AI Infrastructure Brief: Power, Capital, and Silicon Collide in the Next Phase of the Data Center Buildout

The global race to build infrastructure for artificial intelligence is accelerating across North America, reshaping the economics, politics, and physical footprint of the data center industry.

New hyperscale campuses, multi-gigawatt development pipelines, power generation partnerships, and escalating policy debates illustrate how quickly the digital infrastructure landscape is evolving. At the same time, the industry is confronting rising scrutiny from communities and policymakers concerned about energy consumption and grid capacity.

From Memphis and Texas to Wisconsin and Northern Virginia, recent developments highlight a central reality: AI infrastructure is now scaling at industrial magnitude.

AI Cloud Providers Race to Deploy Blackwell Infrastructure

Several announcements this week underscore the enormous scale of compute infrastructure now being deployed to support next-generation AI workloads.

AI cloud provider CoreWeave said it plans to add roughly 5 gigawatts of additional data center capacity by 2030, while projecting that capital expenditures in 2026 could double as the company deploys NVIDIA’s next generation of GPU systems.

Much of that spending is expected to center on NVIDIA’s Blackwell architecture, particularly the GB200 NVL72 platform, which is designed to support massive AI training clusters. Industry analysts note that CoreWeave’s GPU capacity for 2026 is already largely sold out, reflecting the extraordinary demand for AI compute from hyperscalers, model developers, and enterprise customers.

Elsewhere, xAI, the artificial intelligence company founded by Elon Musk, filed permits totaling $659 million tied to expansion of its Memphis “Colossus” campus. The original facility, built in partnership with infrastructure providers including Supermicro, houses one of the largest GPU clusters in the world supporting the company’s Grok models.

At the infrastructure layer, hardware vendors are also adapting their architectures to AI scale.

NVIDIA and optical component manufacturer Coherent announced a strategic partnership to develop next-generation optical interconnect technology, addressing one of the biggest bottlenecks in AI cluster design: moving massive volumes of data between GPUs at ultra-low latency.

Meanwhile, Super Micro Computer continues pushing modular data center architectures designed to accelerate deployment of AI clusters by enabling faster assembly of compute infrastructure at scale.

AI Factories and the Rise of Gigawatt Campuses

The next generation of data center campuses is increasingly measured in hundreds of megawatts and, as you might've heard, even gigawatts.

A prominent example is Nebius, which refers to its facilities as “AI Factories.” The company’s planned campus in Independence, Missouri will serve as its flagship U.S. deployment and recently secured local approval for an extraordinary $150.6 billion Chapter 100 bond issuance, believed to be one of the largest economic development incentives ever approved at the municipal level.

Other projects across the country highlight the same trend toward hyperscale development.

In West Virginia, developer Penzance is targeting a 600-megawatt campus, reflecting growing interest in regions with available land and power infrastructure.

Prometheus Hyperscale has detailed plans for a major Dallas-area campus, positioning the site as a large-scale AI infrastructure hub within the rapidly expanding Texas data center market.

Texas itself is emerging as one of the epicenters of AI infrastructure growth. Google recently filed plans for a fifth data center building at its Midlothian campus, part of a much broader $40 billion investment in Texas through 2027 that includes additional campuses planned in Armstrong and Haskell Counties.

Hyperscale development continues across the Midwest as well. Meta Platforms recently signed a pre-development agreement for a campus in Beloit, Wisconsin, even as another Meta project in Beaver Dam has become a focal point for debate over power consumption. Regulatory filings indicate that the Beaver Dam site could require roughly 220 megawatts of electricity, a load large enough to rival that of many small cities.

Real Estate Capital Floods Into AI Infrastructure

The rapid expansion of AI infrastructure has attracted enormous interest from institutional investors.

Private equity giant Blackstone is reportedly exploring the creation of a publicly traded vehicle focused on acquiring AI data center assets. The proposed entity could raise tens of billions of dollars from retail investors, extending digital infrastructure investment beyond traditional pension funds and sovereign wealth investors.

In Northern Virginia, HighBrook Investors recently closed an inaugural data center fund targeting 300 megawatts of development capacity, reinforcing the region’s continuing status as the largest data center market in the world.

Other recent transactions include GI Partners acquiring two Baltimore-area facilities and Harrison Street selling two powered-shell sites in Maryland, reflecting the ongoing flow of capital into the sector.

Meanwhile, Canadian institutional investor La Caisse committed C$240 million to support construction of Cologix’s MTL8 facility in Montréal, an AI-ready data center designed to leverage Quebec’s hydroelectric power resources.

The Power Question Becomes Central

Behind almost every data center announcement today lies a deeper question: where will the electricity come from?

That challenge was highlighted by a major power infrastructure project involving Babcock & Wilcox, which received approval to develop a $2.4 billion natural-gas generation facility that will supply power to Applied Digital’s expanding AI campuses.

B&W has received full notice to proceed on a $2.4 billion design-build agreement with Base Electron, an independent power producer (IPP) backed by Applied Digital (NASDAQ: APLD) to deliver 1.2 GW of new generation capacity. The generation is intended to supply power to Applied Digital AI Factory campuses under separate power supply agreements. The project includes four 300-megawatt natural gas-fired boilers and steam turbine generator systems.

Base Electron, the IPP backed by Applied Digital, is focused on developing and owning generation assets that deliver new, dispatchable capacity to the grid and to contracted customers, including power supply agreements supporting Applied Digital’s high-density AI data center campuses.

Under its agreement with Base Electron, B&W will engineer, procure and construct the facility, with engineering and manufacturing activities already underway. Siemens Energy has been formally released to design and supply the steam turbine generator sets.

The project reflects the growing industry shift toward dedicated energy infrastructure.

Elsewhere, Atlantic Energy was selected to provide power for a 380,000-square-foot data center in Akron developed by Viking Data Centers, with plans for expansion to 150 megawatts.

In Canada, an agreement among TransAlta, Brookfield, and the Canada Pension Plan Investment Board outlines plans for a 230-megawatt data center project in Alberta, leveraging the province’s abundant energy resources.

These projects illustrate what many industry executives now describe as the “power island” trend: large data center campuses increasingly pairing with dedicated behind-the-meter generation rather than relying entirely on grid capacity.

In effect, the industry is beginning to shift from grid-dependent infrastructure toward hybrid or partially grid-independent models, a structural change that could reshape how AI campuses are designed and financed.

Hyperscale Expansion Continues

Major operators continue expanding their U.S. footprints.

NTT Global Data Centers recently secured four U.S. leases totaling approximately 115 megawatts, reflecting continued demand from hyperscale cloud providers and AI platform companies.

Amazon, meanwhile, has acquired a university campus in Virginia that could eventually be redeveloped into a data center site, another example of hyperscalers securing land for future expansion.

Industrial real estate giant Prologis has also revealed plans for a hyperscale development in Butler County near Cincinnati, Ohio, highlighting the growing role of logistics real estate developers in the digital infrastructure ecosystem. The San Francisco-based real estate development firm is planning the data center campus, called “Project Mila,” on 141 acres in the city of Trenton’s industrial park.

Prologis is recognized as the world's largest industrial real estate investment trust (REIT) and one of the largest, most valuable real estate companies globally, with over $200 billion in assets under management and 1.3 billion square feet of logistics space.

Construction Firms Ride the Data Center Boom

The surge in hyperscale development is also transforming the construction industry.

Turner Construction reported record revenue of $29.2 billion in 2025, with data center projects representing a significant driver of that growth.

The company is one of several engineering and construction firms benefiting from the industrialization of data center development, where projects increasingly resemble large energy or manufacturing facilities in scale.

Turner said its project backlog rose 34% year over year to $44.3 billion, the largest year-end backlog in the company’s history. Data center construction now accounts for roughly 37% of that pipeline, reflecting sustained investment in artificial intelligence infrastructure and digital capacity.

Political Scrutiny and Community Resistance Grow

As the industry expands, political scrutiny is intensifying.

In Birmingham, Alabama, the City Council approved a 180-day moratorium on new data center applications on March 3. The measure exempts facilities under 20 megawatts as well as projects with existing permits, including the Nebius site on Lakeshore Parkway.

In Oregon, lawmakers are considering revisions to House Bill 4084, which would temporarily suspend the ability of data centers to qualify for extended 10-year property tax exemptions in enterprise zones. While not a ban on data centers, the proposal reflects growing debate about the public incentives used to attract large projects.

Community opposition has also halted some projects. A proposed campus in Apex, North Carolina, was withdrawn after local resistance, while a West Louisville data center was approved despite significant community concerns.

Even the political arena is beginning to reflect the industry’s growing influence. Data center development has surfaced as an issue in North Carolina primary elections, where candidates have debated the role of digital infrastructure in local economic development.

The Bottom Line

Look past the individual announcements and a clearer pattern emerges: the architecture of the AI data center economy is changing.

Compute is scaling faster than traditional infrastructure models were designed to support. The rapid deployment of Blackwell-era GPU clusters is pushing facilities toward unprecedented density and power demand. At the same time, developers are pursuing hundreds-of-megawatts campuses as the new baseline, with gigawatt-scale planning increasingly common.

The capital markets are responding in kind. Institutional investors and infrastructure funds continue pouring billions into data center development, while new investment vehicles, such as Blackstone’s proposed public data center platform, suggest that AI infrastructure is beginning to look less like niche real estate and more like a new asset class.

But the real pivot point is energy. For decades, the data center industry largely followed the power grid. Today, the relationship is beginning to invert. Faced with transmission bottlenecks and multi-year interconnection queues, developers are increasingly designing projects around dedicated generation, hybrid grid connections, and behind-the-meter energy systems.

The emerging gigawatt era of digital infrastructure is a phase defined not simply by larger buildings, but by a fundamental rethinking of how compute, capital, and energy systems converge. In effect, the modern AI campus is evolving into something closer to a digital industrial complex: part data center, part power plant.

That shift is already visible in projects pairing hyperscale facilities with gas generation, renewable portfolios, or emerging nuclear technologies. It is also why political debates around energy policy, tax incentives, and land use are becoming inseparable from the data center conversation.

At Data Center Frontier, we talk the industry talk and walk the industry walk. In that spirit, DCF Staff members may occasionally use AI tools to assist with content. Elements of this article were created with help from OpenAI's GPT5.

Keep pace with the fast-moving world of data centers and cloud computing by connecting with Data Center Frontier on LinkedIn, following us on X/Twitter and Facebook, as well as on BlueSky, and signing up for our weekly newsletters using the form below.