controls.

These steps help firms protect margins, keep customers informed, and reduce the risk of sudden market disruption. For governments, transparent rules and predictable licensing will matter more than ever to avoid ad hoc outcomes.

The biggest takeaway: this revenue-sharing approach shifts the AI chip narrative from pure national security controls to a hybrid of trade, investment, and diplomatic bargaining. Expect more creative policy tools and a new set of compliance headaches for global tech companies.

QuarkyByte analyzes these tangled trade-policy moves with quantitative scenario work and supply-chain mapping so leaders can act fast and with clarity.