Trump-Xi Beijing Summit May 2026: H200 Deals, AI Governance, $150B Stakes

President Trump arrived in Beijing on May 13, 2026 for summit meetings with President Xi Jinping on May 14-15 — the first US presidential state visit to China since Trump's own November 2017 trip. The summit follows the May 12 trade truce that cut US tariffs from 145% to 30% and Chinese tariffs from 125% to 10%. On the technology agenda: whether H200 GPU exports to China get a formal licensing framework, what China concedes on rare earth export restrictions, and whether a bilateral AI governance framework gets signed.

At stake beyond the headline trade numbers: Nvidia's $78 billion annual guidance assumes zero China H200 recovery. A licensing framework changes that number. And the bilateral AI governance discussion — however thin on substance — is the beginning of the world's two largest AI-developing nations attempting to talk to each other about what they're building.

The Semiconductor Negotiation: H200 Licenses

The core technology item at the Beijing summit is the H200 export control posture.

Background: Nvidia designed the H20 to comply with 2022 export control thresholds. The Biden administration banned the H20 anyway in April 2025, costing Nvidia $7 billion in Q1 FY2026. Trump subsequently signalled an H200 case-by-case licensing approach in December 2025 — H200 exports with a 25% surcharge and inter-agency approval. But that framework has operated with "presumption of denial" as the default for most Chinese customers.

What China wants from the summit: A formal H200 licensing framework for Chinese commercial AI companies — universities, research labs, and model developers — rather than the current case-by-case presumption of denial. China has separately informed its tech companies that it will approve H200 purchases only under "exceptional circumstances" pending summit outcomes.

What the US wants in exchange: China to resume rare earth element exports under the normal trade framework — gallium, germanium, antimony, and graphite, which China restricted in retaliation for the H20 ban. These materials are critical for compound semiconductors (GaAs, GaN) used in RF components, power electronics, and photonics.

The strategic paradox: China has demonstrated with the May model releases (DeepSeek V4, Kimi K2.6, MiniMax M2.7, GLM-5.1) that it can build frontier-competitive models without H200 access. Beijing's negotiating leverage is that the H200 restriction is hurting Nvidia's revenue more than it is hurting Chinese AI capability. The US argument for maintaining restrictions — preventing capability development — is weaker than it was 18 months ago.

Most likely outcome: A formal H200 licensing framework for Chinese commercial customers that replaces "presumption of denial" with "presumption of approval for commercial AI research under specified end-use constraints" — not unlimited access, but a pathway. This returns Nvidia approximately $3.5-4 billion in annual China revenue without formally lifting the ban.

Rare Earths: China's Leverage

China produces approximately 60% of the world's rare earth elements and processes approximately 85% of global rare earth supply. The export restrictions on gallium, germanium, antimony, and graphite imposed in 2023-2025 were calibrated to apply pressure on US semiconductor manufacturing without triggering World Trade Organisation disputes.

The restrictions affect:

• Gallium: Used in GaAs semiconductors (RF chips in every smartphone and radar system)
• Germanium: Used in fiber optic components, infrared optics, and solar cells
• Antimony: Used in flame retardants (semiconductor packaging) and lead-acid batteries
• Graphite: Used in lithium-ion battery anodes (critical for EV battery supply chains)

US defence contractors and semiconductor manufacturers have been stockpiling and developing alternative sources since the restrictions began, but substitution is incomplete. The summit is an opportunity to restore the supply baseline while China extracts H200 concessions.

Expected outcome: Partial relaxation of rare earth export licensing requirements for US companies, in exchange for the H200 framework. Full removal of rare earth restrictions is unlikely — China would prefer to maintain them as ongoing negotiating leverage.

The AI Governance Framework

The AI governance discussion at the summit is light on substance but significant for its existence. The framework being discussed:

Mutual capability notification: The two countries agree to notify each other when AI systems cross defined capability thresholds — benchmarks for reasoning, code generation, or agentic operation that trigger notification requirements. The purpose is to prevent either side from being surprised by the other's capability development.

AI incident information sharing: A bilateral mechanism for sharing information about AI-related incidents — model misuse, security failures, or unintended behaviour at deployment scale. The model is analogous to hotlines established for nuclear incidents during the Cold War.

Red lines on AI-enabled weapons: Discussion (without expectation of agreement) on whether certain uses of AI in autonomous weapons systems should be mutually constrained. China has been resistant to any constraints on autonomous weapons development; the US is pushing for language that acknowledges the topic exists.

The cynical reading: both sides want to be seen as taking AI governance seriously for domestic audiences. The practical content of any signed framework will be minimal. The optimistic reading: even minimal bilateral AI governance dialogue creates precedent and a communication channel that is absent today.

What Changes After the Summit

Assuming the most likely outcome — H200 case-by-case framework, partial rare earth relaxation, thin AI governance statement:

For Nvidia: The H200 licensing framework restores an estimated $3.5-4 billion in annual China revenue. Against the $78 billion Q1 FY2027 guidance baseline (which assumed zero recovery), this is a 4-5% upside. Nvidia shares have been pricing in zero China recovery — any formal framework is a positive catalyst.

For Chinese AI labs: Access to H200s for qualifying commercial customers removes the training efficiency disadvantage Chinese labs currently face versus Western frontier labs. The gap between Arena Elo scores (Western frontier 1,481-1,503, best Chinese model 1,449) is partly attributable to training compute disadvantage. H200 access for research and commercial AI development would narrow that gap further.

For the 90-day tariff truce: The May 12 tariff reduction was a 90-day ceasefire. A successful summit with substantive semiconductor agreements makes a permanent extension more likely. A failed summit risks reversion to 145% tariffs in mid-August.

For supply chains: Partial rare earth relaxation reduces input cost risk for US semiconductor manufacturers — a second-order effect that takes 6-12 months to flow through to component pricing.

For AI governance globally: A bilateral US-China AI governance framework, however thin, creates pressure on the EU's more detailed AI Act regulatory framework. If the world's two largest AI-developing nations have agreed to a notification framework, EU regulators will want equivalent or stronger commitments from companies operating in Europe.

Why This Summit Matters More Than Past Trade Talks

The May 2026 Beijing summit is happening in a context that makes technology its defining agenda, not its secondary agenda.

Past US-China summits have treated technology as a subset of the broader trade relationship. At this summit:

• The primary trade concession from the US side is a technology export control relaxation
• The primary trade concession from China is a critical materials supply chain normalization
• The AI governance framework is explicitly on the formal agenda, not a side discussion

This is the first US-China summit where AI and semiconductor policy are the core deal structure rather than supporting elements. The outcome will be studied in every technology capital — Seoul, Tokyo, Taipei, Berlin, London — as a signal of whether the US-China technology decoupling trajectory is being managed or merely postponed.

Key Takeaways

• Trump in Beijing May 13-15: First US presidential state visit to China since November 2017; follows May 12 trade truce (tariffs 145% → 30%)
• H200 framework most likely outcome: "Presumption of approval for commercial AI research" replacing "presumption of denial"; restores ~$3.5-4B in Nvidia annual China revenue; formal framework rather than current ad-hoc case-by-case process
• Rare earth exchange: China partially relaxes gallium, germanium, antimony, graphite export restrictions in exchange for H200 framework; removes full restrictions unlikely, partial normalization likely
• AI governance framework: Mutual capability notification, incident information sharing — thin on substance, significant for existence; first bilateral US-China AI governance dialogue
• 90-day truce extension: Successful summit with semiconductor agreements makes permanent tariff reduction more likely; failure risks reversion to 145% tariffs in mid-August
• Technology is the core deal: First US-China summit where AI and semiconductor policy are the central trade structure, not a secondary agenda item

For the Chinese AI model context that has changed Beijing's negotiating position, read 4 Chinese Open-Weights Models in 12 Days: The AI Inference Cost War. For the Nvidia revenue impact of the H200 China ban, read Nvidia's $78B Guidance Hides an $8B China Hole.