The global race over artificial intelligence is often described as a competition for scale: more chips, larger data centers, bigger models, greater energy supply and wider access to data. That framing may be understandable, but it may also push the United States into a contest that favors China’s strengths more than America’s.
After recent high-level discussions between President Donald Trump and Chinese President Xi Jinping, artificial intelligence has again moved to the center of U.S.-China strategic competition. American officials have emphasized the need for AI guardrails that promote both innovation and safety. That is an important starting point, but guardrails alone are not a national strategy.
The more important question is what kind of AI system the United States should build if it wants to maintain long-term leadership.
A race built only around centralized AI could become difficult for the United States to win on its own terms. China’s political system allows Beijing to coordinate state power, industrial policy, energy resources, surveillance infrastructure, capital and data access in ways that democratic societies cannot — and should not — copy. If the competition is reduced to who can centralize the most computing power and data under the strongest government direction, China may be able to close the gap over time.
TRENDING TODAY
That does not mean America is at a disadvantage in the broader AI race. It means the United States should avoid fighting the contest only on terrain that benefits an authoritarian system.
America’s advantage lies elsewhere: open markets, private-sector speed, democratic institutions, individual trust, entrepreneurial experimentation and strong partnerships with allies. Those strengths point toward a different AI architecture — one that is more distributed, more resilient and closer to the people making decisions.
Instead of relying entirely on massive centralized systems, the United States should invest more heavily in distributed AI. In simple terms, that means AI tools should not live only in distant cloud data centers. They should also operate on trusted devices, local networks, mission systems, laptops, workstations, vehicles, ships and edge equipment.
That approach would allow individuals, teams and organizations to use AI in real-world conditions without depending on constant access to the cloud. It would also reduce the risk that one central system becomes a single point of failure.
The military offers one of the clearest examples. In a conflict, forces cannot always assume they will have stable communications, uninterrupted satellite links or reliable access to centralized cloud infrastructure. Cyberattacks, electronic warfare, physical damage and contested environments can all disrupt connectivity.
In those conditions, AI must work where decisions are being made.
A maintainer repairing an aircraft in the middle of the night, a medic moving through a blackout zone, a small military unit operating from a remote outpost, or a shipboard team working under strict emissions controls cannot wait for a distant server to respond. They need trusted tools that can function locally, even when networks are degraded or disconnected.
This is where distributed AI could become a major American advantage. AI systems are becoming smaller, more efficient and more specialized. Tasks that once required large data centers can increasingly be performed on smaller devices. That shift opens the door to AI that is tailored to specific missions, professions and workflows.
A doctor does not need the same AI system as a pilot. A front-line soldier does not need the same model as a logistics planner. A ship engineer, border officer, factory worker, emergency responder or cybersecurity analyst may each need AI trained around specific data, permissions and operational needs.
That kind of specialization is difficult to achieve through one giant centralized model alone. A distributed system can support many smaller models designed for specific environments. Those models can operate locally, protect sensitive data and adapt to the needs of the people using them.
This approach also fits America’s political values better. A centralized AI system controlled by a small number of powerful institutions may create concerns over surveillance, access, control and abuse. A distributed model can give users and organizations more authority over the tools they depend on, while still allowing for standards, oversight and interoperability.
That does not mean the United States should abandon frontier models or large-scale AI infrastructure. American companies still need to build powerful models, advanced chips and major data centers. The cloud will remain essential. But the cloud should be a support layer, not the only foundation of national AI strategy.
A strong AI system should be able to operate in both connected and disconnected environments. It should be secure, flexible, interoperable and trusted by the people using it. It should strengthen decision-making at the edge rather than forcing every decision back to a central command structure.
This is especially important for national security and allied cooperation. The United States works with partners across different regions, militaries, industries and legal systems. Distributed AI could make it easier to share capabilities without requiring every ally to depend on a single centralized platform. It could also make joint operations more resilient when communications are disrupted.
China’s model is likely to continue emphasizing state direction and centralized control. America should not try to imitate that model. Its best chance is to build AI systems that reflect the strengths of a free society: flexibility, trust, competition, decentralization and responsible innovation.
The AI race is not only about who builds the biggest model. It is also about whose systems are more useful, more trusted and more resilient in the real world.
If Washington focuses only on scale, it may miss the area where the United States can build a lasting advantage. The AI race America can actually win is not just in the data center. It is at the edge — in the hands of the people, teams and institutions that need intelligence where decisions are made.
Why It Matters
The debate over AI strategy matters because the United States is competing with China not only in technology, but also in political systems and national security models. A centralized AI race could favor Beijing’s ability to coordinate state power, while a distributed model better reflects America’s strengths.
It also matters because real-world AI must function under pressure. In military, medical, industrial and emergency settings, users may not always have reliable cloud access. Distributed AI could make American systems more resilient, useful and trusted.
What Comes Next
U.S. policymakers should continue discussing AI safety and guardrails, but they also need a clearer strategy for AI architecture. That means supporting both frontier model development and edge-based systems that can operate locally in high-risk environments.
The next stage of the AI race may be decided not only by who builds the largest models, but by who builds the most adaptable and trusted systems for real-world use.
Lawmakers have also framed the U.S.-China AI competition as a national security issue, including proposals focused on limiting adversaries’ access to advanced semiconductor technology.
A world where Communist China leads in AI is bad for the United States and for the entire world. It’s critical that we do not lose this race.
That’s why I introduced the MATCH Act. This bill ensures that Communist China and other adversaries cannot purchase key semiconductor… pic.twitter.com/QzyzE76x1h
— Senator Pete Ricketts (@SenatorRicketts) June 3, 2026
