May 27, 2026

Applied Digital just put a very large flag in the ground in Louisiana: a $3.6bn, 300MW (critical IT load) AI campus on 300 acres in Rapides Parish, targeting first operations in mid‑2027. That kind of single-site power ask is exactly why regulators and governors are starting to treat data centres less like “real estate” and more like “industrial load” that can swing a grid plan. Today’s other big reads—North Carolina moving to make hyperscalers post real commitments, and West Virginia pitching a 50GW-by-2050 expansion—show the new reality: power politics is becoming the fastest-moving part of the data centre market.

The Big Stories

Applied Digital to build $3.6B AI campus in Rapides Parish is the day’s headline because it’s concrete: Delta Forge 1 is planned as a 300MW AI campus, with closed-loop cooling, served by Cleco, and structured to qualify for state and local sales-and-use tax exemptions. Applied Digital is also putting timelines and jobs numbers on it (200 direct roles; 1,000+ peak construction jobs), which is a subtle signal to local stakeholders: this isn’t a “maybe” load, it’s a civic-scale project. For investors, the question to watch isn’t whether 300MW is technically doable—it’s how fast the surrounding power and interconnection machinery can keep up when multiple projects like this land at once.

North Carolina advances bill to limit hyperscale data center costs is what “keeping up” looks like when policymakers decide they’re tired of socialising risk. SB 730 would force hyperscalers to internalise more of the infrastructure bet via long-term contracts, minimum billing, and financial guarantees, and it would ban evaporative/open-loop cooling for facilities over 100MW. The numbers from Duke Energy are the tell: 7.6GW signed with data centres (including 2.7GW added in Q1 2026) and another 15.4GW under discussion—on a retail power backdrop of roughly $0.16/kWh for residential customers (March 2026). This is the beginning of a policy template: if your load is big enough to move the utility capex curve, expect tougher terms.

West Virginia’s 50 by 50 plan aims energy expansion reads like a state trying to pre-empt the “we can’t serve you” problem before it hits. The proposal targets a jump from about 15GW of capacity today to 50GW by 2050, including $1.44bn to refurbish coal-fired plants and efforts around restoring the Pleasants Power Station, alongside transmission and gas upgrades. The inclusion of 7,250 in-home fuel cells over three years is the interesting wrinkle: it suggests policymakers are willing to experiment with distributed solutions as a reliability hedge while big grid builds grind through timelines. Data centres are explicitly in the demand story here—West Virginia wants to be a place that says “yes” when other markets say “not now.”

Gorilla secures 5.5MW (scaling to 18MW) capacity with NeutraDC is smaller in MW, but it’s a clean example of how AI demand is being packaged and sold to investors. Gorilla has an initial ~5.5MW agreement expected to scale to ~18MW by November 2026, tied to deploying GPU infrastructure for AI and enterprise workloads, and it’s talking about a US$2bn+ five-year revenue opportunity at full utilisation. The bigger ambition is the real signal: Gorilla says it wants to aggregate 150MW–200MW across Asia over the next year via partnerships with regional operators. Read this as the “capacity aggregator” model maturing: not everyone building AI compute wants to own the dirt, the shell, and the utility relationship.

BDUK awards CityFibre Project Gigabit contract for Hampshire looks like a broadband story, but it matters for data centre economics because latency, resilience, and metro connectivity are the other half of “AI readiness.” The Hampshire Project Gigabit contract was redesigned in May 2026 from about 75,500 premises (£104.2m) down to around 29,000 premises (£71.7m), with the redesign framed as changing what public money is paying for. Stripped down, this is the public sector acknowledging overlap and reprioritising where it subsidises fibre. For operators and investors, that’s a reminder: government-backed network expansion is getting more targeted—and the easy assumption that “public money fills the gaps” is being stress-tested.

Behind the Headlines

Advanced power electronics push data centers from diesel to batteries is a quiet but profound shift in how AI-scale sites may be designed—and financed. Grid-forming inverters, AI-capable UPS, 800 VDC architectures, and tools like Universal Damping STATCOMs are being pitched as ways to shrink diesel-generator runtime without giving up uptime, with a roster of vendors and names (ON.energy’s 3.5MW AI UPS; Dimaag.ai’s 800 VDC Zenius under ERCOT review; plus Eaton, Vertiv, and even Microsoft in the conversation). The important subtext: as regulators and communities push back on emissions and noise, “diesel as default” becomes a reputational and permitting liability. If battery-forward backup becomes bankable at scale, it changes both the capex mix and the compliance posture of new campuses.

Why local data infrastructure matters for Kenya’s digital future is a reminder that “AI demand” isn’t just a hyperscaler land-grab in the US. Liquid Intelligent Technologies argues Kenya needs stronger local data centres, power, cooling, and secure connectivity to meet rising cloud/analytics/AI demand, pointing to Africa hosting only ~0.6% of global data centre capacity. The drivers listed—data protection rules, performance, cybersecurity, and SME/startup enablement—map to a simple thesis: sovereignty and user experience are becoming intertwined. Investors should read this as early positioning for the next wave of metro and regional builds, where the constraint isn’t just capital—it’s reliable power and the ecosystem around it.

Pope’s encyclical warns AI risks to humanity and society lands in the inbox because it’s a cultural signal with policy legs. “Magnifica Humanitas” explicitly calls out the concentration of power in AI/data systems and highlights the environmental costs—energy, water, carbon emissions, and data-centre infrastructure—while urging regulation and even to “disarm AI.” Whether you agree or not, the point is that data centres are moving from “technical infrastructure” into mainstream ethical and environmental debate, with influential institutions naming the industry directly. That matters because permitting, cooling standards, and disclosure regimes tend to follow attention—and attention is clearly broadening.