The conversation about AI often begins with software. For investors, developers and business owners, one of its most consequential questions is increasingly physical: where will the electricity come from, and when will it be available?
In its October 2026 report, Powering Artificial Intelligence, the International Energy Agency estimates that the largest data centre operators are set to spend more than US$700 billion in capital expenditure during 2026, mostly on new data centres. That is an indication of the scale of ambition, rather than a guarantee that every proposed campus will be built.
The IEA also identifies constraints around network capacity, connection queues and permitting. Its analysis highlights how speculative and duplicate projects can inflate apparent demand in those queues. A substantial pipeline therefore needs careful interpretation.
The UK is changing how it approaches connections
The UK government's response on accelerating electricity network connections for strategic demand, updated on 28 October, sets out plans to prioritise strategically important demand projects and align data centre connections with regional infrastructure targets. Grid capacity, water, land and the needs of other electricity users feature in that approach.
These are policy measures and intentions. They should not be read as confirmation that a particular development has secured an earlier connection. Project-specific evidence remains essential.
What that means commercially
For a landowner, developer or capital partner, the opportunity is easier to assess when the delivery case is explicit. What capacity is contractually secured? What conditions attach to it? When can it be used? Which costs remain uncertain, and who will fund them?
An expansion ambition also needs to be distinguished from an existing entitlement. A credible route towards additional capacity may have commercial significance, but the work, expense and approvals still required should be visible in the investment case.
The same discipline applies to partnerships. If one party contributes land and another funds planning, power and development, the headline equity split provides only part of the picture. Funding obligations, future capital requirements, decision rights and milestones all influence how the arrangement will work in practice.
Let’s talk
Chris McDonald’s perspective
From my perspective, the useful question is how a project becomes deliverable. The opportunity around AI infrastructure is clear enough to deserve attention, but a compelling presentation still needs supporting evidence: power, permissions, a realistic timetable and partners able to meet their commitments. For founders and landowners considering a development partnership, I would start by making the assumptions visible. Understand what exists today, what remains to be secured and how the value created by each stage will be shared. That gives everyone a firmer basis for the conversation.