Summary

• National Grid Ventures is investing $1.75 billion for a 35% stake in Joulent.
• Joulent is developing contracted power infrastructure for large electricity users, including datacentres.
• The deal shows AI compute demand pulling utilities into higher risk commercial power infrastructure.

National Grid is putting $1.75 billion into Joulent, a US energy platform built around power infrastructure for large electricity users, as AI datacentre demand starts to reshape the investment logic of utilities.

The UK listed group’s commercial arm, National Grid Ventures, is taking a 35% stake in Joulent. The platform is developing contracted power and high voltage infrastructure for customers with heavy electricity needs, with its first major project understood to be a 2.67GW gas fired facility in West Texas linked to power supply for a Microsoft operated datacentre.

The deal is notable because National Grid is best known for regulated networks, where returns are shaped by public utility frameworks. Joulent gives it exposure to a more commercial model: power infrastructure built for datacentres and other large users under long term contracts. The potential returns are higher, but so are the risks around development, permitting, fuel, counterparty demand, and political scrutiny.

AI is changing the power conversation. Datacentres were already large electricity users before generative AI, but the latest wave of model training and inference has accelerated demand for dense, reliable compute capacity. That demand flows back through the economy into chips, cooling, land, grid connections, generation, transmission, and energy trading.

Utilities are now being asked to serve a digital infrastructure market that moves faster than traditional grid planning. Hyperscale cloud companies want large blocks of power in specific locations, often on tight timelines. National grids and local networks operate around long investment cycles, regulatory approvals, and physical constraints that cannot be wished away by software demand.

National Grid’s investment suggests that parts of the utility sector may respond by moving closer to the customer. Rather than only connecting large users to the grid, utilities and infrastructure investors can develop dedicated or contracted power assets, particularly where demand visibility comes from long term datacentre agreements.

That model raises uncomfortable questions. A gas fired project serving AI infrastructure may provide reliability and speed, but it complicates the climate narrative around digital growth. Cloud companies have made extensive renewable energy commitments, while the physical expansion of datacentres continues to create demand that grids may meet through a mixture of clean power, storage, transmission upgrades, and fossil generation.

The UK relevance is not diminished because the first project is in Texas. National Grid has said it is targeting more than 10GW of datacentre capacity across the UK and US over five years. The investment therefore indicates where the company sees future growth: not only in regulated wires, but in the infrastructure required to serve compute heavy customers.

The story also carries a warning for UK policy. The country wants more AI infrastructure, more datacentres, and more domestic compute capacity. Those ambitions will collide with grid connection delays, planning constraints, local opposition, and decarbonisation targets unless energy strategy and digital strategy are joined properly.

Enterprise AI buyers tend to see compute as a cloud procurement line. The power system sees it differently. Every AI workload has an infrastructure footprint somewhere, and that footprint becomes material when demand concentrates in hyperscale facilities. The cost and availability of power will increasingly affect where AI services are built, how expensive they are to run, and which countries can host them.

National Grid’s Joulent investment is therefore more than a financial bet on a US platform. It is a sign that AI demand is creating a new class of power customer large enough to influence utility capital allocation. As datacentres move from back office infrastructure to strategic industrial assets, the companies that own and connect electricity systems will become more central to the AI economy.