Nvidia's CEO Highlights Need for US-China Cooperation on AI Development Following Mythos Launch

US-China Cooperation on AI Urged Amid Tensions

Nvidia Corp. CEO Jensen Huang has expressed his concern that US-China tensions over trade and security issues have hindered coordination on crucial research, particularly in the field of artificial intelligence.

The issue came to the forefront in light of Anthropic PBC's breakthrough with its Mythos model, a highly powerful and secure AI technology. Huang believes that the US and China should engage in greater cooperation to ensure the safe use of AI, with a focus on agreeing on what not to use the technology for. This, he argues, is a safer approach than relying solely on unilateral efforts.

The recent easing of export restrictions by the Trump administration has allowed Nvidia to ship its less advanced H200 AI chips to Chinese customers. However, this decision has put Nvidia at odds with Anthropic, whose CEO favors stricter export controls. Despite their differing views on China, Nvidia remains a key supplier and investor in Anthropic, which makes the popular Claude chatbot.

Nvidia's Investment in Anthropic

Nvidia is committing $10 billion to Anthropic, a significant investment that highlights the importance of the partnership between the two companies. As part of a November deal, Anthropic is committed to taking up to 1 gigawatt of computing capacity from Nvidia. This move underscores the growing need for powerful computing resources in the development of advanced AI technologies.

China's Computing Capacity

When asked whether US export controls have constrained China's ability to access computing power, Huang pointed out that the country's abundant energy resources, skill in making mainstream chips, and ability to bundle more processors together make it well-equipped to handle the demands of AI research. The amount of computing capacity required to train Anthropic's Mythos model, for example, is "fairly mundane" and "abundantly available in China."

| Country | Computing Capacity Required to Train Anthropic's Mythos Model | | --- | --- | | China | Fairly mundane and abundantly available | | US | (No specific information provided) |

Huang noted that China has a large capacity for building chips and has many data centers that are sitting empty and fully powered, capable of handling even the most demanding AI tasks. This, he said, would allow China to access the computing resources it needs to advance its AI research without being constrained by US export controls.