This is a working paper (July 2026), cross-referencing the Economy pillar (capital investment strategy) and the Public Office Covenant (institutional guardrails against short-termism). It is published as a standalone working paper rather than folded into either pillar: several sections — including the Jar 1/Jar 2 classification test and ownership of the spinout/IP material — are explicitly unresolved and flagged for dedicated follow-up work. Treat conclusions here as provisional.
The core question this answers
Should the state prioritise clearing existing debt to create headroom for future strategic capital expenditure (e.g. fusion deployment), or should it accept current debt levels and rely on the productivity gains from that future investment to reduce debt-to-GDP organically?
Answer: this is a false binary. The resolution is to stop treating all government spending as one undifferentiated pool and split it structurally.
The framework
Jar 1 — day-to-day expenditure. Ordinary running costs of the state. This budget should move towards balance: don't borrow to fund consumption.
Jar 2 — ring-fenced capital/growth investment. Long-duration strategic capital projects (fusion, grid, nuclear supply chain, other multi-decade infrastructure). This jar can legitimately be debt-financed, because it funds an asset with a genuine future return, not present consumption.
The rule: Jar 2 is never raided to flatter Jar 1's headline numbers, and Jar 1 discipline is never achieved by cutting Jar 2.
This is a revival of the UK's historical "golden rule" (Brown-era: only borrow to invest, balance the current budget) — including an explicit acknowledgement of why it collapsed: governments have a standing incentive to relabel current spending as "investment" to dodge the rule. Any version of this rule needs a hard, independently audited definition of what qualifies for Jar 2, not just good intentions. This classification question is itself contested and needs its own dedicated work — see open question below.
Why the two jars aren't actually independent — the cost-of-capital feedback loop
This is the mechanism that breaks the naive "just wait for growth to pay down debt" argument:
- The entire economic case for the state (rather than private capital) financing very long-duration assets like fusion rests on the state having a low cost of capital (sovereign borrowing rates over 30–50 years vs. private capital's short fund life and high risk premium).
- That low cost of capital is itself conditional on fiscal credibility. If Jar 1 debt keeps rising and markets doubt the government's commitment to stabilising it, gilt yields rise across the board — including on the debt used to finance Jar 2.
- So poor Jar 1 discipline doesn't just look bad on paper — it can directly erode the cheap-financing advantage that made state investment in something like fusion superior to private capital in the first place. You can end up waiting for a cheap-money window that closes before the technology you're waiting to finance is even ready.
- Illustrative sensitivity: the same capital project, financed via 60-year sovereign-grade borrowing at ~3% real vs. 20-year private-equity-grade risk pricing at ~10%, produces roughly a 3x difference in the levelised cost of the resulting energy. Financing structure and cost of capital matter more than the underlying technology cost.
Why over-correcting on Jar 1 is also a failure mode
- Fiscal consolidation that cuts current spending is economically very different from consolidation that cuts capital/R&D spending. The latter does disproportionate long-run damage — cutting skills pipelines, grid capacity, or engineering base now to hit a headline debt number is a false economy that raises the cost of capturing any future dividend.
- The UK's own 2010–2019 austerity decade is the cautionary precedent: debt fell only from ~55% to ~85% of GDP over a full decade of consolidation — much slower progress than expected — partly because the consolidation itself suppressed the growth needed to improve the ratio (the IMF later conceded fiscal multipliers were underestimated, especially near the zero lower bound).
The precedent that should worry us most: North Sea oil vs. Norway
- The UK had a real, one-off productivity/revenue windfall in the 1970s–80s (North Sea oil) and largely spent it on current consumption and unemployment support through a recession, rather than banking it.
- Norway, with a near-identical windfall, built the Government Pension Fund Global — now worth over a trillion dollars — by converting a finite resource windfall into a permanent, intergenerational endowment.
- Implication: any plan that assumes "growth from fusion will pay down the debt" without a binding institutional mechanism forcing that to happen is relying on political discipline the UK has already demonstrated it doesn't have by default. The lock needs to be built before the windfall exists, via legislation requiring a supermajority to redirect Jar 2 funds, or an independent body (OBR-style) with real sign-off power — not a ministerial promise, which is exactly what a future Chancellor under spending pressure will break.
Delivery mechanism precedent (how Jar 2 could actually work)
- The RAB (Regulated Asset Base) model used for Sizewell C is a live example: consumers/ratepayers begin contributing during construction rather than only after completion, which de-risks the project for investors and avoids the interest-during-construction cost spiral that drove Hinkley Point C's overruns.
- This converts a decade-plus construction risk into something closer to a regulated utility return — exactly the kind of steady asset that patient, state-adjacent capital (sovereign wealth funds, pension infrastructure vehicles) is suited to hold.
Technology-selection uncertainty (separate axis from the financing question)
- Once a technology is proven (e.g. Rolls-Royce SMR — known Gen III+ adjacent reactor physics), the financing decision is a pure discounted cash flow problem: pick the cheapest available cost of capital, choose an amortisation schedule.
- Where the winning technology is not yet known (e.g. fusion — HTS tokamak vs. FRC vs. laser ICF), the right framework is a portfolio/real-options approach: fund multiple structurally different bets in parallel, gate capital release on technical milestones (as ARPA-E and the US DOE Reactor Pilot Program already do), and only shift to single-commitment RAB-style infrastructure financing once uncertainty has narrowed from "does this work at all" to "execution risk only."
- Practical implication: a state should not collapse to a single national fusion champion while the physics question is still open, even though the eventual DCF math for the winning technology will look very attractive once it's known.
The classification test (v2) — two independent tests, not one
Early framing conflated "is this capital" with "is this strategic," which reopens exactly the gaming loophole that killed the Brown-era golden rule (any current spending can be dressed up as "job-creating in a strategic sector"). Resolved into two separate tests:
Test 1 — universal capital test (applies to every pound, every sector, no exceptions). Does this spending create a durable asset (physical infrastructure, IP/R&D output, or measurable human capital) that keeps generating a return — productivity, income, or cost-avoidance — for 10+ years after the spending stops? Yes → Jar 2 eligible. No (the benefit disappears the moment funding stops) → Jar 1, regardless of sector.
Test 2 — strategic sector overlay (a priority layer within Jar 2, not a gate into it). Of the spending that passes Test 1, anything in the designated key industries list gets protected/ring-fenced status — first defended when trade-offs bite. Capital spending outside the key sectors (e.g. a new hospital wing) still legitimately sits in Jar 2, it just doesn't get the extra protection.
This closes the loophole (a wage subsidy in a favoured sector still fails Test 1) and avoids wrongly excluding genuine capital projects that fall outside the strategic list.
Governance requirement: Test 1 needs an independent arbiter (OBR-style) rather than self-certification by the department proposing the spend — departments grading their own homework on "does this count as investment" is exactly how the original golden rule got gamed. (Open tension flagged: this is itself a real transfer of power away from elected government to an unelected body — noted, not resolved.)
Keep the strategic sector list short — deliberately, not by default
Once Test 2 exists, the temptation is to keep adding sectors to it ("peanut-buttering"). Resist this — concentration beats spread, for two independent reasons:
- Agglomeration/critical mass: strategic sectors need to cross a threshold of firm density, supplier depth, and skilled labour before spillover effects become self-sustaining. Split the same money across ten sectors and you likely get ten subcritical clusters instead of two or three that reach escape velocity. (South Korea's concentrated bets vs. Latin American import-substitution industrialisation spread across too many sectors at once is the textbook contrast.)
- Institutional attention is scarcer than money: a government can plausibly sustain two or three sectors as genuine cross-party priorities for 20 years. It cannot do that for ten. Industrial policy failures are disproportionately failures of consistency across electoral cycles, not failures of initial funding.
Counter-risk, for balance: concentration cuts both ways — fewer bets means more consequence per wrong bet (British Leyland, Concorde). A short, arbitrary list is exactly as fragile as a long, diffuse one. The discipline has to be in the selection criteria (real or credible UK edge; genuine strategic/security dimension beyond commercial upside; pulls other sectors up with it), not just in list length.
Distinguish two different kinds of "diversification": diversifying how to win once you've picked where to compete (e.g. funding multiple fusion reactor technologies in parallel — see technology-selection section above) is cheap insurance against picking the wrong approach within a committed domain. Diversifying where to compete in the first place (spreading the sector-selection decision itself thin) is the peanut-buttering failure mode. Don't confuse the two.
Evidence base for candidate key industries — UK university spinout data (July 2026)
Rather than picking sectors abstractly, use revealed commercial success as a starting signal: where has UK research already produced independently-validated commercial outcomes. Caution: academic subject-prestige rankings (QS/THE) are a different and largely unhelpful signal here — Oxford topping global rankings in Anatomy & Physiology, Anthropology, Geography, and Modern Languages, or Cambridge in Archaeology and English Literature, says nothing about exportable industrial technology. Spinout commercial value is the metric that actually answers the sector-selection question.
The data: UK university spinouts have grown to £49bn cumulative value since 2010 (nearly tripled since 2020), with £1.3bn in VC funding into spinouts in 2025 concentrated overwhelmingly in three sectors — life sciences/biotech & pharma, AI, and quantum (incl. semiconductors) — which together account for the vast majority of spinout value. Two of Europe's six billion-dollar-plus spinout exits in 2025 came from Oxford alone (OrganOx, medical device, ~$1.5bn; Oxford Ionics, quantum, ~$1.1bn). This converges with the UK Government's own Modern Industrial Strategy, which already flags life sciences, AI, and quantum as priority sectors, backed by a £2.5bn commitment to AI and quantum specifically (sectors the government projects at £212bn value within a few years). Worth checking the full Modern Industrial Strategy sector list against the "keep it short" discipline above — flag if it's already showing peanut-buttering symptoms.
Regional note for levelling-up framing: not purely a Golden Triangle (Oxford/Cambridge/London) story — real depth also in Bristol (AI chips, e.g. Graphcore), Manchester (techbio/enzyme engineering), Newcastle (synthetic DNA manufacturing), and Glasgow (quantum devices).
The additionality principle — why "follow the money" isn't sufficient on its own
Revealed commercial success is a necessary check but not sufficient for deciding where state capital belongs — the operative question is additionality: does state capital change the outcome, or does it just subsidise something private capital would have funded anyway (and potentially crowd out that private capital in the process).
Two structurally different cases, requiring two different roles for state capital — not one flat sector list:
- Tier A — revealed winners (AI, quantum, life sciences). Private capital already functions well here: VC's portfolio model needs the possibility of outsized, fast-compounding returns within a ~7–10 year fund life, and these sectors fit that shape (Oxford Ionics and OrganOx both exited within a normal VC timeframe). State capital chasing the same deals has low additionality and risks crowding out the functioning ecosystem. The state's actual job here is narrower: fix the UK's specific leak point — the scale-up gap, where promising spinouts are acquired early or move their listing abroad rather than growing into large domestic companies (UK spinout IPO activity has slowed since its 2021–22 peak, while acquisition exits continue). That means patient growth-stage capital and domestic-listing incentives, not funding the underlying science.
- Tier B — additionality bets (energy, agriculture). Fails the VC fit-test on every axis: multi-decade payback horizons incompatible with any private fund's life (the same r-g/cost-of-capital logic built out for fusion above), steady/modest rather than power-law returns, and — the dimension that's different in kind, not just degree — a large share of the value is externality (energy security, food security, national resilience against coercion or shock) rather than privately appropriable return. A rational private investor underweights that externality because they can't capture it; the state, by construction, both can and should weight it. This is the tier the RAB/sovereign-patient-capital case earlier in this document was actually built for — here the state needs to be the primary capital source, not a gap-filler.
Practical implication for the strategic sector list (Test 2): don't build one undifferentiated list. Tag each candidate sector by which tier it's in, since the tier determines both the type of intervention (primary funder vs. gap-filler) and the justification you'd give a skeptical Treasury (additionality/externality case for Tier B; scale-up/retention case for Tier A) — these are different arguments and conflating them weakens both.
Tier-specific incentive design — the tools don't transfer between tiers
Because the market failure differs by tier, the policy tools have to differ too. Testing whether a tool is coherent for a given tier: try swapping it to the other tier — if it becomes nonsensical, that's confirmation the two tiers are genuinely distinct interventions, not two flavours of the same one (e.g. a retention clawback solves nothing in Tier B, where private capital never shows up in the first place; a revenue-certainty guarantee is meaningless for a Tier A SaaS/software exit).
Tier A (AI, quantum, life sciences) — problem is retention and pull-through, not capital scarcity: - Growth-stage co-investment alongside private VC/PE at Series C/D+ (e.g. British Growth Partnership-style vehicles), so companies aren't forced into an early trade sale purely for liquidity. - Retention-conditional tax treatment: R&D credit or capital gains relief with claw-back if HQ or IP ownership relocates abroad within a set period. - Public procurement as anchor demand (defence, NHS, security-cleared applications) — demand-side pull that gives companies a domestic scale path instead of needing a foreign strategic acquirer for market access; cheap relative to direct investment. - Pension fund allocation reform (Mansion House Compact-style) — channel a small mandated allocation of private long-duration capital to the growth-stage maturity gap, rather than the state having to be the investor itself. Hard comparator (July 2026): pension funds account for only 10% of UK VC investment vs. 70% in the US; the largest UK DC schemes allocate ~1.4% of assets to growth capital vs. 13% (US) and 21% (Canada) among the largest funds, and ~4.6% (Australia) — roughly two-thirds of UK DC schemes have no VC allocation in their main funds at all. This isn't a modest gap, it's a near-total structural absence of exactly the patient, decades-long capital pool that should naturally match venture-stage risk.
The VC ecosystem gap — scale is improving, sourcing is the real problem
UK VC deployment reached $23.6bn in 2025 (+35% YoY, first annual expansion in four years, ending the post-pandemic "funding winter"), with 2026 tracking toward a further 50%+ increase — deployment volume is not the primary bottleneck it once was.
The sourcing gap is the real issue. UK VC fundraising (new fund capital raised, distinct from capital deployed) was only £2bn in 2025 — far smaller than the $23.6bn actually deployed, meaning a large share of UK deal flow already depends on foreign capital rather than domestic institutional money. This compounds the retention/scale-up problem elsewhere in this document: foreign-funded rounds create foreign expectations about where a company eventually lists or gets acquired.
Already a live policy front, not a green field — position any Generational Reset proposal as accelerating this rather than inventing it: - Mansion House Accord: major UK DC pension providers have pledged to allocate at least 10% of assets into "growth assets" (including VC) by 2030.1 - British Business Bank capacity expanded to £25.6bn; its British Growth Partnership has already attracted commitments from Aegon, NatWest Cushon, and London CIV.2 - Nest (UK's largest pension fund by membership, £68bn) has committed up to £1bn to VC by 2030, with £200m already deployed via Schroders Capital.3
Reinforces the additionality case already in this document: government and government-linked entities are already the dominant funder of UK early-stage business — roughly 40% of VC fund commitments in 2023 were government-linked, rising to 50%+ for some deep-tech companies. The state is not choosing whether to be involved; it already carries most of the load. The actual policy question is whether current support is structured to eventually crowd in a self-sustaining, pension-backed private ecosystem (the US pattern) or remains permanent life support.
Historical honesty check on the US comparator: the US VC industry itself wasn't organically private from inception — the 1958 Small Business Investment Company programme gave early venture funds government-backed leverage, and much of the underlying technology later commercialised by VC (the internet, GPS, semiconductors) originated in DARPA and defence R&D spending decades before VC touched it. The model being held up as the target was itself heavily state-seeded — strengthening rather than undermining the case for the UK doing this deliberately now.
A separate limiting factor, not fixable by capital alone: deploying pension capital well requires experienced fund managers with pattern-recognition and networks built over multiple fund cycles — the same learning-curve argument used for fusion manufacturing and China's rail programme elsewhere in this document. Fixing allocation doesn't instantly fix capacity; that takes time regardless of policy. - IPO-friendly listing reform — make LSE listing genuinely cost- and tax-competitive with Nasdaq so "list abroad" isn't the default.
Tier B (energy, agriculture) — problem is duration/externality mismatch, not retention: - Revenue-certainty mechanisms: RAB or Contracts-for-Difference-style price guarantees, converting an externality-heavy, uncertain return into a regulated-utility-shaped cashflow that matches what long-duration capital (pension infrastructure funds) actually wants to hold — genuine complementarity, not charity, if structured right. - First-loss/blended capital: state takes the riskiest tranche specifically to catalyse private co-investment that wouldn't otherwise touch first-of-a-kind construction risk — crowding in, the opposite dynamic from the Tier A crowding-out risk. - Explicit externality pricing: a formal security-of-supply payment (the logic already used in the UK's electricity capacity market), monetising the food/energy security externality directly rather than leaving it unpriced. - Long-duration gilt-linked infrastructure bonds, purpose-built to match pension liability profiles, as a distinct instrument from ordinary sovereign debt.
Refinement: equity-vs-royalty and the case against a sector-differentiated equity ceiling
Building on the Tier A incentive toolkit above — the equity-reform direction (TenU/USIT Guide, 10% or less for less IP-intensive spinouts, 25% ceiling for IP-intensive ones like life sciences, average already falling from 21.5% to 16.1% in the past year)4 is correct in direction but the sector-differentiated ceiling deserves a harder look, not just a lower number.
The mechanism: a large university equity stake doesn't just capture value after success, it actively reduces the odds of success — a big non-operating shareholder on the cap table is something later-stage VCs price against (complicates option pools, dilutes future rounds, signals extraction rather than partnership). So a 25% ceiling for IP-intensive sectors isn't neutral cost recovery, it's a headwind on the outcome it claims to serve — arguably worse in life sciences than software, given how many funding rounds a biotech company needs before any revenue exists.
Steelman against flattening it further (the "entrepreneurial state" critique, Mazzucato): the state/university funded the highest-risk, otherwise-unfundable early research; near-zero equity on the wins socialises the risk and privatises the reward, shortchanging the public purse that took the risk nobody else would.
Resolution — separate the instrument from the return: equity is a bad tool for capturing university return because it's the thing damaging the venture's odds. Push equity toward flat and low across every sector (drop the sector-differentiated 25% ceiling). Let royalties-on-revenue carry whatever legitimate return case survives from the Mazzucato side — royalties don't sit on the cap table, don't dilute future rounds, and scale with actual success rather than being negotiated as a fixed upfront slice regardless of outcome (some TTOs, e.g. Cambridge, are already moving toward this "mixed model").
Fund TTO/incubator infrastructure directly, not via extraction from the ventures it nurtures: this is a Tier B-shaped public good (high externality, benefits the whole regional pipeline, not fully captured by any single spinout) currently being funded with a Tier A-shaped tool (equity capture). Fund it via direct UKRI/state grant — the same logic already used for the underlying research — and let individual spinout equity/royalty terms be optimised purely for maximising volume and success probability.
Refinement: patents/IP — the friction is process speed, not protection strength
Distinguish two different claims: "IP protection slows spinouts down" (largely false as stated) vs. "the process of securing IP and negotiating terms slows spinouts down" (true, and already the documented problem — spinouts commonly take 12–18 months to launch). The fix already underway (USIT Guide's standardised "landing zones", separate lighter-touch USIT-for-Software guide) targets process speed without weakening protection — correctly, since removing protection would cut against Tier B sectors specifically (life sciences, materials, energy tech, engineering biology), where a discovery published without patent protection can simply be replicated by a better-capitalised (often foreign) incumbent, undermining exactly the sovereign-capability goal Tier B exists to serve. Software/AI is the genuine exception, which is why it already has its own lighter-touch guide. Keep the differentiation by IP-intensity; push hard on speed (non-dilutive translational funding to cover patent-filing costs upfront, faster TTO decision timelines), not on weakening protection uniformly.
Cross-pillar dependency map for this section
This spinout/incubator/TTO work is not solely Economy-pillar material — flag explicitly before drafting, to avoid duplication or contradiction with other pillars:
- Education: academic incentive structures (promotion/tenure still reward publications and grant income over commercialisation) are upstream of whether spinout activity happens at all — equity/royalty reform is necessary but insufficient without this. PhD entrepreneurship training and TTO capacity are themselves a skills pipeline, with a direct cautionary precedent already in this document (National College for High Speed Rail collapse) worth cross-referencing rather than assuming immunity.
- Healthcare: life sciences/medtech spinouts (OrganOx) feed NHS innovation adoption directly; the Tier A "procurement as anchor demand" lever is substantially an NHS-shaped commitment.
- Defence: AI, quantum, robotics, and space/photonics spinouts are substantially dual-use; the same procurement lever is also a Defence-pillar commitment. The National Security and Investment Act 20215 already gives government power to scrutinise/block foreign acquisitions in sensitive sectors (AI, quantum, synthetic biology, satellites, advanced materials, semiconductors) — a harder, statutory version of the retention-clawback tool already proposed; the reform question is whether it's being applied with enough teeth to early-stage spinout acquisitions, not whether a new mechanism is needed.
- Agriculture: UKRI's strategy explicitly targets robotics, AI-enabled systems, and engineering biology for agricultural transformation — meaningful Tier B agritech spinout activity is Agriculture-pillar substance; check for duplication/contradiction with Agriculture's existing robotics/precision-ag threads.
- Public Office Covenant: the governance questions running through this whole document (OBR-style independent arbiter, direct-funding-not-extraction principle, continuity-lock mechanism) are institutional-design questions as much Public Office Covenant's territory as Economy's — this document may need to be jointly owned rather than filed under Economy alone.
Retail vehicle idea (unplaced) — "UK Growth Investments" (UGI)
Flagged as not yet slotted into the Tier A/B structure — worth its own placement decision later.
Origin correction, for the record: this idea was prompted by a belief that Chancellor Reeves had tied ISA tax relief to LSE-listed holdings specifically. As of July 2026 this doesn't appear accurate — the actual Autumn 2025 Budget policy cuts the Cash ISA allowance from £20,000 to £12,000 for under-65s (from April 2027) and proposes a 22% tax on interest earned on cash held inside Stocks and Shares ISAs, both aimed at nudging savers from cash into investment generally, not specifically into UK-listed assets. May be conflating the earlier, apparently-shelved 2024 "British ISA" proposal. Worth re-verifying before citing as precedent in the pillar draft.
The core design tension: growth/VC-style returns are inherently variable and power-law distributed (most investments fail, a few pay for the rest) — a literal "fixed return" promise on a vehicle actually investing in growth-stage companies is a contradiction unless someone else absorbs the variance. No way around this; it's the same "who holds the risk" question as the RAB/first-loss discussions elsewhere in this document.
Two honest structural options, not a free combination of fixed-return-plus-growth-exposure: 1. State-guaranteed floor, housed at NS&I. National Savings & Investments is already the UK's trusted, state-backed guaranteed-return retail institution (Premium Bonds, Guaranteed Growth Bonds) — natural home rather than a new institution. State absorbs downside variance to offer a genuine floor, investing pooled capital via British Business Bank/British Growth Partnership channels (see VC ecosystem section above). This is a real, sizeable contingent liability — needs honest disclosure and sizing, not concealment behind the word "fixed." 2. Tiered/first-loss structure, retail-facing — reuses the Tier B blended-capital tool (state/BBB takes first-loss tranche; retail savers hold a senior tranche behaving close to fixed-return in exchange for giving up most upside). Genuinely novel contribution here: extending an institutional-only tool (first-loss capital) down to mass retail participation.
Existing UK precedent, partial: VCTs and EIS/SEIS already give retail investors substantial tax relief (30% VCT/EIS, 50% SEIS, plus CGT benefits) for investing in early-stage/growth UK companies — but genuinely risk-bearing, no fixed return; relief compensates for real risk rather than disguising it. Also relatively niche/complex vs. ISA accessibility. UGI's real contribution, if it proceeds, may be mass-market simplification of an already-existing mechanism rather than a wholly new one.
Governance flag, consistent with this document's HS2/North Sea precedents: a "fixed return" that later proves unsustainable and gets quietly reduced under fiscal pressure is exactly the broken-commitment pattern already warned against. Any floor guarantee needs the same continuity-lock treatment as Jar 2 generally — legislated, not just promised.
Worked case study: HS2
Test 1 verdict: passes trivially. Physical rail infrastructure, multi-decade asset life — uncontroversially Jar 2. Classification was never HS2's problem.
Test 2 verdict: does not automatically qualify for top-tier protection. HS2 is connectivity infrastructure that indirectly serves whichever sectors sit along the corridor — "it enables growth generally" is precisely the vague, sector-agnostic justification the short-list discipline is designed to screen out. It should stand on its own transport-economics business case, not borrow legitimacy from a protected sector list.
The real lesson is a governance/lock failure, not a classification failure. Cost-to-complete rose from an original ~£32–33bn (2010, full London–Manchester–Leeds network) to £87.7–102.7bn (2026, London–Birmingham only, Euston link deferred to 2040–43) — roughly triple, for a much smaller asset. The eastern leg (Leeds) was cut in 2021, the western leg (Manchester) in 2023, top speed cut from 360 to 320 km/h, and Automatic Train Operation — the one genuinely novel technical element — was dropped entirely in the May 2026 reset. This is successive governments each finding it politically survivable to hack off a piece: exactly the failure mode the Jar 2 "lock" mechanism (supermajority requirement / independent sign-off, see North Sea oil precedent above) is meant to prevent — just applied to scope continuity rather than fund-raiding directly. Contributing mechanisms worth carrying into any future fusion/strategic-asset procurement design: construction started in 2020 while design was only ~10% complete; optimism bias in the original business case (the general, well-documented megaproject pattern — over budget, over time, under benefits — not a UK-specific pathology); and risk transferred onto contractors, who then priced bids defensively once holding cost-overrun risk they hadn't caused.
Skills/IP legacy — the sharpest illustration of the lock problem. The National College for High Speed Rail opened 2017 with capacity for 3,000+ learners (~£26m Doncaster site, ~£40m across two sites), explicitly framed as leaving "a lasting skills legacy." It closed July 2023 with just 49 students and 119 apprentices, plus a written-off £2.8m loan. Cause, per Doncaster's Chamber of Commerce: the government's scope "flip-flopping" meant employers couldn't commit apprentice volumes with confidence contracts would still exist by graduation.6 The instrument built specifically to generate the exportable human-capital dividend collapsed as a direct, documented consequence of the same continuity failure that gutted the physical scope.
Three national models for capital-project technology/skills dividends
Relevant to whether a future strategic bet (fusion or otherwise) should be explicitly designed to generate an exportable technology/skills legacy, not just the physical asset:
- China (forced transfer + indigenisation + export): technology transfer made a condition of market access (JVs with Kawasaki, Alstom, Siemens, Bombardier), progressively indigenised into domestic designs (Fuxing/CR400), then exported (Jakarta–Bandung HSR, Belt and Road projects). Requires domestic market leverage large enough to force supplier concessions, and programme scale (~40,000km of HSR) big enough to generate a real manufacturing learning curve. Neither lever is available to the UK at HS2's scale.
- Japan (organic domestic build, patient decades-long payoff): Shinkansen (1964, predates TGV by ~17 years) built substantially from domestic engineering capacity — notably aviation/precision engineers redirected after post-war restrictions on Japanese aircraft manufacturing — as an entirely new standalone standard-gauge network, not an upgrade of existing lines. Framed from inception as building sovereign technological capability, not just a transport service. Export dividend was real but took decades: Taiwan HSR (2007) and the Mumbai–Ahmedabad corridor both run Shinkansen-derived technology. The consistent thread across decades of Japanese governments is the enabling factor — the same continuity/lock property HS2 lacked.
- UK (HS2): neither forced-transfer leverage nor sustained multi-decade continuity. Treated skills/IP legacy as an ancillary byproduct (a training college) rather than a locked, protected primary objective — and it collapsed for exactly that reason.
Implication for any future strategic bet: if an export-grade technology/skills dividend is wanted (not just the physical asset), it has to be an explicit, contractually locked objective from inception — protected by the same continuity mechanism proposed for Jar 2 generally — not something assumed to fall out the side of the main project.
Open question — still needs dedicated follow-up work
What actually goes in Jar 1 vs. Jar 2 (Test 1) is a political question at the margins, and reasonable people will disagree on specific cases. Contested cases flagged for further work: - Education (human capital investment vs. current service delivery — capital spending on buildings is clean Jar 2; teacher salaries produce the actual human capital but look like Jar 1) - Military spending (equipment/infrastructure vs. personnel costs — a defence hawk treats readiness itself as the asset; a fiscal conservative treats only hardware as capital) - NHS (capital estate vs. staffing costs — preventive care spending has a stronger investment case than most current spending, since it demonstrably reduces future acute costs) - R&D spending (opex in form, capex in function) - Welfare vs. active labour market programmes (pure income support is clean Jar 1; retraining/childcare-enabling-work schemes have a real investment logic)
The pattern across all of these: the fight is rarely about the technical definition of investment, it's about whose preferred spending gets to claim Jar 2's protected status. This needs its own session: map the plausible classification schemes per case, identify the genuine partisan fault lines, and stress-test whether Test 1's independent-arbiter mechanism can survive the same political gaming that killed the Brown-era golden rule.
For public discussion. Not affiliated with any political party. | generationalreset.org
The Generational Reset | In Discussion: The Two-Jar Fiscal Framework | For public discussion. Not affiliated with any political party. | generationalreset.org