Nscale Eyes 8GW Data Center Acquisition in West Virginia
Nscale is set to acquire an 8GW data center campus in West Virginia, a game-changer for AI infrastructure. #DataCenters #Nscale
A UK-based neocloud is in talks to acquire what could become the largest AI data center campus in the United States. If the deal closes, it won't just be a milestone for Nscale — it will signal a fundamental shift in how serious European cloud players are about competing on American soil.
According to reporting by *The Information*, Nscale is in active negotiations to acquire American Intelligence Power (AIP), the entity that controls a massive undeveloped site in Mason County, West Virginia. The property, branded as Monarch Compute Campus, carries permits and power agreements already in place for up to 8GW of total capacity — a number so large it demands context. For reference, the entire US data center market consumed roughly 17-20GW of power as recently as 2023. An 8GW campus, fully built out, would represent a meaningful fraction of that figure in a single location.
This isn't a done deal. But the parties at the table — and what the site already has secured — make it worth paying close attention to.
What Nscale Would Actually Be Buying
The headline number is 8GW, but the more important details are buried beneath it.
AIP was founded at the start of 2026 as a joint venture between Fidelis New Energy and 8090 Industries. Fidelis originally unveiled plans for the Mason County site back in 2023, when the vision was considerably more modest: a 1GW campus powered by an adjacent hydrogen facility. That concept has since been dramatically scaled up — and the power source has shifted to fast-response natural gas generation, with Caterpillar already contracted to supply generator sets for the first 2GW phase.
That first 2GW phase is expected to come online by 2027, which is an aggressive but not implausible timeline given that permits are reportedly already cleared.
What Nscale would acquire under this deal isn't just land. It would absorb AIP entirely — meaning it inherits the permits, the power agreements, and the Caterpillar relationship. For a company looking to scale quickly in North America, buying regulatory certainty is often worth as much as buying the dirt itself. Greenfield data center development in the US can take years just to clear interconnection queues and secure grid agreements. This site has already navigated those bottlenecks.
The catch: the land is undeveloped. Nscale would still need to fund and execute the actual construction of data centers, the electrical infrastructure buildout, and the fit-out of facilities. At this scale, that's a multi-billion-dollar capital program — on top of whatever the acquisition itself costs.
Why Mason County, West Virginia
West Virginia doesn't often show up in data center site selection conversations. That's part of why this site matters.
The state has been quietly positioning itself as a viable alternative to the saturated markets of Northern Virginia, Phoenix, and Dallas. Land costs are a fraction of what you'd pay in Loudoun County. Labor is available. And critically, West Virginia sits on substantial natural gas infrastructure — relevant context for why a natural gas generation strategy makes sense here rather than, say, queuing for renewable interconnection in an overloaded PJM market.
Mason County specifically has additional geographic advantages. It's within reasonable distance of major fiber routes and East Coast demand centers, without the premium cost basis of being inside them. For hyperscale and AI workloads that don't require sub-millisecond proximity to end users, that tradeoff is increasingly attractive.
The fact that Amazon, Meta, and Fluidstack are also reportedly interested in this property tells you everything about its strategic value — three very different types of buyers don't converge on the same undeveloped land by accident.
The original hydrogen-powered concept Fidelis pitched in 2023 was ahead of its time in some ways and impractical in others. Pivoting to natural gas generation for the near-term phases is a pragmatic call. It gets the first 2GW online faster, using proven equipment from a supplier — Caterpillar — with an established track record in mission-critical power applications.
What Nscale's Position in This Deal Actually Means
Nscale completed a $2 billion Series C funding round recently, explicitly earmarked for accelerating expansion across Europe, North America, and Asia. The company currently operates data centers in England, two locations in Norway, and Texas. It's a genuine infrastructure operator — not a paper company — but its US footprint is thin relative to its ambitions.
Acquiring Monarch Compute Campus would be a transformational bet. It would vault Nscale from a mid-sized European neocloud into a credible large-scale US AI infrastructure player almost overnight — at least on paper.
The insider reality is more nuanced. Operating a 2GW-and-growing campus in rural West Virginia is a fundamentally different challenge than running colocation facilities in Norway and a Texas data center. Supply chain logistics, staffing at scale, and managing hyperscale customer relationships all get more complex at this level. Nscale has capital. Whether it has the operational depth for this scope of buildout is the more interesting question.
The reported IPO preparation — Goldman Sachs and JPMorgan were reportedly hired in February 2026 to help prepare for a listing — adds another layer of strategic logic to this deal. A credible 8GW development pipeline in the US would be a compelling asset on an IPO roadmap. It anchors a growth narrative that purely European infrastructure can't provide.
The AI Demand Equation Behind All of This
None of this makes sense without understanding the demand signal driving it. AI model training and inference are extraordinarily power-hungry. A single large GPU cluster for frontier model training can consume 50-100MW on its own. Hyperscalers and AI-native companies are scrambling to secure capacity years in advance, which is why the pipeline of announced data center projects in the US has exploded.
The 8GW target at Monarch isn't a number someone invented for marketing purposes — it reflects a genuine thesis about where AI infrastructure demand is heading over the next decade. The question isn't whether demand will fill 8GW eventually. The question is who builds it, how fast, and whether they can finance the capital stack required to get there.
For Nscale, winning this site would mean competing directly for AI infrastructure contracts alongside the hyperscalers — not just offering GPU cloud services, but providing the underlying physical capacity that AI workloads run on. That's a different business model and a different risk profile, but the margin potential at scale is substantial.
What Comes Next
The deal value hasn't been disclosed. Competing interest from Amazon, Meta, and Fluidstack means Nscale isn't negotiating in a vacuum. And the undeveloped nature of the site means any acquirer is committing to a long-duration capital program before meaningful revenue flows.
Watch for a few things: whether Nscale's reported IPO timeline accelerates or stalls depending on whether this deal closes, how the first 2GW phase financing is structured (project finance versus balance sheet), and whether hyperscalers end up as anchor tenants rather than competing buyers — a scenario that would de-risk the project considerably.
The broader signal here is that the race to secure AI-ready land with permits and power agreements already in place has moved past the point of incremental plays. Eight gigawatts in a single campus isn't a data center project — it's infrastructure at the scale of a small power grid. Whoever ends up with the keys to Mason County will be shaping AI infrastructure capacity in the eastern United States for the better part of the next decade.
Explore more about the InfraSale Marketplace here!