S06.7 · Energy, Power & Climate
Regulated electricity and gas delivery and grid-enabling equipment, now creating new large-load tariff categories to handle AI datacenter demand.
Regulated utilities and grid infrastructure covers regulated electricity and gas delivery along with the equipment that enables the grid. Global grid investment reached $470B in 2025 (BloombergNEF), the first year past that threshold, and is growing roughly 10% a year on electrification, renewables interconnection and AI datacenter demand; global electric utility revenue is estimated at $3-3.5T (2025E, industry aggregates). In a regulated business the capex line is the earnings story, because rate base is how a utility grows. The structural novelty is that AI datacenter load is now large enough that utilities are creating entirely new large-load interconnection tariff categories that did not previously exist.
Regulated cost-of-service markets across most of the US, the EU and Asia earn an allowed return on rate base; deregulated markets — ERCOT, the UK, EU wholesale and Australia's NEM — earn wholesale and capacity-market prices instead. That is a structural difference in how a utility's revenue is built, not a matter of degree. Ownership is highly fragmented: thousands of utilities operate globally, with US investor-owned utilities dominating regulated rate base above a long tail of municipal and cooperative utilities.
Utilities buy power from generators and wholesale markets and source transformers, cable and switchgear from equipment suppliers. The model earns a regulated return — allowed ROE around 9-10% in the US — on a highly capital-intensive base with low volume risk in regulated markets. The constraints are physical and procedural: permitting, siting, and transformer and cable supply-chain bottlenecks are severe, and they gate the pace of build-out more than financing does. Rate base growth is the revenue growth mechanism — every dollar of prudently incurred grid investment ultimately becomes a dollar of allowed return — so the current wave of grid capex driven by electrification and datacenter interconnection is a direct earnings driver for regulated utilities, not simply a cost to be recovered.
Vertical integration into grid-equipment manufacturing answers supply-chain shortages; distributed-energy integration and EV-charging infrastructure open new rate-base categories, which in this business model is what expansion means.
Agentic AI automates outage prediction, load forecasting and interconnection-queue processing, collapsing the grid-planning engineering cost line. The moat is indifferent to all of it — franchise territory is regulatory, and AI does not touch it. The change worth flagging is tariff design. AI datacenter load is large and concentrated enough that utilities are building large-load interconnection tariff categories from scratch, with take-or-pay minimum-volume commitments and exit-fee provisions that had no precedent in prior tariff design. AI-driven load forecasting is already in use; automated interconnection processing is expected within 2-5 years.