Zuckerberg Warns US AI Lead at Risk Without Faster Infrastructure

Meta CEO Mark Zuckerberg has cautioned that the United States could lose its strategic edge in artificial intelligence unless it accelerates the development of energy capacity and data centers. He framed the competition as a defining factor for the future global balance of power, influencing democratic values, economic prosperity, and national security.

Zuckerberg acknowledged that the US currently holds an advantage in silicon design but is falling behind in the speed of building physical infrastructure. He pointed to China's rapid deployment of nuclear capacity, noting that countries like China are bringing online 1GW+ of nuclear capacity every other week, and argued that the US must speed up its own buildout to remain competitive.

While advocating for faster domestic development, Zuckerberg supported the continuation of export controls on silicon, saying they have been effective in slowing foreign AI labs during a critical period. He also warned that any policy delaying American model releases, even by a month, could significantly jeopardize US leadership and allow foreign models to surge ahead.

Zuckerberg proposed closer collaboration between frontier AI companies and the US government on national security. He suggested that leading labs provide the government with intermediate training checkpoints and technical staff to harden critical systems against emerging risks, without delaying public releases.

On the open-source front, he argued that the US and its allies should lead the ecosystem, rather than restricting access to foreign models. He stated that American open-source models should be the best globally, as restricting access could reduce quality and increase centralization.

Beyond geopolitical competition, Zuckerberg emphasized the importance of distributing advanced AI widely rather than concentrating it in a few institutions. He said Meta is focused on developing personal superintelligence for billions of people, which could strengthen individuals and create a balance of power. He also highlighted potential benefits in business, education, scientific research, and healthcare, while acknowledging risks related to cybersecurity, bioterrorism, surveillance, job displacement, and human control over powerful systems.

Zuckerberg said Meta would implement a governance structure where its independent board of directors can approve safety criteria for releasing AI models and review whether individual releases meet those criteria.