Good morning,

Welcome back to Delivery Signal — a weekly read on what UK government strategy documents actually commit to, and whether delivery is keeping pace. Built from the same tracking that sits behind strategygov.uk.

What government published last week

The big addition to the tracker this week is the UK's AI stack — now mapped end to end. We ingested the founding AI Opportunities Action Plan (Matt Clifford, Jan 2025 — 50 recommendations, all accepted by government), alongside the June 2026 delivery cluster: the AI Hardware Plan, the AI Economics Institute, the AI Champions' adoption response, and the Digital & Technologies Sector Plan Year One.

Put together, you can now trace a single line from the 2025 recommendations to the 2026 spending — and see where the joins are under strain. One join stands out.

Tension of the week

The UK's AI ambition is, by the government's own words, constrained by power — and the plan to supply it isn't sequenced.

The compute build-out is large and concrete: a £750m national supercomputer, an AI Research Resource expanding at least 20x by 2030, and 5 AI Growth Zones drawing £28.2bn of mostly private data-centre investment. All of it is electricity-hungry, and all of it lands on the same grid that Clean Power 2030 is simultaneously racing to decarbonise and expand.

The documents don't hide the problem — they site Growth Zones for "greater energy capacity" and point to "the UK's first small modular reactor and a renewables-powered microgrid" as power sources. But an SMR and a microgrid are not credible near-term supply at data-centre scale, and grid-connection queues are already a flagged constraint on clean power itself. The mitigation the Action Plan created — an AI Energy Council co-chaired by the DSIT and Energy secretaries — is advisory. Nowhere is there a published plan that says how much grid headroom is reserved for compute versus the wider electrification Clean Power 2030 needs.

The trade-off is named. It isn't yet resourced or sequenced. That gap is the story.

Full tension — evidence and current status: strategygov.uk/wiki/tensions/ai-compute-energy-clean-power

Commitment watch

This week's ingest added the founding AI commitments to the tracker — the dated promises the 2026 spending is delivering against:

One more worth flagging: the Action Plan put the prize from AI adoption at "£400bn by 2030," while the government's own later documents use an OECD-based "£55–140bn of added GVA by 2030." That's a roughly 3–7x gap in the headline benefit case — the kind of thing the new AI Economics Institute exists to reconcile. (the tension)

What to watch

• Summer 2026 — the £150m advance commitment to buy novel AI chips is due to be awarded
• AISI legislation — still unconfirmed; the frontier-model regulation bill is the one to watch
• UK Compute Roadmap — the Action Plan's long-promised 10-year compute strategy; next on the tracker's ingest list
• AI Growth Zone power — whether DESNZ/NESO publish a compute-demand allowance within the grid-connection reforms

Delivery Signal tracks 170+ pages of UK government strategy commitments, updated as new documents land and lint-checked weekly for status changes and contradictions. Reply to this email with anything you'd like us to track.

This week added the full UK AI stack to the tracker — 1 founding strategy, 4 new dated commitments, and 2 fresh cross-government tensions.

Know someone who'd want this?