Give the AI Export Program What It Needs to Win

In 90 days, the Commerce Department’s American AI Exports Program persuaded companies that normally sell separate—and sometimes competing—parts of the artificial intelligence stack to assemble 78 proposals for exporting complete American-led AI systems. The packages span hardware, data, models, cybersecurity, and sector-specific applications. That response is a remarkable show of private-sector readiness. It is not yet an export strategy.

Designation alone cannot power a data center, finance a server fleet, prepare a procurement, train a workforce, or get an export license approved in time to close a deal. The program will matter only if Washington can convert “designated packages” into operating systems abroad faster than China can install and embed technologies that become too expensive to replace. China has learned to combine infrastructure, technology, finance, and diplomatic attention into an integrated offer. The United States still presents many buyers with a collection of agencies and vendors—and expects them to assemble Team USA.

Washington now has an opportunity to upgrade that model. The Commerce Department should select from the 78 applications that meet its demanding standards to create the supply base for a portfolio of American-led AI ecosystems. With finite deal-team capacity, Commerce should prioritize the packages tied to the most strategically sensitive markets and components first, so early designees receive support substantial enough to actually close deals, rather than a shallow show of interest spread across the full list. The harder task begins after designation: matching the right package to the right market and mobilizing the government behind it.

Designation Needs an Owner

The U.S. government has not yet made that operating model clear. Commerce has announced an integrated American AI export team and a National AI Center. Approved packages are also supposed to receive support through an interagency economic diplomacy group. But public materials do not identify an accountable federal quarterback for each transaction, nor do they disclose enough about staffing to show that the government can provide coordinated support at commercial speed.

It should. Each designated consortium should receive a federal account lead and a short deployment plan tied to specific countries, buyers, financing needs, licensing risks, and deadlines. Commerce should be explicitly responsible for driving the process, supported by a small interagency delivery unit that maintains a unified opportunity pipeline, convenes deal teams, assigns tasks, and elevates disputes. A foreign buyer should encounter one coordinated American offer—not a sequence of referrals from Commerce to the State Department, the Export-Import Bank, the U.S. International Development Finance Corporation (EXIM), and then back again.

There is already a useful institutional model. DFC has established a Strategic Working Office for Rapid Deployment, led by Vice President Marques Coleman, to bring experienced investment professionals to high-priority transactions, while its New York office connects the agency to financial and legal talent. The AI export program needs the same deal-making mentality across the government: small teams, clear ownership, access to senior decision-makers, and service standards measured in weeks rather than quarters.

Embassies must serve as the field force, and Commerce’s own Foreign Commercial Service (FCS) is the natural core of it. Ambassadors and country teams, including FCS officers posted in roughly 80 embassies, can identify buyers, convene ministries, businesses, and reveal obstacles invisible from Washington. What they lack is scale: enough federal funding and staffing to match the volume an AI export push will require. And many posts lack enough commercial, digital, technical, and finance expertise to assemble an AI infrastructure transaction. Priority markets should have properly staffed teams, reinforced by regional specialists who can surge to a deal rather than leaving each embassy to improvise.

Project Preparation Is Strategy

A robust project preparation capability determines whether an attractive concept becomes a bankable deal. Resources for feasibility studies, power and connectivity assessments, technical design, cybersecurity architecture, regulatory alignment, pilot programs, and procurement preparation are essential. China often provides these functions as part of an integrated offer. The United States too often waits for a fully formed transaction, limiting its ability to shape opportunities early and deploy capital where it can have the greatest strategic impact.

The U.S. Trade and Development Agency is designed to fill this gap. It funds project preparation and partnership-building that connect infrastructure priorities abroad to U.S. technology and services. Its scale, however, is minimal relative to its importance to this initiative. USTDA has reported an average of $231 in U.S. exports for every program dollar. Washington should have paired the launch of the AI export program with at least a doubling of USTDA funding and a dedicated AI Export Project Preparation Facility. It still should.

Match the Right Tool to the Right Deal

Commerce and State should identify priority markets and buyers. USTDA should prepare projects and shape transparent, technically sound procurements. Export-control officials should identify licensing risks before a proposal reaches a foreign cabinet or bank credit committee—not after months of negotiation. But U.S. Government advocacy and project preparation alone cannot make U.S. products cost-competitive, particularly in price sensitive markets in the Global Middle. EXIM should provide early financing indications through its ExportAI initiative. DFC should assess eligible investments in data centers, energy, connectivity, and other enabling infrastructure. The Millennium Challenge Corporation should incorporate power, connectivity, digital infrastructure, and workforce capacity into eligible, country-owned programs.

This machinery should support a diverse portfolio. A hospital network in Southeast Asia will not need the same combination of compute, models, connectivity, and local expertise as an industrial platform in Eastern Europe. Packages should therefore be modular, interoperable, and adaptable. Frontier systems will anchor U.S. technological leadership, but affordable mid-tier, open-weight, and domain-specific models will determine whether American technology reaches the volume market. The goal is not one national stack. It is an ecosystem broad enough to compete across countries with different costs, infrastructure, localization, and sovereignty requirements.

Do Not Let Content Rules Concede the Market

The program also exposes a tension between rebuilding American production and winning foreign markets now. In industry discussions, at least one major U.S. company questioned whether it should apply because the expected share of domestically produced hardware appeared difficult to meet. The objective is sound: taxpayer-backed programs should support American production and jobs. But electronic systems contain thousands of parts drawn from supply chains that cannot be relocated on a program deadline.

If Commerce’s designation standards or EXIM’s financing rules on content requirements are too restrictive, U.S.-led consortia may be sidelined while Chinese systems are installed. Once a competitor’s power architecture, cloud environment, models, data pipelines, workflows, and trained personnel are embedded, replacement becomes economically prohibitive and politically disruptive. The first deployment can determine the next decade of purchases, standards, and commercial relationships.

Commerce and EXIM therefore need a common transition policy. Initial eligibility should recognize the full American contribution: architecture, intellectual property, chips and other strategic components, software, cybersecurity, engineering, integration, and continuing services—not only the origin of every electronic part. Packages could commit to transparent, time-bound increases in domestic content, with narrowly tailored treatment for components that are not available at scale in the United States. Higher U.S. content should earn greater financing support, while sensitive components from adversarial suppliers remain subject to strict security rules.

That is not a case for abandoning domestic-content goals. It is a sequencing strategy: secure the installed base, then use procurement commitments and financing incentives to pull more production into the United States. America must build domestic capacity and win foreign markets at the same time. A policy that accomplishes only the first risks leaving U.S. factories without global customers; one that accomplishes only the second fails American workers and weakens supply-chain security.

The Race Begins After Selection

Every month Washington spends perfecting its process gives competitors more time to add users, improve local applications, train developers, lower costs, and embed standards and institutional relationships on the ground.

The American AI Exports Program should measure results that reveal market adoption: operating installations, time from designation to a financing indication, time to licensing decisions, active users, locally deployed applications, workers trained, private capital mobilized, and repeat purchases. Designations, memoranda, advocacy meetings, and announced dollar values are intermediate steps, not victories.

The U.S. industry is prepared to assemble exportable AI ecosystems and eager to work with the U.S. government. Washington must show that it can coordinate agencies, prepare projects, finance infrastructure, staff embassies, resolve licensing questions, and adapt content rules quickly enough to win actual deployments.

The United States still possesses the world’s strongest collection of AI technologies, companies, capital, and alliances. Whether American AI becomes the preferred foundation for countries building their digital future will be decided by deployment—and by whether Washington can move at the speed of the market.