President Donald Trump said he discussed artificial intelligence guardrails and Nvidia’s advanced chips with Chinese President Xi Jinping, putting two of Silicon Valley’s most consequential policy fights — AI safety and semiconductor access to China — into the center of U.S.-China diplomacy after a two-day Beijing summit.

Trump told reporters aboard Air Force One on Friday that he and Xi had talked about the possibility of working together on guardrails for artificial intelligence. Asked what kind of guardrails he had in mind, Trump described them only as the standard safeguards discussed around AI, according to reports from the trip. He also said Nvidia Corp.’s H200 chips came up during the conversations, a sign that the future of U.S. AI hardware exports remains politically sensitive even after Washington moved to allow limited sales to some Chinese buyers.

The remarks did not amount to a new bilateral technology agreement. They instead pointed to a more tentative stage in which Washington and Beijing are testing whether AI safety can be separated, at least partially, from the broader rivalry over chips, data centers, military applications and industrial policy. U.S. officials have increasingly framed frontier AI as both an economic opportunity and a national-security risk, while Chinese officials have sought to reduce dependence on American semiconductor technology and accelerate domestic alternatives.

For Silicon Valley, the distinction matters. Nvidia remains the dominant supplier of AI accelerators used to train and run large models, and its China business has been constrained for years by U.S. export controls designed to prevent advanced chips from supporting Chinese military modernization or other sensitive capabilities. The H200 sits near the center of that debate because it is powerful enough to matter for commercial AI workloads but has also been treated as a strategic asset by policymakers concerned about compute access.

U.S. Trade Representative Jamieson Greer said separately that semiconductor export controls were not a major topic in the China talks, according to Reuters. That statement tempered expectations that the summit would produce a clear breakthrough for chip companies or investors hoping for a more predictable path for U.S. AI hardware sales into China. Reports around the summit indicated that while the United States had approved some H200 exports to selected Chinese firms, actual deliveries remained uncertain amid Chinese reluctance and broader policy friction.

Trump’s comments also reflected the administration’s argument that the United States remains ahead in the AI race. He said Xi was aware of U.S. progress and portrayed China as a strong competitor rather than an equal leader. That message tracks with the White House’s broader technology agenda, which has emphasized domestic AI infrastructure, semiconductor capacity, and the use of export policy to preserve U.S. advantages in advanced computing.

The guardrails discussion is likely to draw close attention from U.S. AI developers because any U.S.-China channel on AI safety could affect how governments approach model evaluations, misuse prevention, incident reporting, military applications and cross-border standards. Even a limited dialogue could matter if it reduces the risk of misunderstanding around powerful AI systems. But the same talks could become politically vulnerable in Washington if lawmakers view cooperation as weakening pressure on Beijing or creating a route for China to gain insight into U.S. capabilities.

That tension has become more pronounced as AI systems become more capable and more expensive to build. Advanced models require massive computing clusters, specialized chips, high-bandwidth memory, energy infrastructure and sophisticated software ecosystems. Because Nvidia’s accelerators are central to that stack, export decisions are no longer narrow trade questions. They influence the pace at which Chinese cloud providers, model developers and research institutions can scale frontier systems relative to U.S. companies.

The China market remains commercially important for U.S. chipmakers, but the policy environment has become increasingly unpredictable. Nvidia has repeatedly adapted its product lineup in response to U.S. restrictions, while Chinese companies have invested in domestic chips and alternative architectures. If Chinese firms accelerate away from Nvidia hardware, the long-term effect could be a partial decoupling of AI infrastructure markets, with U.S. companies losing revenue and influence even as Washington seeks to limit strategic technology transfer.

At the same time, U.S. officials face pressure not to allow commercial considerations to override security concerns. Critics of advanced chip sales to China argue that even civilian AI hardware can support military planning, cyber operations, surveillance systems, weapons development or other dual-use applications. Supporters of calibrated exports argue that maintaining some commercial relationship can preserve U.S. leverage, support American companies, and reduce incentives for China to build entirely independent supply chains faster.

The summit therefore left technology markets with mixed signals. Trump’s remarks suggested that AI safety has reached the level of direct leader discussion, which could eventually create room for a formal risk-management channel. But Greer’s comments suggested that export controls were not settled in the meeting, and the absence of a major announcement left Nvidia and other semiconductor companies facing continued uncertainty over access to Chinese customers.

Chip stocks had rallied in recent months on expectations that AI infrastructure spending would remain strong across cloud computing, enterprise software and sovereign AI projects. Any shift in U.S.-China chip policy can affect that trade because China is both a major end market and a strategic competitor. Investors are watching not only whether Nvidia can ship H200 chips, but also whether broader U.S. controls might tighten again if national-security agencies conclude that China is closing the AI performance gap.

The policy stakes extend beyond Nvidia. U.S. cloud providers, chip equipment suppliers, memory makers, AI startups and research labs all operate in an ecosystem shaped by export controls and international standards. A U.S.-China AI guardrails framework, if it ever becomes concrete, could influence how companies document model risks, report dangerous capabilities, handle cybersecurity threats and design systems for high-stakes uses. For now, however, Trump’s comments offered few specifics about what those guardrails would cover or who would negotiate them.

The immediate takeaway is that AI has moved from a background issue in U.S.-China competition to a core diplomatic subject. The Trump-Xi talks did not resolve the conflict between commercial access and strategic control, but they confirmed that the future of AI governance and advanced chip exports will be negotiated at the highest levels of government. For Silicon Valley, that means business plans for AI hardware, frontier models and China exposure will remain tied to diplomatic signals as much as product road maps.