How the US plans to dominate global AI infrastructure

Summary

• What’s the plan? Through its AI Exports Program, the US is using development finance to push the world into using the ‘full stack’ of American AI – chips, data centers, cloud services, models, and applications.
• Trade-offs: Partner countries gain cutting-edge technology, but with implications for their sovereignty and ongoing dependence on American tech. 
• What to watch: The Trump Administration’s designation of Anthropic as a supply chain risk will make partner governments think twice about relying on the US AI stack. 
• The bigger picture: If the US AI ecosystem is perceived as less reliable, partner countries will hedge – by focusing on sovereign capability or looking to China.

‍

The US currently leads the world in frontier AI models, chip design, and cloud infrastructure. But while it’s one thing to have the world’s best technology, it’s another to have the world use it.

The Trump Administration's Executive Order 14320 on Promoting the Export of the American AI Technology Stack, signed in July 2025, represents Washington’s most explicit policy lever to press America’s AI advantage globally.

The idea is that influence flows not only through frontier AI models, but through the infrastructure underpinning them. Compute is physical, quantifiable, and concentrated among a few vendors, making it uniquely tractable for exerting influence over how AI is used worldwide. 

The strategic rationale

White House Office of Science and Technology Policy (OSTP) Director Michael Kratsios told lawmakers in September that exporting the AI technology stack was ‘the most important part’ of the administration's AI Action Plan. Kratsios argued it was incumbent on the US government to promote its AI technologies broadly, 'so that when [China] has the capacity to actually export chips themselves, we are already there and already around the world.'

Kratsios said the idea for the program originated during his experience in the first Trump administration, when he was trying to convince allied governments to replace Huawei in telecommunications infrastructure. The lesson was that early Chinese dominance in 5G generated enduring strategic vulnerabilities for Washington and its allies. 

Understanding the AI stack

The Executive Order defined the stack across the following components:

• Chips, servers, and other AI-optimized hardware 
• Data center storage and cloud services
• Data pipelines and labeling systems
• AI models and systems
• AI cybersecurity measures
• AI applications for specific use cases, such as healthcare or agriculture 

Companies accepted into the AI Exports Program receive federal financial and diplomatic support to export the US AI stack to foreign markets. 

The program works by encouraging so-called industry-led consortia – groups of companies that together can offer integrated ‘full-stack’ packages to foreign governments. This echoes the NVIDIA-AWS partnership to deploy sovereign AI clouds globally, where NVIDIA supplies the hardware and AWS provides the cloud platform and security framework.

What about AI sovereignty?

The AI Exports Program faces headwinds from the global push for AI sovereignty. Countries increasingly want AI infrastructure within their borders and are seeking to prioritize domestic firms.

The EU announced a EUR 1 billion plan in October 2025 to accelerate AI development in key industries, while France, Japan, and the UAE are all pursuing national AI infrastructure strategies. 

Partner countries may welcome US investment while hedging against long-term vendor dependence. As the Information Technology and Innovation Foundation (ITIF) has argued, for the AI Exports Program to succeed, Washington must offer partners ‘a genuine stake in building a global AI ecosystem in which they can build products and services, capture value, and create jobs.’

Washington appears alert to this tension. At the India AI Impact Summit in February 2026, Kratsios told attendees: ‘Real AI sovereignty means owning and using best-in-class technology for the benefit of your people, and charting your national destiny in the midst of global transformations.’ 

Kratsios appeared to be trying to recast dependency as agency – nations could keep sensitive data within their borders while enabling access to American frontier AI capabilities. 

What to watch

The Department of Defense’s designation of Anthropic as a supply chain risk this month has put the AI Exports Program on shaky ground.

One of the policy’s primary authors, Dean Ball, who has since left his position in the OSTP, went so far as to call the program ‘dead on arrival’. As Ball wrote, ‘If corporations and foreign governments just cannot trust what the US government might do next with the frontier AI companies, it means they cannot rely on that US AI at all.’

For US allies and partners, the supply-chain risk designation will only reinforce existing concerns about the durability of Trump-era technology arrangements. Such concerns had already been heightened for UK policymakers in December, when the US suspended the USD 40 billion US–UK Tech Prosperity Deal over broader trade and market‑access disputes. 

Adding to the uncertainty, the Commerce Department is reportedly drafting new tiered rules for AI chip exports that would require allied governments to provide security assurances and make investments in US domestic data centers. While the AI Exports Program has been pitched as the carrot, these new rules would give Washington a regulatory stick over its most critical component.

If this pattern continues, US partners will likely hedge more openly toward domestic AI strategies and away from costly AI infrastructure deals with an administration they see as unreliable on AI policy.

America’s AI industry will also scrutinize whether Washington can maintain policy continuity within and between administrations. If flagship AI firms find themselves caught between competing government agencies, the export program's industry coalition could fracture before it consolidates.

Another pressing question is whether the Department of Commerce can issue revised guidance that resolves previous confusion around the consortia requirement. Without clarity, major firms will likely pursue bilateral arrangements outside the program framework.

The bigger picture

So long as the AI Exports Program struggles with the fallout from the Anthropic decision and its own implementation delays, the US risks ceding ground to Chinese alternatives. 

Through its Digital Silk Road, Beijing has paired state-backed finance with exports of its own AI stack to governments across Asia, Africa, and the Middle East. Companies like Huawei Cloud now market integrated cloud‑and‑AI packages in Saudi Arabia, the UAE, and North Africa, positioning Chinese infrastructure as the backbone for local e‑commerce and ‘smart’ public services. Notably, Beijing’s greater willingness to export surveillance and security technologies makes its stack particularly attractive to less democratic governments. 

The decision for partner governments is less about choosing a single ‘stack’ and more about managing exposure to competing US and Chinese AI ecosystems. Neither offer is without strings. But abrupt policy shifts in Washington mean that American AI – even though it is world-leading – may no longer be the obvious choice.