Project Macedonia: Building Infrastructure for the AI Age

The current global data center buildout is an unprecedented growth catalyst for Texans. Texas has the unique opportunity to house the physical foundations of the next technology frontier and keep all the investment which comes with it. Texas understands that ambition, and Governor Greg Abbott is ensuring that all Texans share in its effectuation.

Texas has earned its place among the world's economic leaders. The governor's office reports that its economy reached $2.9 trillion in 2025, ranking eighth against the nations of the world. This is a state with the scale to matter and the confidence to think bigger.

Our view is that Texas is on track to become the world's third-largest economy as this AI infrastructure buildout matures. That is our claim about its potential, not an official forecast, because we expect the state to attract over two-thirds of the projected $5 trillion investment in U.S. AI infrastructure by 2030. The point is the direction: Texas is building for a future larger than the one it currently occupies.

Our view is that Texas is on track to become the world's third-largest economy as this AI infrastructure buildout matures.

Our team has spent more than two decades witnessing the growth of data centers, through cycles, crashes, and comebacks. This excitement around AI is like nothing we have witnessed before. The underlying story is familiar.

Every age has its infrastructure, and the places that build it become the places where the next economy takes root. Our founder's note, "Every age has its infrastructure. States can build the AI age," follows the evolution of infrastructure buildouts from railroads through electrification, to digitization and into AI. Each era changed everyday life through physical systems that someone had to finance, build and maintain.

AI now belongs in that history, which is being written in real time even as you read about it here. The experience on a phone or computer may feel weightless, but the infrastructure behind it is anything but. It runs on buildings full of servers, cooling systems, and electricity. For Texas, that means the opportunity is tangible. It is construction, equipment, energy, and an enduring place in the global economy being built around those facilities.

Governor Abbott is right to put Texans at the center of that opportunity. His Sept. 21 directive halted permits sought by data centers until the state's audits are completed. His stated terms are clear: data centers must pay their own way, protect the grid and water, and respect the communities where they operate.

That is a demonstration of true leadership while facing unprecedented growth which can often go unchallenged. Power use, water consumption, and community impacts are part of the project, not details separate from it. Texas is treating them that way. Scrutiny of those commitments recognizes the scale of what is being built and the importance of getting the public bargain right.

These policies are the right response to speculative projects. The audits examine power supply, water consumption, financial assistance and ownership. They recognize something builders know: a rendering is not a completed facility. Distinguishing credible commitments from speculative proposals is sound stewardship of resources that Texans depend on.

Cooling technology is part of that scrutiny. The audit expressly asks whether facilities will use air-cooled, closed-loop or other water-efficient cooling systems. It puts the technology behind a developer's promises into the public assessment of the project.

Energy receives the same serious attention. The audit examines developers' progress in procuring generation with a focus that they ought to fund the electrical infrastructure serving them. That is the right framework for evaluating how renewable supply, including solar and wind farms, is funded alongside clean generation such as nuclear. The support for advanced nuclear development is progressive energy policy management.

There is no contradiction between welcoming the future and taking those responsibilities seriously. Confidence does not mean indifference to costs. The state's approach expresses confidence that it can attract investments, while concurrently ensuring it serves all its residents.

The benefits are bigger than any one facility. Construction work, a durable tax base and investment in energy infrastructure can support the communities that host this new economy. Those benefits matter precisely because the buildings are real, the commitments are long-lived, and the people around them have a stake in the outcome.

Texas is doing this right. It is a state accustomed to thinking at scale, and it is approaching the AI age with both ambition and accountability. Governor Abbott deserves credit for keeping Texans first as this next chapter takes shape. This data center boom is a once-in-a-generation opportunity for Texans to occupy a podium position alongside global superpowers.

Texas is doing this right.

The pause wave

On July 14, 2026, Governor Kathy Hochul signed Executive Order No. 62. It temporarily holds certain state environmental permit applications for data centers that had not yet been deemed complete, while New York develops an environmental review and standards. It does not bar every data center project or local permit. The scale behind that decision is real: as of May 2026, nearly 12 gigawatts of data center load requests were in New York's interconnection queue, more than 8 gigawatts of them added in 2025. New York must use this pause to write clear, demanding rules and then give serious projects a path forward.

New York is not alone. In August, Governor Greg Abbott announced a moratorium on new data centers connecting to the Texas grid, and on September 21 he extended it to environmental permits, citing water and energy. Governor Josh Shapiro took a different path in Pennsylvania: no pause, but an executive order requiring developers to meet strict public requirements and win local approval before the state reviews a single permit, with state agencies barred from signing nondisclosure agreements on data center deals. Maine's legislature had passed a pause on large data centers before Governor Janet Mills vetoed it in April. In 2025, more than 40 states considered 267 bills touching data centers. The pressure is national. So is the opportunity to replace indefinite prohibition with a serious public bargain.

We say that as a team that has spent more than two decades witnessing the growth of data centers. We have watched this industry through cycles, crashes, and comebacks. And we truly believe states can meet this moment with higher standards instead of closed doors.

Every age has its infrastructure. The states that set sound terms for building it shape what comes next.

The railroad age

In the nineteenth century, the infrastructure of the age was steel rail. Towns fought and struggled to get a depot. Counties that won the junction became cities. Counties that blocked the line, or got bypassed, stayed farms. Everyone understood the stakes, because they could see them: the railroad decided where grain got sold, where factories got built, and where young people stayed or left.

The buildout was messy. Speculators ran wild. Towns handed out subsidies to paper railroads that never laid a mile of track. The panic of 1873 was, at its core, a railroad bubble popping, and it dragged the country into a depression. Calls to slow construction after that crash would have been understandable. The lesson was to deal with speculation and risk while keeping useful routes open.

Here is what actually happened. The speculators went broke. The rails stayed. By 1916 the network peaked at 254,251 miles of road, according to the Historical Statistics of the United States, and those miles carried the entire industrial age: the steel, the coal, the cattle, the mail-order catalogs, the Sears houses. The bust punished the gamblers. The infrastructure rewarded everyone else for a century.

Lesson one: the infrastructure outlives the hype cycle. The capital that chases a boom can evaporate. The capacity that gets built keeps compounding. A state that permanently blocks the buildout does not protect itself from the bust. It just opts out of the century.

The electrification age

At the turn of the twentieth century, electricity was terrifying. People genuinely feared it. Newspapers ran stories about horses being electrocuted in the street. Utilities were rapacious monopolies and everybody knew it. There were serious, respectable arguments that wiring every home was dangerous, premature, and a scheme by power companies to bill the public for something nobody needed.

Now look at the adoption curve. In 1912, about 16 percent of American dwelling units had electric service. By 1920, about 35 percent. By 1930, 68 percent. By 1950, 94 percent. That is the Historical Statistics of the United States, not a promotional brochure. In one working lifetime, electricity went from a frightening novelty to something so assumed that we only notice it when it stops.

The places that wired early attracted the factories. The factories brought the jobs. The jobs built the tax base, and the tax base built the schools. Nobody today argues that Pennsylvania or Ohio should have paused electrification for a decade to study it. We know now how important that infrastructure became.

Lesson two: public concern about new infrastructure deserves a practical answer. The response to a dangerous, monopoly-prone technology was not a permanent prohibition. It was terms: rate regulation, universal service, and safety codes.

The digital age

In 1995, Pew Research found that 14 percent of American adults used the internet. Serious people dismissed it as a toy for hobbyists. News anchors stumbled over the word "browser" on live television. By 2000, 52 percent of adults were online. By 2010, 76 percent. By 2021, 93 percent. The internet went from curiosity to utility in about fifteen years, faster than electrification by a wide margin.

This period also had its infrastructure bust. Telecom companies borrowed and spent staggering sums laying fiber optic cable in the late 1990s, far more than the era's traffic could use. The bubble burst. Global Crossing and WorldCom collapsed in two of the largest bankruptcies in American history. The gamblers were wiped out so completely that the phrase "dark fiber" entered the language: cable in the ground with nobody lighting it.

And then? That dark fiber got lit. The overbuilt backbone the speculators left behind is a large part of why broadband got cheap and fast in the 2000s. The bust destroyed the balance sheets. It did not destroy the capacity. Once again: the infrastructure outlived the hype cycle, and the regions sitting on top of it collected the next twenty years of growth.

In March 2005, when Pew first measured it, 5 percent of American adults used social media. Five percent. It was a college kid's distraction. By 2021, 72 percent of Americans used it, and it had reorganized politics, advertising, news, and friendships. Facebook went from a dorm room to a company running gigawatt-scale data center campuses in Oregon, Iowa, and Texas in under two decades.

Nobody in 2005 wrote a zoning code for what social media would require. Nobody could have. The demand arrived first, the infrastructure followed, and the communities that landed the facilities collected the construction crews and the tax base. The social media age, like every age before it, ran on buildings full of servers. We just called them something friendlier.

The iPhone launched in 2007. The App Store opened in 2008 with 500 apps. In 2011, when Pew first measured it, 35 percent of American adults owned a smartphone. By 2021 it was 85 percent, and today it is 91 percent. The app economy, mobile banking, ridesharing, maps in every pocket: an entire layer of American life rebuilt itself on top of one device in a single decade.

Here is the part nobody thought about while they were double-tapping photos: every one of those apps lived in a building. The "cloud" is not a metaphor. It is a physical facility with servers, cooling, and a power feed. The mobile age felt weightless on the consumer end because all the weight was concentrated somewhere most people never looked: in data centers, quietly built in places like Northern Virginia, Hillsboro, and the Texas prairie.

Lesson three: every era's magic trick has a physical back end. The chatbot in your pocket is a building somewhere, drawing megawatts. The only choice a state gets is whether that building, and everything that comes with it, is inside its borders or outside them.

The AI age, in numbers

Now the current buildout. Strip away the hype and look at the measured data, because the measured data is more dramatic than any hype.

Lawrence Berkeley National Laboratory, which produces the federal government's definitive accounting of data center energy, reports that US data centers consumed 58 terawatt-hours in 2014. By 2023 that was 176 terawatt-hours, about 4.4 percent of all electricity used in the United States. The lab's 2024 report projected 325 to 580 terawatt-hours by 2028, somewhere between 6.7 and 12 percent of national electricity use. Its June 2026 update raised the reference case again: 649 terawatt-hours and roughly 11.8 percent of all US electricity by 2030, inside a scenario range of 521 to 843 terawatt-hours.

Globally, the International Energy Agency counts about 415 terawatt-hours of data center electricity consumption in 2024, roughly 1.5 percent of world electricity, growing at 12 percent a year for the last five years. Its base case doubles that to around 945 terawatt-hours by 2030, just under 3 percent of global electricity. The United States alone adds about 240 terawatt-hours by 2030 in that case, a 130 percent increase from 2024.

The IEA adds a number that should be on the wall of every governor's office. The United States already has the highest per-capita data center electricity consumption in the world, about 540 kilowatt-hours per person in 2024, and that figure is projected to more than double to over 1,200 kilowatt-hours by 2030, roughly a tenth of what an American household uses in a year. That is what "the AI age runs on buildings" means in practice. The load is coming. The only question is who hosts it, on what terms, and who gets paid.

Goldman Sachs Research forecasts global power demand from data centers rising 50 percent by 2027 and as much as 165 percent by 2030, compared with 2023.

McKinsey estimates $6.7 trillion in capital spending on AI infrastructure by 2030. Money that would qualify as the third largest economy in the world, and it is being allocated right now, county by county, state by state.

Money on that scale does not vanish when a legislature frowns at it. It moves. Capital is a river, and a moratorium is not a wall; it is a diversion channel pointed at your neighbor. That is an economic choice states can avoid by setting predictable terms.

Aspects to consider before imposing a moratorium

Let us brainstorm aspects of a state's economy which are impacted by a long-term moratorium.

First, the construction economy. A large campus is billions of dollars in construction spending and thousands of construction jobs over several years. Those wages get spent at the diner, the car dealer, the hardware store, and the little league sponsorship.

Second, the permanent tax base. A data center is one of the largest property taxpayers a county can land, and it arrives without demanding new lanes of traffic every rush hour. It is, from a county budget's point of view, close to the perfect taxpayer: enormous revenue, and modest service load.

Third, the utility economics. A large, steady, and creditworthy load lets a utility spread fixed grid costs across more kilowatt-hours, which can relieve pressure on everyone else's rates, if the deal is structured correctly.

Fourth, the option value. The state that becomes known as a place where compute gets built attracts the companies, the engineers, and the follow-on development that cluster around it. Ask Northern Virginia what three decades of that clustering is worth.

Now flip it and look at what the welcoming states are collecting in a full sprint to land this buildout, competing on speed, land, and power. The counties winning these projects are watching their tax bases transform: new substations, new transmission, new generation, construction payrolls measured in the billions, and permanent revenue that funds schools and hospitals long after the construction crews leave. A short pause to write strong rules can be worthwhile. But it should lead to a clear path for projects that meet them.

A long-term moratorium also does not stop residents from using AI tools. Demand can be served from elsewhere. When a project moves, the host state may lose associated construction work, tax revenue, and grid improvements. The question is whether the host community can secure those benefits without taking on unfair power, water, and financial costs.

The concerns driving the pause

Why do communities and their representatives ask for a pause? Data centers can be large, power-hungry neighbors. If a project arrives before the grid, water supply, and local benefits are understood, the costs can fall on people who never agreed to bear them. Asking for proof before granting permits is responsible. The choice ahead is what standards should allow the right projects to proceed.

The concern we hear is about the deal: will the developer take power, water, and public support while leaving the community with the transmission line and the bill? That concern calls for better terms, written into permits and utility agreements, with costs and benefits measured in public. A temporary pause is one answer. We think clear and enforceable standards can do more.

Every state can choose that path, regardless of leaning or region. Ratepayer protection, water use, and community benefits are the foundations of a stronger framework, not reasons to shut the door permanently.

The fair questions

Now let us also be clear about what a pause does, because dismissing the concerns is how the industry loses the argument. The concerns are real, and the industry has earned some of them.

Power. A 500-megawatt campus is a city-sized load landing on a regional grid. Who pays for the transmission upgrades, the substations, and the new generation? If the answer is "the average household's monthly bill," the deal is badly structured, full stop.

Water. Cooling takes water, and in water-stressed counties that is a genuine constraint that deserves hard limits, real measurement, and honest disclosure. Communities are right to demand the numbers before the concrete is poured.

Speculation. Some of what is getting proposed right now is not serious. When utilities build for phantom load that never materializes, ratepayers can get stuck with the cost.

Every one of these is a legitimate policy problem. Each needs a specific rule and a way to enforce it. The test is whether it produces a clear path for projects that meet them.

Define the conditions, don't stop altogether

Here is a playbook any state can adopt. It is not a free pass. It asks developers to prove how they will serve the communities that host them.

First: renewables before the grid. Require every large data center developer to build or fund new generation for the community before it draws a single megawatt off the shared grid. The project should arrive as a net contributor to local energy supply, not a drain on it. A developer who adds solar, wind, or hydro power to a county before taking power from that county has changed the entire politics of the project. Then the community can see and measure what the project contributes before accepting its load.

Second: pay your own way, in writing. The developer signs its electric service agreement, the ESA, under which the utility commits to serve the site at the regulated rate, and the developer posts real security for the grid upgrades its load requires. If the load never materializes, the security covers the stranded cost, not the ratepayers. This one clause kills the phantom-load problem dead. Speculators do not post security. Builders do.

Third: water honesty. Publish the projected draw. Commit to closed-loop cooling where the watershed requires it. Put the number in the permit where everyone can see it, and let the community hold the developer to it.

Fourth: screen ruthlessly. No land, no path to power, no capital behind the project, no permit. States should be openly hostile to speculative proposals, because every ghost project that collapses poisons the well for the builders who can actually deliver. Serious developers should welcome this scrutiny. It clears the field.

Do these four things and the whole frame flips. The data center stops being something done to a town and becomes something built with it. The state becomes a partner setting terms that communities and developers can rely on. And the capital, the jobs, and the tax base stay home. Good leadership protects the community and gives serious builders a clear path.

A word to data center developers

Before the industry enjoys this argument too much, let us aim it inward for a minute. The industry has work to do. If developers treat a community as an obstacle, withhold power and water numbers, or request capacity without a credible delivery plan, they should expect scrutiny.

States need a way to separate credible commitments from speculative ones.

Weak proposals and opaque costs make public trust harder to earn. If the industry wants terms instead of pauses, it has to act like a partner before it is forced to: publish the water numbers, post the security, build the community renewables first, and show up to the town hall in person. The developers who cannot operate that way are doing the damage, and we would rather states screen them out up front.

The verdict of history

Every age asks how to build the infrastructure people need while protecting the public. Railroads, electrification, and digitization came with real risks and real opportunities. We learned to set standards, correct mistakes, and keep building. We can apply that same approach to the AI age. Protect households. Make developers build new renewable supply before drawing from the shared grid. Measure water use. Screen out projects that cannot deliver. Then say yes to the ones that can. That is a practical offer to every state.

The buildout is happening. Let us decide the terms together.