Reading about the proposed Stratos AI project in Utah, I found myself thinking about Dwight Eisenhower. Not because of the technology. Not because of artificial intelligence. Because of incentives.
Eisenhower’s warning about the Military Industrial Complex is one of the most quoted political observations of the last century. It is also one of the most misunderstood. He was not warning Americans about a conspiracy. He was warning them about what happens when government priorities, corporate interests, capital investment, national security concerns, and public policy all begin reinforcing one another. When enough powerful interests benefit from the same outcome, a self-sustaining ecosystem emerges. Nobody needs to coordinate it. Nobody needs to secretly control it. The momentum comes from aligned incentives and enormous amounts of money.
The proposed Stratos project may prove to be a perfect example of that dynamic. The project envisions an AI-focused campus spread across roughly 40,000 acres with power requirements that could eventually reach as much as 9 gigawatts. To put that into perspective, that is the kind of number normally associated with regional infrastructure planning, not a single technology project. Whether Stratos ultimately gets built exactly as proposed is less important than what it represents. What caught my attention was the coalition of interests surrounding it and the assumptions increasingly being made about AI infrastructure across the industry.
Technology companies need more compute. GPU manufacturers need larger deployments. Utilities need new demand growth. Pipeline operators need customers. Construction companies need projects. Investors need returns. Politicians want economic development. National security officials want America to maintain technological leadership over China. Every participant has a rational reason to support expansion. No conspiracy is required. The incentives are already aligned.
That alignment should not automatically concern us. Large infrastructure projects have always required coalitions of economic and political interests. Railroads, highways, airports, electrical grids, telecommunications networks, and semiconductor fabrication plants were all built because multiple stakeholders saw value in making them happen. The difference is that those projects generally produced benefits that were relatively easy for the public to understand. People could see the roads, drive on the highways, work in the factories, and understand the connection between the investment and the resulting economic activity.
The AI infrastructure buildout is different. It is happening at a scale that is difficult for most people to comprehend while producing benefits that are often indirect and concentrated. We are talking about projects measured not in acres but in tens of thousands of acres. We are talking about power requirements measured in gigawatts rather than megawatts. We are talking about capital expenditures that rival major public works projects. Yet the economic case is often presented with surprisingly little scrutiny.
One of the first questions local communities should ask is also one of the simplest: How many permanent jobs are being created?
The answer matters because these projects are increasingly being sold as economic development engines. Construction employment is real and significant. Large projects create opportunities for contractors, tradespeople, engineers, and suppliers. But construction eventually ends. Once operational, modern AI data centers are designed for efficiency, automation, and scale. They are not labor-intensive facilities. The long-term employment numbers often bear little resemblance to the public perception created by the size of the investment.
That does not mean these projects lack value. It does mean we should be honest about what kind of value they create. A manufacturing plant employing thousands of workers has a different economic profile than a highly automated computing facility consuming massive amounts of energy. The distinction matters because communities are being asked to make long-term commitments involving land use, infrastructure, utility capacity, and public resources. Those commitments deserve careful analysis rather than assumptions that every large investment automatically translates into broad-based prosperity.
The shift in language surrounding AI infrastructure is equally noteworthy. A decade ago, large data centers were typically discussed as technology projects. Today, they are increasingly framed as strategic assets. Discussions about AI infrastructure quickly become discussions about competition with China, national security, technological sovereignty, and America’s future economic leadership. There is truth behind these concerns. Artificial intelligence will almost certainly become a foundational capability for governments, businesses, and military organizations. Countries that lead in AI development will enjoy meaningful advantages.
The danger is that national security arguments have a tendency to overwhelm other legitimate questions. Once a project is labeled strategically important, skepticism can be portrayed as opposition to progress. Questions about costs, tradeoffs, and local impacts become politically more difficult to raise. We begin evaluating projects primarily through the lens of what they contribute to the broader strategic competition rather than what they contribute to the communities expected to host them.
This is where Colossus: The Forbin Project unexpectedly comes to mind.
Most people remember the 1970 film as a story about a supercomputer taking control. What made the film interesting, however, was not the machine itself. The lesson of Colossus was never that machines become evil. The lesson was that dependence changes behavior. Once society concluded it could not function without the system, the system stopped being optional.
I am not suggesting AI infrastructure is leading us toward a science-fiction dystopia. I am suggesting that dependency has a way of reshaping priorities. As AI becomes more deeply embedded in economic growth, national competitiveness, financial markets, defense planning, and corporate strategy, the pressure to continuously expand supporting infrastructure will only increase. Every new generation of models will require more compute. More compute will require more power. More power will require more infrastructure. More infrastructure will require more capital and more political support. The cycle reinforces itself.
That is why the emergence of what might reasonably be called an AI Industrial Complex deserves attention. Not because it is sinister, but because it is becoming powerful. The combination of technology companies, infrastructure providers, investors, government agencies, and national security interests is creating an ecosystem with enormous momentum. Such ecosystems are not inherently bad. They helped build many of the systems modern society depends upon. But they do deserve scrutiny, particularly when the costs and benefits are distributed unevenly.
The most important debates about AI over the next decade may have less to do with model intelligence than many people assume. The industry spends enormous amounts of time discussing benchmarks, capabilities, and hypothetical future breakthroughs. Meanwhile, the real-world constraints are becoming increasingly obvious. Energy availability, infrastructure capacity, financing, permitting, public acceptance, and resource allocation may ultimately determine the pace and direction of AI adoption more than any single technical advancement.
The question facing communities, policymakers, and business leaders is not whether AI matters. It clearly does. The question is whether we are asking sufficiently hard questions about the infrastructure being built in its name. History suggests that once industrial ecosystems achieve enough scale and momentum, changing direction becomes extraordinarily difficult. That is precisely why the time to examine incentives, tradeoffs, and long-term consequences is before the infrastructure becomes indispensable rather than after.
Shimmy’s Take
The debate over AI has largely focused on what the technology will do. We spend far less time discussing what supporting the technology requires. The Utah project caught my attention not because it was unusually large, but because it made visible something that is happening all around us. AI is no longer just a software story. It is becoming an infrastructure story, an energy story, a capital allocation story, and increasingly a national security story.
Eisenhower understood that powerful systems do not emerge because someone planned them. They emerge because incentives align.
Looking at the AI infrastructure race, those incentives are aligning right now.
That is why the most important question may not be whether AI changes the world.
It may be whether we understand the world we are building around AI before it becomes too difficult to build anything else.
