The projects, markets, power constraints and network infrastructure behind the next AI buildout in the U.S. and Latin America
The United States is experiencing one of the largest infrastructure buildouts of our generation.
It is easy to describe this simply as a data center boom. I think that understates what is happening.
AI is forcing the convergence of four industries that historically operated somewhat independently: data centers, energy, telecommunications and computing infrastructure.
The result is a new generation of projects measured not only in megawatts, but increasingly in gigawatts and tens of billions of dollars.
For people working across the data center ecosystem, the important question is no longer whether demand will continue. The more interesting questions are: Where will the next generation of capacity be built? Who is building it? Who has access to enough power? And what infrastructure will be required to connect these new AI campuses to the rest of the world?
I spent time researching the major U.S. data center developments currently under construction or moving through the development pipeline. The scale surprised even me.
From Data Centers to AI Infrastructure Campuses
For most of the history of the data center industry, a 20 MW or 50 MW facility was significant. Then 100 MW campuses became increasingly normal. The newest generation of projects is operating on an entirely different scale.
Among the major developments moving forward in the United States are projects associated with Meta, Microsoft, Amazon Web Services, Google, OpenAI, Oracle, Vantage Data Centers, QTS, Crusoe, STACK Infrastructure, Digital Realty and others.
Several individual campuses are approaching or exceeding 1 GW of planned capacity. A gigawatt-scale data center campus is no longer simply a real-estate development. It starts looking like a combination of a power plant, industrial complex, telecommunications hub and computing factory.
Some of the Largest Projects to Watch
One of the most significant developments is Meta’s massive expansion in Richland Parish, Louisiana. Meta has since increased its committed investment there to more than $50 billion, with the campus now expected to deliver over 2 GW of compute capacity, making it the largest single project referenced in this analysis. Turner Construction, DPR Construction and Mortenson are the general contractors publicly associated with the development.
Meta is simultaneously developing major infrastructure in markets including Indiana, Wisconsin, Texas and Ohio. In Lebanon, Indiana, Meta is developing an approximately 1 GW AI-oriented campus representing roughly $10 billion of investment.
Wisconsin is emerging as another important AI infrastructure cluster. Microsoft continues expanding its Mount Pleasant development while Vantage Data Centers’ Project Lighthouse near Port Washington represents approximately 902 MW of planned capacity and roughly $15 billion of investment. Meta is also developing its Beaver Dam campus.
Three major operators developing infrastructure in one state tells us something important: Wisconsin is becoming a serious hyperscale and AI infrastructure market.
Texas May Become America’s AI Infrastructure Capital
Texas deserves special attention. The Abilene area has become one of the most interesting AI infrastructure markets in the world. Crusoe’s development there forms an important part of the infrastructure supporting the OpenAI and Oracle ecosystem.
Additional expansion could push the broader Abilene campus toward approximately 2 GW of capacity. At the same time, Vantage is developing major Texas infrastructure, Meta is expanding in El Paso, QTS and Lancium are developing additional capacity, and AWS is pursuing another massive campus in Wharton County.
Texas has several advantages that are difficult to reproduce elsewhere: large amounts of land, an enormous energy industry, renewable generation, natural gas infrastructure and a business environment capable of supporting very large industrial developments.
But Texas also illustrates the biggest constraint facing the entire industry. It isn’t land. It isn’t buildings. And increasingly, it isn’t capital. It is power.
Power Is Becoming the New Location Strategy
For years, the three most important considerations in data center development were often summarized as: Power. Fiber. Land.
Today I would change the emphasis: POWER. Power infrastructure. Fiber. Then land.
The electricity requirements of AI infrastructure are forcing developers to think very differently about energy. Waiting years for conventional utility interconnection is simply incompatible with the speed at which AI companies want to deploy compute.
This is why we are increasingly seeing data centers developed alongside dedicated natural-gas generation, utility-scale substations, battery storage, renewable generation, private transmission infrastructure, nuclear power agreements and behind-the-meter generation.
In other words, the boundary between data center developer and energy developer is disappearing. I believe this will be one of the defining changes in our industry over the next decade.
America’s Data Center Map Is Changing
Northern Virginia will remain one of the world’s most important data center markets. But the AI boom is changing the geography of American digital infrastructure.
Capital is increasingly moving toward Texas, Wisconsin, Louisiana, Indiana, Ohio, Iowa, Pennsylvania, Mississippi and the Carolinas. Even within Virginia, development continues moving farther from the traditional Ashburn core as developers search for available power.
This geographic decentralization will have consequences far beyond real estate.
The Next Data Center Boom Will Also Be a Fiber Boom
There is an important second-order effect that deserves more discussion. When hundreds of megawatts of compute appear in a market that historically had relatively little hyperscale infrastructure, that compute must communicate with the rest of the world.
AI campuses require enormous connectivity. They need diverse long-haul fiber, cloud connectivity, high-capacity IP networks, DDoS protection, multiple physical entrances and connections between training clusters, inference infrastructure, cloud regions and major Internet exchanges.
Eventually, some of these new data center markets may develop their own interconnection ecosystems. This means the AI data center boom will inevitably trigger another wave of fiber and network infrastructure investment across the United States.
The physical Internet map will evolve alongside the data center map.
The Companies Actually Building the AI Boom
Hyperscalers may announce the projects, but an enormous ecosystem has to actually build them.
Construction organizations including Turner Construction, Mortenson, DPR Construction, Holder Construction, McCarthy Building Companies, JE Dunn, Hensel Phelps, The Weitz Company and Whiting-Turner are becoming critical participants in the AI infrastructure economy.
But even these companies depend on another layer of infrastructure providers: electrical contractors, transformer manufacturers, switchgear suppliers, cooling companies, generator manufacturers, engineering firms and skilled mission-critical labor.
We talk constantly about GPU shortages. The next shortage may be everything required to put those GPUs into production: transformers, generators, switchgear, cooling equipment, substations, electricians, engineers and experienced mission-critical construction teams.
Announced Megawatts Are Not Real Megawatts
There is also a reason for caution. The industry increasingly talks about enormous numbers: 500 MW, 1 GW, 2 GW and sometimes considerably more.
But announced capacity should never be confused with energized capacity. A developer may control enough land for 2 GW while initially building 100 MW. Another project may have land and financing but still be waiting for utility infrastructure. Others may have power agreements but face permitting challenges.
The most useful way to evaluate the U.S. data center pipeline is therefore not simply, “How many gigawatts have been announced?” It is: How many megawatts have land, permits, financing, power, equipment, contractors and customers—and when will those megawatts actually become operational?
That distinction will become increasingly important as the market grows.
Latin America Is Becoming a Core Part of the AI Infrastructure Map
The U.S. is the largest part of the current AI infrastructure cycle, but it would be a mistake to view this as a U.S.-only story. Latin America is entering its own accelerated data center investment cycle.
CBRE reported that Latin America led global data center inventory growth in early 2026, with 41.3% year-over-year growth across the markets it tracks. The region is still smaller than North America in absolute capacity, but its growth rate and development pipeline are increasingly difficult to ignore.
Brazil remains the regional center of gravity. São Paulo is Latin America’s deepest hyperscale and colocation market, while new AI-oriented investment is expanding the map beyond the traditional core. In May 2026, Ascenty announced US$1.2 billion of investment tied to 150 MW of new AI capacity and four new data centers in the São Paulo region.
Scala Data Centers says it has more than 100 MW of operational IT capacity across Latin America and more than 900 MW in projects under construction or development. Its published pipeline spans Brazil, Mexico, Chile and Colombia, including large-scale expansion around São Paulo, Querétaro, Santiago and Bogotá.
Mexico is particularly important. Querétaro has developed into one of the region’s strongest hyperscale clusters because of its proximity to Mexico City, cloud demand and growing concentration of data center infrastructure. The challenge is familiar: securing power and transmission capacity fast enough to match demand.
Chile has built a strong cloud and data center ecosystem around Santiago, supported by renewable-energy potential and international connectivity. Colombia is earlier in the cycle, but Bogotá is attracting new development and offers an important gateway into the northern Andean market.
The regional opportunity is therefore becoming more distributed: Brazil provides scale; Mexico provides proximity to the U.S. and a large domestic economy; Chile combines cloud demand with energy advantages; and Colombia offers an emerging connectivity and enterprise hub.
Latin America’s Next Constraint Will Also Be Power
The same lesson visible in the United States is beginning to apply across Latin America: the data center market can grow only as fast as deliverable power allows.
Latin America has a potentially powerful advantage. Several markets have access to large renewable-energy resources, including hydro, wind and solar. But abundant generation at a national level does not automatically mean a data center can obtain hundreds of megawatts at the exact site and on the exact schedule it needs.
Transmission, substations, permitting, grid reliability and the distance between generation and demand can become the real bottlenecks. The winners will increasingly be developers that secure energy early rather than simply securing land.
Brazil’s August 2026 announcement of roughly US$444 million in public investment for new AI supercomputing infrastructure is another signal that compute sovereignty is becoming a strategic issue in the region, not merely a commercial cloud decision.
Miami and the Connectivity Between the Two Continents
There is another reason the U.S. and Latin American data center stories should be analyzed together: the networks connecting them.
Miami and South Florida remain a natural interconnection bridge between North America, Latin America and the Caribbean. At the same time, submarine cable systems landing in Brazil, Colombia, Chile, Mexico and the Caribbean continue to diversify the routes through which regional traffic reaches the United States and the rest of the world.
As AI infrastructure expands in both directions, the strategic value of diverse international fiber, subsea capacity, regional Internet exchanges, cloud interconnection and high-capacity IP networks should rise with it.
The result may be a more integrated digital-infrastructure market across the Americas. Compute will increasingly be placed where power, economics, latency, regulation and data-sovereignty requirements make sense—and networks will have to connect those islands of compute as if they were one system.
The Bigger Picture
I have spent much of my career around Internet infrastructure, including the networks connecting North and South America. What is happening now feels different from previous infrastructure cycles.
Cloud computing created enormous centralized data centers. Streaming created enormous bandwidth requirements. Mobile connected billions of additional devices. But AI is simultaneously increasing demand for compute, electricity, cooling, fiber and capital.
That combination makes this cycle unique.
We may eventually stop thinking about these projects as data centers altogether. They are becoming AI infrastructure factories.
And the locations capable of supplying abundant electricity, diverse fiber, land, skilled construction and favorable regulation could become some of the most strategically important infrastructure markets in America.
The race to build AI models gets most of the headlines. Behind that race is another one that may ultimately be just as important: the race to build the physical infrastructure that makes AI possible.
Ready to deploy edge infrastructure in Latin America? Contact EdgeUno today
