What matters. Why it matters.
- The IEA’s 2026 outlook puts global data-centre electricity use at 485 TWh in 2025 and projects 950 TWh in 2030.
- Ofgem is examining how to clear speculative demand from Britain’s connections queue. A queue position is not usable power.
- The commercial checkpoint is a credible connection, funded equipment and a paying customer—not an industry growth forecast.
The signal in the updated numbers
AI’s infrastructure story is moving from the size of a market to the ability to deliver a service. The International Energy Agency’s April 2026 assessment estimates that data centres consumed 485 TWh of electricity in 2025 and projects 950 TWh by 2030. It also describes supply-chain bottlenecks that constrain the more aggressive near-term paths. The figures update the 2025 report discussed in our founding edition; they should not be combined into a single series without checking the revised assumptions.[1]
Britain’s queue is not a demand forecast
Ofgem’s July statement says demand connection applications rose from 41 GW to 125 GW in under a year, with data centres accounting for at least 80 GW. Those are applications for capacity, not operating facilities or guaranteed future consumption. The regulator proposed commitment fees and project milestones to address speculation. Its consultation page records a 16 September closing date. That establishes the consultation’s status, not the subsequent adoption of every proposal.[2][3]
A larger electricity system behind the screen
TWh, global data centres
Where the commercial question becomes interesting
Our interpretation is that delivery evidence deserves more weight than headline exposure to AI. Imagine two otherwise similar proposed facilities. One has an agreed route to energisation, costed equipment and customers whose contracts depend on a workable commissioning date. The other has a large proposed capacity figure and a broad market forecast. They may appear in the same industry pipeline, but the documents needed to evaluate them are different. This is a framework for examining projects, not a claim about the value of a particular listed company.
Follow the chain from equipment to cash
A connection can still be insufficient if the rest of a project is not ready. A practical review links each delivery stage to the next: power availability, equipment installation, commissioning, customer acceptance, billing and cash collection. A supplier’s order announcement belongs at a different point in that chain from collected revenue. The useful questions are which obligations remain conditional, who bears a delay and whether the funding plan covers the period before payment begins. These are analytical questions for the relevant contracts and filings; the sector-level documents do not answer them for individual businesses.

The counter-case matters just as much
A constrained input can attract investment, but it does not guarantee attractive returns for every supplier. If supply expands, customer plans change or financing becomes more expensive, the economics can move even while aggregate demand grows. Better equipment efficiency can also alter the amount of infrastructure required for a unit of useful computing. The mistake would be to turn a physical bottleneck into an automatic stock-market conclusion. A good case needs to explain the path to cash and the conditions that could interrupt it.
A compact diligence file
For a repeatable review, build a one-page record for each project. Put the proposed capacity, the status of the connection, the expected customer start date and the funding assumptions side by side. Mark an item as unknown when the supporting document is unavailable. That is more informative than filling the gap with a sector average. The same discipline works across countries: change the regulator and the network documents, while keeping the delivery questions consistent. It also makes later updates useful, because a reader can see which condition improved rather than being asked to accept another broad statement about momentum.
What would change our reading
We would place greater weight on a project when its public record reconciles the connection date, construction budget and customer timetable. We would become more cautious if the same milestone repeatedly moves, if announced demand cannot be tied to a contract or if a financing plan depends on uninterrupted access to new capital. For the regulatory story, the next document to read is the formal outcome of the consultation. Until that is established, proposed requirements stay in the proposal column.
The case and the counter-case.
The signal
Power delivery is becoming a commercial filter for the AI buildout. A strong demand story cannot energise a site.
What could change it
Faster model efficiency, financing constraints or delayed customer commitments could weaken the case for proposed capacity.
The next checkpoint
Look for final connection rules, binding project milestones, energisation dates and contracted customer demand.
Inspect the evidence.
Sources checked 16 October 2026. Inline references connect factual sections to the documents below. Analysis is the publication’s interpretation; multiple documents from one institution are not independent corroboration.
- IEA · Key Questions on Energy and AI (2026), executive summary ↗2025 electricity use is an estimate; 2030 is the IEA central projection, not a realised outcome. Document date: 16 April 2026.
- Ofgem · Speculative data-centre connection proposals ↗Regulator statement dated 29 July 2026. Queue applications are not energised capacity. Document date: 29 July 2026.
- Ofgem · Proposed data centre connection reforms ↗Consultation closed 16 September 2026; this page does not establish that the proposals have become final rules. Document date: 29 July 2026.
Prepared with AI assistance under a disclosed fictional editorial identity. General analysis, not personalised investment advice. No paid placement is attached to this article. AI & authorship · Request a correction.


