The United States and nine partners have signed the San Francisco Declaration on the Secure Foundations of Enduring Prosperity, setting out a shared approach to protecting the supply chains that advanced artificial intelligence depends on. The agreement, announced after a November 15 gathering of the US-led Pax Silica initiative, signals a move to treat chips, energy, industrial equipment and other enabling infrastructure as a joint economic-security concern rather than separate national purchasing problems.

Who signed and what was agreed

The signatories named in the US State Department's November 15 media note and declaration text are the United States, Australia, India, Israel, Japan, South Korea, Qatar, Singapore, the United Arab Emirates and the United Kingdom. The event was convened by Under Secretary of State for Economic Affairs Jacob Helberg.

The declaration links economic and national security, calling for the protection of sensitive technologies and critical infrastructure underlying what it terms “super intelligence” supply chains. It urges partners to reduce excessive dependencies, coordinate against coercive economic practices and address unauthorized access to key technologies. It also calls for working groups covering policy, adoption, supply-chain cooperation, security hubs and resilience.

Those are substantial political commitments. They are not, on their own, a new export-control law, a binding purchase contract or a guarantee that any shortage in advanced computing hardware will be resolved. Implementation will depend on later decisions by the participating governments and institutions.

Why AI supply chains are unusually exposed

Modern AI products rely on a fragmented physical and industrial system. Specialized chips may be designed in one country, manufactured with equipment sourced elsewhere, packaged in another location and installed in data centers dependent on reliable energy and networking. Software access and research collaborations introduce further dependencies. A disruption at one bottleneck can affect an entire chain, even when the final model is developed elsewhere.

The declaration tries to give partners a common vocabulary for mapping those vulnerabilities. In practical terms, this could mean better information sharing about critical components, closer scrutiny of investments or arrangements that give outsiders control over strategic infrastructure, and greater incentives to build alternative suppliers. But none of those outcomes should be reported as completed projects merely because the founding statement has been signed.

Competition with China, but wider commercial questions

Reuters framed the initiative against deepening rivalry with China. That geopolitical context matters, especially for advanced semiconductor manufacturing and critical materials. Yet the alliance's commercial consequences remain uncertain: friend-shoring may increase redundancy and political trust while also raising costs, complicating international research or leaving difficult questions about which suppliers qualify.

For businesses, a declaration does not automatically change current compliance requirements. Procurement teams should watch for subsequent national rules, funding programs, official technology classifications and new reporting obligations before revising supplier terms. Universities and research institutions should likewise distinguish research-security guidance from actual legal restrictions.

What happens next

The clearest test will be whether the proposed working groups produce specific measures, timelines and accountable delivery owners. Their outputs could reveal which parts of the AI stack governments regard as most vulnerable—from chipmaking machinery and mineral processing to data-center power and cloud services—and how they intend to share the expense of strengthening them.

The larger story is that AI leadership increasingly depends on more than model weights and software talent. It also depends on trustworthy, resilient physical systems. The San Francisco Declaration represents an attempt to coordinate that foundation, not proof that it has already been secured.

For a related perspective, see Nextwith.ai's analysis of why AI compute contracts are not the same as delivered capacity.

Sources