November 2025 left a hard-to-ignore image: America is becoming the main battleground for the new digital infrastructure, and data centers—especially those designed for Artificial Intelligence—have transitioned from being a technical issue to an energy, industrial, and geopolitical matter. In just a few weeks, announcements of hundred-megawatt campuses, investment plans totaling tens of billions of dollars, and agreements connecting, for the first time so directly, the expansion of computing with the future of power generation occurred.
The pattern repeats: the demand for compute power to train and operate AI models is pushing tech companies, data center operators, and investment funds to secure land, electrical grids, backbone fiber, water—or alternatives to reduce consumption—and long-term energy contracts. The question is no longer whether there will be more data centers but where they will be built, what energy they will use, and who can sustain the pace.
United States: from the traditional “campus” to the AI superfactory
Among the most notable moves is the jump to projects conceived as “AI superfactories.” Microsoft unveiled in Atlanta a facility linked to its Fairwater initiative, designed to operate as part of a network of centers connected across states via dedicated links, functioning as a single distributed “factory” for model training. The company described a deployment aimed at massive loads, with compute and storage capacity at the exabyte scale, and an architecture designed to work in concert with other nodes in its network. The clear ambition is to build planet-scale infrastructure without concentrating everything in a single point.
In parallel, Anthropic announced a $50 billion investment to develop data centers in the U.S., partnering with Fluidstack, starting with locations in Texas and New York, with plans to add more sites later. Known for its family of Claude models, the company frames the decision around the need for “custom” infrastructure to support its own compute demands.
The investment fever also showed up in regional-scale projects. Amazon Web Services announced plans to invest $15 billion to develop new campuses in Indiana, a state positioning itself as a hub for digital infrastructure. Google revealed its intent to invest around $40 billion in data centers in Texas, reinforcing a concentration of investment in states with abundant industrial land and expanding energy ecosystems.
In the Midwest, Meta announced it would invest over $1 billion in an AI data center in Beaver Dam (Wisconsin), another example of how AI training and operational infrastructure are moving to locations that until recently were outside the main map of traditional hubs.
Virginia consolidates as a “corridor” for infrastructure, but scale alters the rules
If there’s one region summarizing the rise of data centers in the U.S., it’s Virginia, where the ecosystem of networks, energy, and operators has been maturing for years. The difference now is the scale.
Vantage Data Centers announced a 192 MW campus near Fredericksburg (Stafford County), with a reported investment of $2 billion. This figure alone places the project among large-scale deployments of critical infrastructure.
Following the same “megacampus” logic, CleanArc Data Centers launched its first major project in Caroline County (Virginia): a 900 MW campus, with an initial 300 MW phase scheduled to enter service in Q1 2027, with phased expansions planned through the next decade. The announced power capacity positions it equally as energy infrastructure and digital infrastructure.
Additionally, Powhatan (Virginia) approved a campus with a initial 365 MW, supported by Dominion Energy and managed by JLL. The key point isn’t just the capacity but the implicit message: data center deployments are now being negotiated and planned with energy players and real estate managers as core partners, not peripheral providers.
Energy: the major constraint driving nuclear deals and mega-electric projects
November also conveyed a deep message: AI is forcing a direct confrontation with energy issues. Westinghouse—an advocate for accelerating the construction of up to 10 AP1000 reactors in the U.S. by 2030—and Google Cloud announced a partnership to apply digital tools and AI to nuclear construction and operation, promising to reduce friction in a sector where timelines and complexity often incur the highest costs.
The pressure isn’t theoretical. In North Carolina, Energy Storage Solutions (E22) announced a 900 MW campus valued at $19.2 billion in the Kingsboro area (Edgecombe County), with work expected to start in 2026. The project’s magnitude sparked local debate: regional media highlighted that the anticipated power consumption would be equivalent to the total output of the state’s only nuclear plant, illustrating why data centers are now at the heart of public discussions.
In South America, energy once again surfaced as a structural element: Guyana agreed with Cerebras to build a 100 MW AI data center powered by natural gas, linked to the Wales Gas-to-Energy complex. The country seeks to turn its new energy landscape into a driver of technological development.
Latin America: Mexico, Brazil, Chile, and Colombia enter a phase of big bets
Parallel to the U.S. push, several announcements reinforced Latin America’s positioning as a growth region, especially around cloud and AI.
Actis launched the TERRANOVA platform, with a plan to invest $1.5 billion in developing data center campuses in Mexico, Brazil, and Chile. The first project will be based in Querétaro and aims to start development in early 2026, with subsequent plans for Brazil and Chile.
In Mexico, a project was announced to install Latin America’s first High-Performance Computing and AI Center in Nuevo León, with an initial investment of $1 billion, led by AI-GDC and Cipre Holding. However, the project drew market attention: NVIDIA denied committing any direct financial investment, though the project promotes support through its technology.
Brazil had one of the most high-profile movements: the data center project linked to TikTok at the Pecém port complex (Ceará) advanced with the addition of Omnia, backed by Patria Investimentos. Reuters estimated the project at R$50 billion (about $9.25 billion), with 300 MW of planned consumption and operations scheduled for 2027 in partnership with Casa dos Ventos.
Chile, for its part, greenlit the environmental approval for the Data Hall Quilicura by Ascenty, a $140 million project scheduled to start construction in January 2026. In Colombia, Claro completed the modernization of the Triara Megacenter in Medellín, investing over $13 million to bolster cloud, AI, and cybersecurity capabilities.
Regionally, NextStream announced the creation of an integrated management center to operate in real time 10 data centers across five Latin American countries, a key step toward professionalizing and centralizing large-scale operations.
Other notable announcements during this period
Beyond the main list, the industry radar captured other relevant movements during the same period, reported by specialized media: from the end of AWS’s Project Rainier to new plans in Wisconsin for massive AI capacity, signaling that the data center landscape in the U.S. continues expanding beyond traditional hubs.
Frequently Asked Questions
What does it mean for a data center campus to have 900 MW, and why is this such an important figure?
It indicates the project’s scale competes with large energy infrastructures and requires planning around power grids, substations, and supply contracts as core components of development.
Why are nuclear reactors and natural gas mentioned when announcing AI data centers?
Because training AI models demands continuous, massive electricity; ensuring reliable generation—nuclear or gas—is becoming a strategic advantage to attract investment.
What distinguishes an “AI data center” from a traditional cloud data center?
Typically designed for high computational density (many GPUs), faster internal networks, and advanced cooling, with the goal of training models and operating inference at scale.
Which regions in the Americas are attracting most data center investment, and why?
States like Virginia, Texas, and Indiana lead due to their available land, connectivity, and energy capacity; in Latin America, hubs such as Querétaro, Nuevo León, and Pecém are favored for their industrial strategy and connectivity.