The UK Defence Investment Plan commits £298 billion over four years, including £15 billion of new investment. The headline number is not the story. Where the money goes is.

More than £5 billion is earmarked for autonomous systems: an Army the plan describes as transformed by AI and autonomy, a Hybrid Navy, and next-generation autonomous combat air. £63.6 billion goes to the Defence Nuclear Enterprise through Dreadnought and SSN-AUKUS. £8.6 billion goes to the Global Combat Air Programme with Japan and Italy. Dan Jarvis, Defence Secretary when the plan was published, framed it as a response to a character of warfare changing at the fastest rate in history.

He is no longer in post. John Healey resigned from Defence in June 2026, Jarvis published the plan and left the department the following month, and Wes Streeting inherited its delivery in July: three Secretaries of State in roughly two months. The plan runs to 2030. The estate it funds will still be in service long after that. Whatever continuity Defence gets across that horizon is not going to come from the ministerial layer.

Read past the totals and a second signal appears, one that matters more to how the money is spent than to how much of it there is. The Defence Investment Plan broadens the supplier base beyond the traditional primes toward firms that deliver AI, autonomy, cyber, and digital integration. The plan is the funding vehicle for the June 2025 Strategic Defence Review. The window from strategy to procurement is opening now.

The fourth largest line, and the one nobody leads with

There has been a fair amount of noise recently about the nuclear deterrent swallowing the defence budget. The plan's own Chart 2, planned spend by chapter over the next four years, lets anyone check it.

People and Veterans is the largest chapter at £78.5 billion, and rightly so. People are the centre of defence. Everything else in the plan is equipment that someone has to operate, maintain and be housed alongside, and the Strategic Defence Review was explicit that poor accommodation was driving a crisis in recruitment and retention. Funding people ahead of everything else is the correct order, not an anomaly to be explained away.

The Defence Nuclear Enterprise is second at £63.6 billion, which is close to what Air, Land and Maritime receive combined. So the deterrent does not take most of the budget. It does take about as much as the three conventional fighting domains put together, which is the version of the claim that survives contact with the source.

Fourth on that list, ahead of Land at £19.2 billion and Maritime at £18.0 billion, is Defence Infrastructure and Estate at £22.7 billion.

The plan is candid about why. Project Castra commits around £24 billion over ten years against what the plan describes as "decades of conscious underinvestment in an ageing estate". Project Trenchard commits around £10 billion "to reverse decades of estate decline". The Strategic Defence Review had already confirmed that "the Defence estate is the foundation on which our Armed Forces operate", and identified poor accommodation as a driver of a crisis in recruitment and retention.

The estate is not a supporting line in this plan. It is the fourth largest one. It is also the one whose asset data moves fastest, because multiple suppliers write to it every day and none of them owns the whole record. That is where the autonomy layer is most likely to be pointed at records nobody has reconciled.

Autonomy inherits whatever data it is fed

An autonomous system does not generate its own understanding of the estate it operates in. It inherits one. It inherits the asset records, the configuration baselines, the maintenance histories, and the criticality classifications that already exist. If those records are incomplete, inconsistent, or ungoverned, the autonomy layer does not correct them. It industrialises them.

This is the part of the plan that no line item captures. You can procure an autonomous capability. You cannot procure the governed data foundation it depends on, at least not as a bolt-on after the contract is signed. In a high-consequence Defence environment, ungoverned data is not an administrative inconvenience. It is a capability risk, and in some cases a safety and audit risk.

The plan does not disagree. Its own account of the Integrated Force describes "underpinning the force with a shared digital and data foundation for software-defined warfare", and states that "modern military success is not about individual platforms (like tanks or jets) but about how seamlessly those systems connect across every domain, maximising the benefits of autonomy and data".

Connection across domains is a data problem before it is a technology problem. So is autonomy. The plan says so in its own words, on its own pages. What it cannot do is fund the reconciliation work that would make those words true on the estates the money lands on.

Anyone who has worked inside defence infrastructure knows the shape of the problem. Through-Life Management assumes you can trust the asset data across decades of service. The Asset Data Standard assumes a discipline that many estates have never fully closed out. Systems such as JAMES hold enormous value, and they also sit alongside legacy records that were never fully reconciled. The gap between what the standard requires and what the data actually contains is where AI and autonomy projects quietly fail.

That is the backward-looking half of the problem, and it is the half most people mean when they talk about data quality. The forward-looking half is harder, and this plan makes it urgent.

Asset data standards were written for things with long, stable lives. A capability is procured, it has a configuration baseline, that baseline changes through controlled modification, and the record follows behind it. The plan buys something that does not behave that way. A Hybrid Navy mixes crewed and uncrewed. Next-generation autonomous combat air changes what it can do through software rather than through refit. Software-defined warfare means capability arrives on a release cycle rather than a procurement cycle.

So ask the question that follows. When a platform's function changes because its software changed, does the asset record change with it? Is criticality reassessed when the role shifts? Does the maintenance regime follow the new function or the original one? Is the thing described in the register still the thing operating on the front line?

A data model built on the assumption that an asset is a physical object with a stable configuration cannot describe an asset whose capability is a moving target. That is not a data quality problem, which can be closed with effort and money. It is a data model problem, which can only be closed by design, and it has to be designed before the capability arrives rather than retrofitted after it.

The risk in a rapidly changing force is not that the data is dirty. It is that the data model is faithfully describing the last generation of asset while the front line is already operating the next one.

None of this argues against the plan. It argues for getting the sequence right. The organisations that will capture the autonomy dividend are not the ones that move first. They are the ones whose asset data can be trusted when the autonomy arrives.

ISO 55000 was designed for exactly this class of problem, and it remains a defensible, applicable discipline for governing long-life asset data in regulated environments. The UK-sovereign, security-cleared, practitioner route into this work matters in Defence in a way it does not in most sectors.

There is a second question this plan raises, and it is the harder one. Most of the estate the money lands on is not maintained by the department. It is maintained by contractors, and the asset record moves between them on dates the plan itself publishes. That is a separate piece of work, and a separate conversation.

Foundations before automation.

Sources and further reading