US President Donald Trump and Chinese officials discussed artificial intelligence safety protocols and the status of Nvidia H200 chip shipments during a state visit to Beijing, Trump confirmed to reporters on Friday. The talks signal a potential, if tentative, opening for cooperation on a technology both nations view as critical to economic and military dominance. “We talked about possibly working together for guardrails,” Trump said on Air Force One.
The discussions centered on what Trump described as “standard guardrails that we talk about all the time,” a phrase that suggests the conversation may have revisited existing frameworks rather than breaking new policy ground. The South China Morning Post first reported the exchange on May 15, 2026. On the specific matter of Nvidia’s H200 graphics processing units, Trump confirmed the topic “did come up” and that he “thinks something could happen.” The chips, which are subject to US export controls, have not been approved for sale to China by Beijing.
Trump’s framing placed the onus squarely on China. “They chose not to” buy the hardware, he said, because China “wants to try and develop their own” domestic alternatives. That characterization was met with a pointed but non-specific response from Beijing. Chinese foreign ministry spokesman Guo Jiakun said on Friday that China had repeatedly stated its principled position on the issue, without elaborating.
He referenced Beijing’s long-standing opposition to the “abuse of export controls.”
US Trade Representative Jamieson Greer added a layer of diplomatic complexity. In an interview with Bloomberg TV on Friday, Greer stated that allowing the H200 shipments would be a “sovereign decision” for China. The remark effectively acknowledges that the primary barrier to the chips entering the Chinese market is not just Washington’s export restrictions, but Beijing’s own regulatory and strategic choices.
The exchange highlights a fundamental tension in the US-China technology relationship. The United States seeks to limit the flow of advanced semiconductors that could accelerate China’s military AI capabilities. China, in response, is pouring state resources into indigenous chip design and fabrication to circumvent those controls.
Here is what the study actually says. The H200 is not a consumer product. It is a data-center GPU designed for large-scale AI training and inference, representing a significant leap over its predecessor, the H100.
Restricting its sale is a cornerstone of US policy aimed at maintaining a technological edge. Yet Trump’s comments suggest a transactional view, where chip access could be linked to broader agreements on AI safety. Guo’s statement on AI cooperation was notably more expansive. “China has always advocated that all parties jointly promote the development of artificial intelligence in an open, inclusive, beneficial and good-for-all direction,” he said.
The language, while standard diplomatic phrasing, does not close the door on joint efforts. It contrasts with the more transactional tone from Washington. The backdrop to these talks is a global race to establish AI governance.
The European Union has pressed forward with its AI Act, a comprehensive risk-based regulatory framework. The United Kingdom hosted a global AI Safety Summit in 2023. The United Nations has established a high-level advisory body on AI.
China and the US, as the world’s two leading AI powers, have engaged in sporadic bilateral talks on the issue since 2024, but no binding agreement has emerged. A key question is what “guardrails” means to each side. For the US, the term often encompasses preventing AI from being used to develop biological or chemical weapons, launch cyberattacks, or generate sophisticated disinformation.
For China, the concept is frequently tied to “controllable” AI that aligns with socialist core values and does not threaten state authority. The semantic gap is wide. The chip discussion also unfolds against the reality of China’s domestic semiconductor progress.
Huawei’s Ascend series of AI chips has made significant strides, reportedly matching some capabilities of the Nvidia A100, though still lagging behind the H200 in performance and energy efficiency. Trump’s acknowledgment that China “wants to try and develop their own” is a recognition of this industrial policy reality. That timeline concerns US chipmakers.
Nvidia has consistently lobbied for a calibrated approach to export controls, arguing that overly broad restrictions would only accelerate China’s indigenous development and cost American companies billions in revenue. The company did not immediately comment on Trump’s remarks. The economic toll extends beyond a single company.
The Semiconductor Industry Association reported that US chip companies lost an estimated $8 billion in revenue due to export controls in 2025 alone. The H200 decision, if it moves toward approval, could signal a shift in the cost-benefit analysis within the Trump administration. Behind the diplomatic language lies a hard strategic calculus.
The US needs China’s cooperation on AI safety to prevent risks. China needs advanced chips to fuel its AI ambitions. Both needs are in direct conflict.
The Beijing talks did not resolve this contradiction. They merely restated it in public. A vivid detail from the trip: Trump delivered his remarks on Air Force One during the return flight, a setting that often produces more candid commentary than formal press conferences.
The informality of the venue did not extend to the substance, which remained carefully hedged on both sides. Why It Matters: The talks represent the highest-level US-China engagement on AI governance since Trump’s return to office. Any agreement on AI guardrails, even a non-binding one, would reshape the global regulatory landscape and affect how companies from OpenAI to Huawei develop and deploy advanced models.
The H200 chip decision, meanwhile, will signal whether Beijing is willing to remain dependent on US hardware or is fully committed to a decoupled domestic supply chain. Key Takeaways: - Trump confirmed AI guardrails and Nvidia H200 chips were discussed during his Beijing visit, but characterized the talks as routine. - Beijing has not approved H200 imports, a choice Trump attributed to China’s desire to develop domestic chip alternatives. - US Trade Representative Greer called the H200 decision a “sovereign” matter for China, shifting the narrative from a pure export control issue. - China’s foreign ministry reiterated its opposition to the abuse of export controls while leaving the door open for inclusive AI cooperation. What comes next is a series of technical and diplomatic steps.
The US and China are expected to hold working-level talks on AI safety in Geneva later this quarter, according to a Reuters report from April 2026. The outcome of those meetings will test whether the presidential-level conversation translates into concrete risk-mitigation measures. For the H200, the next milestone is a decision by Beijing’s Ministry of Commerce on whether to certify the chips for import.
No timeline for that decision has been announced. The global semiconductor industry will be watching.
Key Takeaways
— - Trump confirmed AI guardrails and Nvidia H200 chips were discussed during his Beijing visit, but characterized the talks as routine.
— - Beijing has not approved H200 imports, a choice Trump attributed to China's desire to develop domestic chip alternatives.
— - US Trade Representative Greer called the H200 decision a "sovereign" matter for China, shifting the narrative from a pure export control issue.
Source: South China Morning Post
