America’s AI Problem Isn’t Technology—It’s Thinking Too Transactionally

As Washington moves to support exports of full-stack artificial intelligence systems—including a recent Commerce Department initiative—it seeks to avoid repeating a familiar mistake.

In rare earths, the U.S. allowed processing capacity to migrate abroad, only to discover too late that supply chains had been reshaped. In 5G, it underestimated the power of integrated systems deployed at scale, ceding ground in global telecommunications infrastructure.

AI presents the same test.

The question is not whether the U.S. can lead in technology. It can. The question is whether it will compete in systems.

I first encountered that distinction in 1982, as an intern at Eastman Kodak. A colleague asked: Are you a project thinker or a process thinker?

The difference is simple. Project thinkers optimize individual outcomes. Process thinkers build systems that generate outcomes over time.

Kodak, founded by George Eastman, once mastered systems thinking. It built an ecosystem—film, processing and distribution—that strengthened with every user. Over time, it shifted toward optimizing products within an inherited model rather than adapting the model itself.

The result was not sudden collapse. It was gradual irrelevance.

That same shift—from systems to transactions—now defines a central challenge for the U.S.

At its most consequential moments, the U.S. has understood this. The Constitution created a governing system. The postwar order did the same globally—Bretton Woods, the Marshall Plan and NATO. The interstate highway system transformed commerce as a network, not a series of projects.

When America has led, it has done so by building systems—not by making better deals.

American business has often grasped this more clearly than policy. Henry Ford built a production system. Bill Gates and Andy Grove built platforms. Jeff Bezos built logistics and cloud systems. Elon Musk’s Starlink extends that logic into space.

In AI, U.S. policy is starting to move in this direction. The push to export full-stack AI systems reflects a broader recognition: competition is no longer about individual technologies, but about deploying integrated systems—compute, data, models, cybersecurity and applications—at scale. Efforts to pair these exports with financing from institutions such as the U.S. International Development Finance Corporation and the Export-Import Bank point in the same direction—toward building complete systems, not just selling components.

But unless that financing operates as part of a repeatable process—rather than a series of case-by-case transactions—the U.S. will struggle to match competitors that deliver integrated, scalable solutions.

The U.S. still leads at the frontier. But the contest will not be decided in the lab. It will be decided in deployment.

And that deployment is happening not primarily in the U.S. or China, but across the global middle—the countries in Asia, Africa, the Middle East and Latin America where infrastructure is still being built and system choices remain open.

These are not peripheral markets. They are decisive.

AI rewards scale. The global middle will generate the next wave of users, data and system adoption. The systems that take root there will shape standards and long-term influence.

Yet in much of the global middle, countries are not choosing between technologies. They are choosing between systems—who can build, finance, deploy and sustain them.

Here is where the U.S. risks falling short.

It still operates transactionally. Technologies are exported. Deals are negotiated. Financing is assembled case by case.

Others operate differently. In rare earths and telecommunications, China built integrated systems—combining infrastructure, financing and deployment into scalable offerings adopted across emerging markets. The return was not in any single deal. It was in the system those deals created.

This is not a failure of innovation. It is a failure of alignment.

Building world-class technology is hard. Building globally deployable systems is harder. It requires aligning capital, policy and execution—and integrating local partners, not just exporting solutions.

The U.S. is beginning to recognize this. But recognition is not execution.

Kodak did not fail because it lacked innovation. It failed because it began to think transactionally.

The risk for the U.S. is not sudden decline. It is gradual erosion.

It may continue to lead in innovation, yet find that others shape supply chains, define infrastructure and embed their systems across the global middle—the places where long-term influence is determined.

America has built systems that defined eras. The question is whether it will do so again—before others define the next one.

Because the future will belong not to those who make the best deals, but to those who build the systems within which all deals are made.

Development Impact: The countries of the global middle will determine how AI systems scale, setting the terms of access, standards, and long-term influence. If the U.S. fails to deploy integrated, accessible systems, developing economies risk dependence on alternative ecosystems with fewer choices and less resilience.