May 21, 2026

Washington just told the market, in effect, that AI-era data centres are now “infrastructure” in the same bucket as mines and megaprojects. The US Federal Permitting Improvement Steering Council has expanded FAST-41 coverage to include a QTS hyperscale build in Richmond (~3.2M sq ft) alongside Alaska’s Arctic copper project — a neat, slightly alarming pairing that captures the moment.

The Big Stories

US treats AI growth as industrial infrastructure problem — FAST-41 is being stretched to cover AI-linked developments, with a QTS Richmond hyperscale campus (~3.2M sq ft) pulled into the program and Alaska’s Arctic copper and critical minerals project also granted coverage. The connective tissue is supply chain and power: the piece flags copper constraints, transformer lead times of 18–24 months, and the pace of load growth (QTS reported nearly 4 GW; NERC projects a 224 GW summer peak rise). This matters because it’s a tacit admission that permitting and grid bottlenecks are now first-order constraints on AI build-out — and that governments may start “industrial policy-ing” the inputs (metals, transformers) as much as the campuses.

UK strikes historic multi-billion-pound trade deal with GCC — the UK signed a free trade agreement with the GCC (Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, UAE), projected (via modelling) to add £3.7bn to UK GDP per year in the long run and lift real wages by £1.9bn annually. It removes an estimated £580m in annual duties once fully implemented (£360m day one) and includes digital trade provisions like data free-flow commitments and ambitious customs timings. For data centre and cloud investors, the interesting bit isn’t tariffs; it’s the signal that cross-border digital services and data movement are being written directly into trade architecture between two capital-rich regions.

Ford Energy to supply up to 20 GWh of BESS to EDF — Ford Energy and EDF power solutions North America signed a framework for up to 20 GWh of DC Block battery storage over five years (up to 4 GWh annually), with deliveries expected to start in 2028. Ford is explicitly positioning this as a way to repurpose excess US EV battery manufacturing capacity after taking a $19.5bn charge tied to rationalising its US EV assets and roadmap. The subtext for the data centre world: grid constraints are turning storage into a mainstream procurement item, and non-traditional industrial players are looking for demand sinks big enough to matter.

Greenpeace exposes AWS ties to controversial fossil and surveillance firms — Greenpeace Germany alleges AWS supports fossil fuel, surveillance and autonomous weapons firms, claiming relationships with at least 38% of firms on exclusion lists and support for at least 100 “high-risk” entities. Protestors showed up at the Hamburg AWS Summit on May 20, and Greenpeace is pushing for mandatory ethical minimum standards for cloud providers (including service restrictions and independent oversight). Even if you discount the framing, this is a reminder that “cloud neutrality” is getting harder to defend politically — and that enterprise procurement, regulation, and activism are converging on the same question: what workloads should hyperscalers refuse?

Italy views Modi visit as new chapter for partnership — Italy elevated relations with India to a Special Strategic Partnership under a 2025–2029 action plan spanning defence, advanced manufacturing, energy transition, space and AI. For infrastructure watchers, the tangible nugget is the Blue & Raman submarine cable project, plus a reassessment of IMEC timelines due to West Asia instability. This is geopolitics bleeding into fibre routes and digital corridors: the physical map of connectivity is being shaped as much by security and diplomacy as by latency charts.

Behind the Headlines

Selector adds multi-cloud network visibility across hybrid infrastructure is the kind of “plumbing” story that quietly dictates operator efficiency. Its NetOps update aims to normalize telemetry via a “data hypervisor” and correlate an end-to-end view across branches, colocation, on-prem data centers, and public cloud — pulling from VPC flow logs, hyperscaler event streams, SNMP, and streaming telemetry. The important thread is that as estates sprawl across colocation and multiple clouds, teams are being forced into a single operational truth layer before they can credibly automate anything. Selector telegraphing foundational AI/ML models now, with a full generative AI release planned for the fall, is also a sign that vendors think the buyer conversation has moved from “visibility” to “operators want decisions and actions.”

EU-LISA report: AI’s potential for energy-efficient data centers lands in a very specific corner of the market: an operator that runs three on-prem data centers (main site in Strasbourg, plus sites in Austria and Estonia), has rack-capacity constraints, and must expand while complying with regulations that forbid third-party cloud use. The report’s stance — AI can optimize operations but isn’t mature enough to be a first-line solution — is telling because it’s coming from an organization that can’t just “burst to cloud” when constrained. This is what the next phase of efficiency looks like: not marketing claims about AI-run facilities, but incremental gains in cooling energy and PUE under tight regulatory and architectural constraints.

T5 Services names Mason Thornburg as CFO for growth is a people move with a scale signal embedded in it. T5 says it has surpassed $1.6bn in annual revenue and supports over 90 third-party facilities globally; Thornburg will oversee accounting, reporting, treasury, forecasting, and “investment discipline” across T5 Construction and T5 Operations. When services businesses get this big, CFO appointments aren’t just about clean books — they’re about how aggressively you can fund expansion, manage risk across a global delivery footprint, and decide which growth you don’t chase. For investors, it’s another datapoint that the picks-and-shovels side of the data centre boom is professionalising fast.