May 14, 2026

A Texas county just hit pause on the data centre land-grab — and it won’t be the last. Hill County, Texas approved a one-year moratorium on new data centre and energy storage projects, citing water use, public safety, and “quality of life” as proposals outpaced local oversight. Pair that with grid experts warning that transmission timelines are measured in seven-year chunks while hyperscalers want interconnects in one to three, and you’ve got the makings of a real constraint: not demand, but permission.

The Big Stories

The Hill County moratorium is the clearest signal yet that local politics is becoming an operational risk, not just a PR problem. The county explicitly tied its freeze to water and safety concerns, and the story notes lawmakers in at least 14 states are considering similar restrictions — a fast-moving pattern as mega-projects (including those backed by OpenAI, Oracle, and Microsoft) collide with rural governance capacity. For investors, this isn’t “NIMBY noise”; it’s schedule risk and optionality risk, especially for greenfield campuses that assumed local permitting would be the easy part.

On the supply side of power, Fervo Energy priced its IPO at $27, valuing the company at roughly $7.7 billion after upsizing the offering. Fervo’s pitch is enhanced geothermal at meaningful scale: over 658MW in contracted offtake (including Google and utilities), with Cape Station targeting first power in 2026 and ~100MW by early 2027 on a path to 500MW. The market’s reception matters because data centre power conversations are drifting from “find more gas and solar” to “find firm, clean, financeable megawatts” — and geothermal is trying to graduate from concept to procurement.

Corporate buyers are still hoovering up clean electrons at a pace that makes policy rollbacks look like background noise. CEBA reports 27GW of corporate clean energy contracted in 2025 and estimates 17GW in Q1 2026, arguing 2026 could be the biggest year ever. CEBA also explicitly points to permitting and transmission reform (citing the SPEED Act), which is the tell: the bottleneck is shifting from “who will buy” to “can the system actually connect and deliver,” particularly as hyperscalers drive demand and plan up to $700bn in 2026 capex.

Grid timelines are now the industry’s most stubborn physics problem — and the numbers being thrown around are getting uglier. A study on PJM presented May 13 found that permitting and generator interconnection delays can add about $100 billion in system costs over 20 years in the most constrained scenario under high load growth tied to AI, data centres, electrification, and advanced manufacturing. That dovetails with a blunt message from regulators: AI-driven, gigawatt-scale data centres are increasing electricity demand faster than utilities can expand generation and transmission, with interconnection requests arriving on 1–3 year horizons while transmission takes ~7 years. Translation: even in markets with plenty of capital, the queue — and the paperwork behind it — is becoming the binding constraint.

India’s data centre momentum keeps widening beyond the usual cloud suspects. Uber is setting up its first data centre in India in partnership with Adani Group, expected to be operational later this year and framed as supporting Uber’s global technology ambitions. It also plugs into Adani’s much bigger narrative: a plan to invest $100bn in renewable-powered AI data centres by 2035, alongside other moves mentioned in the story (OpenAI with TCS, Amazon, Reliance). The key point isn’t the single facility — it’s that large, non-hyperscale platforms are now building sovereign-ish infrastructure footprints where power-and-land industrial groups can bundle sites, energy, and delivery muscle.

Behind the Headlines

Virginia is about to replay a familiar drama: transmission as the enabler of data centre growth, and the flashpoint for everything communities don’t like about it. The Piedmont Environmental Council and partners are teeing up a press conference over Valley Link Transmission’s proposed 115-mile, 765kV Joshua Falls–Yeat line across nine Virginia counties, which would require a 200-foot-wide corridor and clear over 2,600 acres. The groups’ argument explicitly links the line to data centre-related development — a reminder that “grid buildout” is not politically neutral when it’s perceived as subsidising one industry’s load growth. With route approval targeted for September 2026 and comment processes starting in fall 2026, this is the long runway version of the same constraint Hill County just acted on: social licence can be as slow as steel-in-the-ground.

South Africa is trying to do two things at once: widen access and make the digital economy investable — while also tightening the reins on AI policy. The government’s plan to expand “meaningful” connectivity includes R2.549 billion for 2026/2027 and support for extending REIT tax rules to digital fibre, towers and data centres. That REIT angle is a quietly big deal for capital formation: it’s a policy lever that can lower the cost of domestic funding for long-life digital infrastructure. At the same time, the minister withdrew the Draft National AI Policy after misuse of generative AI and appointed an independent expert review panel — a signal that regulation may arrive via risk management and governance, not a grand “sovereign AI” industrial plan.

Softbank’s latest move is a bet that AI data centres will pay for something other than lithium — and that Japan can manufacture it at scale. Softbank launched a zinc-bromine “water battery” manufacturing business, aiming for GWh-scale mass production around FY2028 and targeting annual revenue over JPY100bn by FY2030, with factories at Sharp’s Sakai site and partnerships with COSMOS LAB and DeltaX. The company is explicitly aiming at AI data centres above 150MW (scalable to >400MW), which tells you where it thinks the storage market is headed: long-duration and non-lithium chemistries positioned as safer, more supply-chain-resilient options for giant campuses. If this works, it’s not just a battery story — it’s a potential re-shaping of how hyperscale-grade sites think about behind-the-meter resilience and peak management.