Anthropic advocates for stronger export controls on AI chips while Nvidia opposes restrictions.

Anthropic has urged the White House to tighten so-called AI diffusion rules that are already set to hurt Nvidia and other chip makers by limiting or blocking the sale of higher-end GPUs and accelerators outside the US and select allies from mid-May. In a briefing note and formal submission to the US Department of Commerce, the San Francisco-based chatbot maker argued that these looming export controls won't go far enough to stem the flow of smuggled chips fueling China's continued advancement in artificial intelligence. The company wants tougher limits on the legitimate sale of AI-targeted processors—particularly from Nvidia—to various nations, along with a hard crackdown on smugglers using countries where such chips remain legal as a proxy to sneak them into China against US export rules.

The Biden administration published these export rules in mid-January, establishing three progressively more restrictive tiers with China placed in tier three. Countries in the first tier retain unfettered access to American-designed accelerators, while those in the second face strict import limits. The third tier effectively blocks purchases unless the US grants an exception. Anthropic's push for tighter controls follows Western AI companies' alarm at China's DeepSeek, which released a free LLM in late January that beat their models on some benchmarks without requiring quite as many resources to train as its American rivals.

In its submission, Anthropic stated: "The US currently leads in advanced semiconductor technology and export controls capitalize on the trend of computing power doubling every two years, so while US chip technology continues advancing, China's progress is slowed." The company projected that "by 2027, countries using older chips could face AI training costs that are ten times higher than those with cutting-edge American technology." Specifically, Anthropic wants the Trump administration to slash the amount of compute tier-2 nations can purchase without permission, noting that "currently, countries in tier two can buy the equivalent of 1,700 Nvidia H100 advanced chips without needing government permission — roughly $40 million worth of technology. This creates a potential loophole for smuggling." The company proposed increasing allocations to tier-2 countries only if they maintain "robust" datacenter security and establish "government-to-government agreements that prevent smuggling and align technology controls," while calling for "enhanced resources for the Bureau of Industry and Security" to improve control effectiveness.

Anthropic's position stands in stark contrast to Nvidia CEO Jensen Huang, who on the same Wednesday called on the Trump administration to loosen restrictions, stating: "We need to accelerate the diffusion of American AI technology around the world. The policies and encouragement from the administration really need to support that." Huang's efforts reportedly included attending a $1 million-a-head dinner with the President at Mar-a-Lago. Earlier in the month, Nvidia booked a $5.5 billion charge tied to newly implemented export controls on H20 accelerators to China, including Hong Kong, and other nations of concern.

Anthropic's stance is less risky than it might appear given Nvidia's dominance in the AI ecosystem. The California lab maintains a close relationship with Amazon Web Services, one of its largest backers, and makes extensive use of AWS's custom Trainium AI accelerators to develop its Claude family of models. Late last year, AWS revealed Project Rainier, an AI supercomputer containing hundreds of thousands of Trainium2 accelerators set to power on this year, providing Anthropic with "5x the number of exaFLOPS used to train its latest generation of AI models." For Anthropic, stifling Chinese progress in AI through tighter export controls is arguably more important than maintaining favorable relations with Huang.