Stakeholder landscape · working document
Around 700 PFI contracts reach an ending nobody designed, peaking in the mid-2030s. This maps every party in that landscape by role, responsibility and influence — and asks of each one the only question that matters for the strategy: what do they contribute that no one else does?
Every actor by role, responsibility and influence, with a verdict on what it contributes. The verdict column is the point: a landscape crowded with bodies is not the same as a landscape with an owner. Read the verdict column alone and the group's central diagnosis becomes visible — only three actors are decisive, and none of them currently treats this as their problem.
| Actor | Role | Responsibility | Influence | What it adds to this landscape |
|---|
The contract lifecycle as a value stream. Top lanes are who is acting; red is where public value drains out; green is the intervention that would have stopped it. Read the red lane left to right — every leak downstream is cheaper to prevent upstream, and the system only ever engages at stage 5.
The public side by altitude, with accountability lines between levels. Select a node to trace its chain upward.
The relationships that carry actual control, grouped by type. Ownership behaves very differently from advice, and the cases where an actor has more than one parent are where the strategy either breaks or unlocks.
the group's three end-of-contract scenarios, extended into who is exposed under each — action PFI-08. Scenario 3 was judged most likely absent intervention, and it is the one in which the public side does worst.
The contract simply expires. Assets transfer for nominal sum; maintenance and service liabilities fall wholly on the public sector. Few players — the authority and the Treasury.
Exposed: the receiving authority, in full and immediately. The s151 officer carries an unprovided liability. TUPE staff transfer with up to thirty years' accrued service.
Who gains: the SPV and its equity, exiting cleanly at the lowest contractual standard.
A flip involving SPV owners or sponsors, potentially including retained government interests or golden shares.
Exposed: depends entirely on where any retained public interest sits — which is what the UKGI check was meant to establish. On current evidence no such interest has been found, which makes this scenario thinner than hoped.
Who gains: whoever structures the flip. The authority is a spectator unless a golden share exists.
The incumbent SPV rolls into a further long concession. Judged the most likely outcome absent intervention.
Exposed: the authority negotiates renewal from the weakest possible position: no condition baseline, no asset register, no in-house FM capability, and a live service it cannot switch off.
Who gains: the incumbent equity holder, twice — on the tail of the original deal and again on the renewal.
Computed from the 720-asset catalogue by mapping each English procuring authority to its mayoral combined authority. The critical split is in remit versus geography only: local authority, fire and police schemes sit within functions a mayor holds or can reach, while NHS trust schemes sit inside the same boundary but answer to DHSC, not the mayor. Only the first column is exposure a mayor could act on.
| Mayoral area | In remit | £m in remit | Geography only | £m geography | £m total | In remit, expiring by 2032 |
|---|---|---|---|---|---|---|
| Greater London GLA · 2000 | 66 | 2,953 | 10 | 981 | 3,934 | 34 |
| West Yorkshire WYCA · 2014 | 24 | 1,763 | 4 | 427 | 2,190 | 8 |
| South Yorkshire SYMCA · 2014 | 18 | 1,414 | 2 | 41 | 1,455 | 5 |
| Greater Manchester GMCA · 2011 | 31 | 1,324 | 6 | 742 | 2,067 | 12 |
| East Midlands EMCCA · 2024 | 20 | 1,176 | 5 | 422 | 1,598 | 10 |
| Liverpool City Region LCRCA · 2014 | 11 | 1,125 | 0 | 0 | 1,125 | 4 |
| North East NECA · 2024 | 25 | 1,064 | 8 | 552 | 1,615 | 11 |
| West Midlands WMCA · 2016 | 20 | 1,008 | 8 | 1,420 | 2,429 | 8 |
| West of England WECA · 2017 | 6 | 363 | 2 | 502 | 865 | 3 |
| Tees Valley TVCA · 2016 | 6 | 151 | 0 | 0 | 151 | 5 |
| Cambs & Peterborough CPCA · 2017 | 3 | 137 | 3 | 518 | 655 | 0 |
| York & North Yorkshire YNYCA · 2024 | 5 | 78 | 1 | 312 | 389 | 4 |
| Total | 235 | 12,556 | 49 | 5,917 | 18,473 | 104 |
Andy Burnham became PM on 20 July 2026. the group's "figure with genuine exposure at the top of the pyramid" is therefore literal: the PM personally carried Greater Manchester's 31 in-remit PFI projects (£1.3bn) for nine years.
Within the mayoral tier itself, GM ranks fourth by in-remit capital — London (£2.95bn), West Yorkshire and South Yorkshire are larger. The current GM mayor inherits the portfolio; the argument that reaches No.10 fastest is the one framed in the former mayor's own experience of it.
195 English projects sit in areas with no mayor at all — Kent, Lancashire, Essex, Staffordshire, Cornwall, Northamptonshire and others. A mayoral strategy addresses a majority of the exposure and leaves roughly two-fifths of it structurally unowned.
That is an argument for using mayors as the salience engine rather than the delivery route: they can make it a national argument, but they cannot be the national answer.
WMCA carries £1.42bn of geography-only exposure against £1.01bn in remit — the only area where what the mayor cannot touch exceeds what he can. Almost all of it is large acute hospital PFI.
This is the clearest single illustration of the accountability split: a mayor can be asked about a hospital in his city region and has no locus whatever to act on it.
Of those, 34 are in London and 12 in Greater Manchester. These are live within one mayoral term, which makes them a legitimate manifesto and scrutiny item now rather than a next-decade abstraction.
London's concentration is the strongest available argument for approaching the GLA — and it has been absent from the strategy entirely.
pfi_catalogue.db (720 assets) over English projects only. Authority names were normalised and matched to combined-authority membership by string rule; the constituent-council mapping is our own and has not been checked against each authority's constitutional order, and the 2024 bodies (North East, East Midlands, York & North Yorkshire) carry the most risk of membership error. Central and national bodies (MoD, National Highways, GCHQ, HMRC, DfE, prisons) are excluded from geographic attribution. Capital values are as recorded in the catalogue and are not inflation-adjusted, so they are not comparable across contract vintages. The source data contains authority name variants (Leeds and Leeds City Council appear separately) which the normalisation is designed to catch but may not catch completely. Treat this as a directional first cut for targeting, not as a published exposure statement.Current state · the delivery model
There is no single standard PFI, but there is a standard shape. This is the machine as designed — who puts money in, who builds, who operates, who maintains, and where each party's incentive points once construction is finished.
The SPV is the hub: a thinly-capitalised company whose only assets are the concession and the subcontracts. Everything the public sector wants is delivered through it; everything the private sector earns is extracted through it. Note how many boxes are affiliates of the same sponsors — the construction contractor and the FM provider are frequently related parties to the equity holders, which is the structural reason self-reported condition data is not independent evidence.
The same project across time. The gold lane shows where cash actually flows; red is where public value is lost. The pattern to notice: the public sector's leverage is highest at the two ends and near zero in the middle, which is the reverse of when it pays attention.
The position paper was criticised for being insufficiently positive about what PFI genuinely delivered. That correction is right on the merits and useful tactically: a proposition that opens by trashing the model cannot credibly propose the next one.
A very large programme of schools, hospitals, roads and housing was built when direct capital was constrained. Whatever the cost of finance, the buildings are real and in use.
Construction risk genuinely transferred. The discipline of a lender with capital at risk produced completion performance the public sector had struggled to achieve directly.
The unitary charge bundled maintenance and could not be raided. Against the non-PFI estate, where maintenance is the first cut every year, PFI assets were protected from backlog for a generation.
The honest critique is narrower and more damning than "PFI was a rip-off". The model worked as designed for twenty-five years and had no design for year twenty-six. Handback was a clause, not a plan. That is a fixable design fault, which is exactly what makes a next-generation model arguable.
Future state · the models
Two questions, not one. What happens to the assets now reverting — and what model the next generation of public infrastructure should use. The first is forced on us by 2037. The second is the thirty-year mission.
Every expiring contract resolves into one of these, ordered by the capability the authority needs — which is also the order of how much value it can capture.
Scenario 3. The incumbent extends on its own terms because the authority has no baseline, no capability and a live service. Most likely absent intervention.
No capability needed Value captured: none
Asset returns publicly; services go to market. Works if the condition baseline is sound and records survived. Fails quietly if not — the new provider prices the unknown.
Procurement capability Value: moderate
Direct delivery; Islington's housing insourcing is the reference. Needs in-house technical leadership most authorities no longer have, and inherits the TUPE population with up to thirty years' accrued service.
High capability Value: high
Take control before expiry, then use the moment to re-baseline a decarbonisation and modernisation capital case rather than paying a repair bill. Hexham is the buy-out reference.
Capability + capital Value: highest
If the honest critique is that PFI worked for twenty-five years and had no design for year twenty-six, then the test for any successor is what it does at the end, not what it does at financial close.
| Model | Asset ownership | Condition risk | Return control | End-of-life design | Read |
|---|---|---|---|---|---|
| PFI / PF2the incumbent | SPV, reverting at expiry | SPV in theory; authority in practice, once the SPV dissolves | Uncapped. Refinancing and secondary-sale gains largely retained | None. Handback is a clause, not a plan | The problem |
| Mutual Investment ModelWales | SPV with a public-sector minority equity stake | Shared — the public stake buys a seat and sight of the data | Partially capped; public co-investor shares upside | Assurance built in: Independent Tester, Clerk of Works, BIM Information Manager | Closest live answer |
| NPD / hubScotland | SPV with capped private return | Shared, with a public-interest director in the room | Capped by design. Surplus returns to the public side | Standing vehicles may retain transferable client capability between deals | Solves the capability cliff |
| Regulated Asset BaseTideway, Sizewell C | Regulated company, indefinite | Regulator sets and enforces condition | Regulated rate of return, periodically reset | No end. The asset is never handed back — the problem is dissolved rather than solved | Works at scale only |
| Alliance / integratedNEC4 Alliance | Public throughout | Shared pain and gain across the alliance | Open-book, incentivised | Continuous — no expiry cliff exists to design for | Needs a capable client |
| Public + escrowed lifecyclethe minimum reform | Public throughout | Public, but funded — lifecycle ring-fenced and audited | Not applicable | The fund is the plan. Condition is a reported metric, not a terminal dispute | Cheapest fix |
The first version of this document carried no source list, against A7C's own production standard. This is the repair. Claims are graded by what actually stands behind them, because the mix is uneven and the reader is entitled to know which is which.
Local Partnerships ownership and structure — localpartnerships.gov.uk/about-us; LLP OC346845. Companies House filing not yet pulled; do that to make the check repeatable.
NISTA — established 1 April 2025 from the IPA and NIC as a unit within HM Treasury, lead minister the Chief Secretary. Parliamentary and gov.uk sources.
Local Audit Office — MHCLG local audit reform strategy (Dec 2024) and transition plan (Nov 2025).
Mayoral exposure table — computed from pfi_catalogue.db. Method and caveats stated in full on that tab.
Schools cohort figures, Stoke handback and liquidation detail, NAO condition-data and dispute findings, expiry-health-check ratings, the seven-year doctrine, SOPC4 condition and survey requirements, the 28-day adjudication point and Lancashire Schools [2024] EWHC 37 (TCC).
All sit in research/01–11 and docs/ in the catalogue with URLs attached. They are not yet carried into this document as inline citations. Do that before any external circulation.
Aggregate problem-size and adviser-savings figures circulating in discussion are excluded from this edition pending verification.
Authority-readiness anecdotes; sinking-fund exhaustion; investor concentration estimates; the proportion of councils without asset registers.
The completed-expiry record. Scenario 3 is a group judgement with no cited cases, and the document itself notes custodial is already through a first wave. Tabulating what actually happened — rollover, re-procurement, insource or buyout, and on whose terms — is the single highest-value outstanding task.
Procurement law on rollover. A continuation "on the incumbent's terms" is generally a direct award and is constrained. Untested here, and it may make scenario 3 harder than assumed.
Scottish Futures Trust. Page 3 borrows NPD/hub without asking whether SFT already runs the national expiry programme this document says nobody owns.
Every option — near-term route or successor model — is tested against the same seven dimensions. This is the scoring frame the workbench (page 5) uses, and it is deliberately a national-interest test, not a lowest-cost test: an option can win on price and fail the country.
Whole-life cost including the liabilities an option hides — TUPE, backlog, renewal pricing — not just the headline fee.
Schools open, wards safe, lights on through the transition. The dimension incumbents weaponise.
Statutory compliance, backlog trajectory, and whether the condition baseline is independently evidenced.
Does the option leave the authority stronger or permanently dependent? The capability cliff is the system's deepest wound.
Does the route sustain competitive tension, or hand the incumbent a renewal on its own terms?
LED, heat pumps, retrofit, changed demand. Original handback standards rarely cover any of it.
Open data, auditable decisions, a named owner. The dimension the whole landscape currently fails.
A generation of infrastructure was delivered by a model that worked as intended and was never designed to end. Around 700 contracts now reach that undesigned ending, peaking in the mid-2030s.
Every expiry is a decision on what replaces it. Taken together and taken deliberately, they are the largest single opportunity to reset how Britain builds and maintains public infrastructure for thirty years.
No body has the mandate, the authority and the interest simultaneously. The ask is not money first — it is an owner, with the funded trigger following.
The group structured the original PFI deals, is watching them expire, and is working on what comes next. That critique cuts both ways — that history is the asset, provided the proposition is reform rather than repetition.
Data explorer · the full catalogue
All 720 catalogued projects, filterable live. Slice by expiry window, geography, sector or commissioning body — every tile, chart and row below recomputes from the current filter. Click a bar in the expiry chart to jump to that year.
| Project ▾ | Authority | Sector | Region | Expiry | Capital £m | UC 23–24 £m | Flags |
|---|
pfi_catalogue.db (720 projects, HMT/IPA-derived catalogue with A7C enrichment: pre-SoPC flags, distress flags from Companies House, derived expiry years). Capital values as recorded, not inflation-adjusted — not comparable across contract vintages. 57 projects carry no recorded expiry date and are excluded from the expiry chart but included elsewhere. Unitary charge is the 2023–24 reported figure and is missing for some rows. Regenerate the dataset with scripts/mca_exposure.py siblings against a refreshed database before external use.Options workbench · live scoring
Ported from the July strategy canvas and reseeded to the 13 August position. Every option scores 0–3 against the seven national-interest dimensions (page 3). Your edits persist in this browser; export before sharing a position. Statuses: Decided · Shortlist · Explore · Parked · Rejected.