OpenAI Restrains Public Comments on China Chip Controls

Nvidia’s dominance in the AI chip market gives it significant influence over customers’ business decisions and U.S. export policy, The Information reported on the 2nd, local time. The outlet said OpenAI has also been cautious about publicly discussing controls on advanced semiconductor exports to China, concerned that doing so could strain its relationship with Nvidia.

According to the report, OpenAI Vice President Noam Brown recently shared an article in the company’s internal Slack about changes to its China semiconductor export policy and asked about the policy. Policy staff pointed him to previously published materials on advanced AI safety, but did not directly address China export controls. They said specific discussions should take place with global policy officials rather than on Slack. CEO Sam Altman and other executives reportedly held back public comments over the past year because they were concerned about upsetting Nvidia CEO Jensen Huang. During U.S. lawmakers’ discussion of related legislation last year, a policy staffer reportedly told a congressional contact that OpenAI was speaking less openly than under the Biden administration for fear of provoking Huang.

The report said Nvidia’s influence extends to customers and Washington. Some cloud companies weigh business decisions carefully to avoid antagonizing Nvidia, which controls supplies of GPUs used to train and run AI models. One company that uses AMD GPUs said customers were reluctant to disclose their use of AMD chips because they feared Nvidia’s reaction. Nvidia reportedly challenged researchers and policy experts who criticized China export policy and warned that renewing their funding could be difficult. Some research institutions were also accused of using funding and influence to push back against researchers critical of the controls.

Chris McGuire, a senior fellow at the Council on Foreign Relations, said Nvidia has overstated the performance of China’s domestic AI semiconductors, increasing policymakers’ distrust. He said data show Chinese AI chips lag Nvidia substantially in both quality and volume, while Nvidia describes the situation differently. Nvidia has argued that its chips should be used in overseas markets to strengthen U.S. competitiveness in AI infrastructure. Spokesperson John Rizzo said that if cloud computing around the world relies on foreign chips, models and applications will also be built overseas.

OpenAI says strong export controls support AI infrastructure development in democracies and give the United States and its allies time to strengthen safety standards, oversight systems and international cooperation. The company remains cautious about the specific controls that should apply to China. OpenAI is also developing its own AI chip, Jalapeño, as a competitor to Nvidia, and claims it outperformed Nvidia’s Blackwell GPUs on some benchmarks. The company plans to use the chip only in the United States and allied countries, not sell it to China.

The U.S. Congress is expected to resume discussions of advanced semiconductor export-control legislation after the midterm elections. The report noted speculation that Nvidia’s aggressive lobbying could increase opposition in Washington and weaken its policy influence. McGuire said Nvidia’s technical expertise and market insight could help shape policies that protect national security while supporting business growth, but that doing so requires trust.

Nvidia’s dominance in the AI chip market gives it significant influence over customers’ business decisions and U.S. export policy, The Information reported on the 2nd, local time. The outlet said OpenAI has also been cautious about publicly discussing controls on advanced semiconductor exports to China, concerned that doing so could strain its relationship with Nvidia.

According to the report, OpenAI Vice President Noam Brown recently shared an article in the company’s internal Slack about changes to its China semiconductor export policy and asked about the policy. Policy staff pointed him to previously published materials on advanced AI safety, but did not directly address China export controls. They said specific discussions should take place with global policy officials rather than on Slack. CEO Sam Altman and other executives reportedly held back public comments over the past year because they were concerned about upsetting Nvidia CEO Jensen Huang. During U.S. lawmakers’ discussion of related legislation last year, a policy staffer reportedly told a congressional contact that OpenAI was speaking less openly than under the Biden administration for fear of provoking Huang.

The report said Nvidia’s influence extends to customers and Washington. Some cloud companies weigh business decisions carefully to avoid antagonizing Nvidia, which controls supplies of GPUs used to train and run AI models. One company that uses AMD GPUs said customers were reluctant to disclose their use of AMD chips because they feared Nvidia’s reaction. Nvidia reportedly challenged researchers and policy experts who criticized China export policy and warned that renewing their funding could be difficult. Some research institutions were also accused of using funding and influence to push back against researchers critical of the controls.

Chris McGuire, a senior fellow at the Council on Foreign Relations, said Nvidia has overstated the performance of China’s domestic AI semiconductors, increasing policymakers’ distrust. He said data show Chinese AI chips lag Nvidia substantially in both quality and volume, while Nvidia describes the situation differently. Nvidia has argued that its chips should be used in overseas markets to strengthen U.S. competitiveness in AI infrastructure. Spokesperson John Rizzo said that if cloud computing around the world relies on foreign chips, models and applications will also be built overseas.

OpenAI says strong export controls support AI infrastructure development in democracies and give the United States and its allies time to strengthen safety standards, oversight systems and international cooperation. The company remains cautious about the specific controls that should apply to China. OpenAI is also developing its own AI chip, Jalapeño, as a competitor to Nvidia, and claims it outperformed Nvidia’s Blackwell GPUs on some benchmarks. The company plans to use the chip only in the United States and allied countries, not sell it to China.

The U.S. Congress is expected to resume discussions of advanced semiconductor export-control legislation after the midterm elections. The report noted speculation that Nvidia’s aggressive lobbying could increase opposition in Washington and weaken its policy influence. McGuire said Nvidia’s technical expertise and market insight could help shape policies that protect national security while supporting business growth, but that doing so requires trust.