This is a working paper (July 2026, revised October 2026). It draws on the Economy pillar (capital investment strategy) and the Public Office Covenant (institutional guardrails against short-termism). It is published on its own rather than folded into either pillar because several parts, including the Jar 1/Jar 2 classification test and who owns the spinout and IP material, are unresolved and flagged for further work. Treat its conclusions as provisional.
The question this answers
Should the state pay down existing debt first, to make room for future strategic capital spending such as fusion, or should it accept current debt levels and rely on the productivity gains from that investment to bring debt down as a share of GDP over time?
Answer: neither, because the choice is a false one. The way out is to stop treating all government spending as one pool and split it in two.
The framework
Jar 1: day-to-day spending. The ordinary running costs of the state. This budget should move towards balance. Don't borrow to pay for consumption.
Jar 2: ring-fenced capital and growth investment. Long-term strategic capital projects such as fusion, the grid, the nuclear supply chain and other infrastructure that lasts decades. This jar can reasonably be financed by borrowing, because it pays for an asset with a future return rather than for consumption today.
The rule: Jar 2 is never raided to make Jar 1's numbers look better, and Jar 1 is never balanced by cutting Jar 2.
This revives the UK's old "golden rule" from the Brown era: borrow only to invest, and balance the current budget. It also has to face up to why that rule collapsed. Governments always have a reason to relabel current spending as "investment" to get round the rule. Any new version needs a hard, independently audited definition of what counts as Jar 2, not just good intentions. That classification question is itself contested and needs separate work (see the open question at the end).
Why the two jars aren't independent: the cost-of-capital loop
This is the mechanism that breaks the simple argument that we can wait for growth to pay down the debt.
- The economic case for the state, rather than private investors, financing very long-lived assets like fusion rests on the state's low cost of capital. Governments can borrow for 30 to 50 years at sovereign rates. Private funds have short lives and charge a high risk premium.
- That low cost of capital depends 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 borrowing that funds Jar 2.
- So weak discipline on Jar 1 does more than look bad. It can wear away the cheap financing that made state investment in something like fusion better value than private capital in the first place. A government can end up waiting for a cheap-money window that closes before the technology is ready to finance.
- An illustrative sensitivity: the same capital project financed with 60-year sovereign borrowing at about 3% real, against 20-year private-equity pricing at about 10%, gives roughly a threefold difference in the lifetime cost of the energy it produces. For capital-heavy assets, the financing structure and cost of capital can matter more than the cost of the technology itself.
Why over-correcting on Jar 1 is also a failure
- Cutting current spending is economically very different from cutting capital or R&D spending. Cuts to capital and R&D do much more long-run damage. Cutting skills pipelines, grid capacity or the engineering base to hit a headline debt number is a false economy that raises the cost of capturing any future return.
- The UK's austerity decade after 2010 is the warning. On the measures published at the time, public sector net debt rose from about 54% of GDP at the end of March 2010 to 83.5% at the end of March 2016, despite spending cuts explicitly aimed at bringing it down.1 Part of the reason is that the cuts themselves held back the growth needed to improve the ratio. The IMF later acknowledged that it had underestimated fiscal multipliers, especially when interest rates were near zero.
The precedent that should worry us most: North Sea oil and Norway
- The UK had a one-off revenue windfall in the 1970s and 1980s from North Sea oil. It largely spent it on current consumption and unemployment support through a recession, rather than saving it.
- Norway had a similar windfall and built the Government Pension Fund Global, worth just under NOK 21,300 billion (over $2 trillion) at the end of 2025, by turning a finite resource into a permanent endowment for future generations.2
- What follows: any plan that assumes "growth from fusion will pay down the debt", without a binding mechanism to make that happen, relies on political discipline the UK has already shown it doesn't have by default. The lock has to be built before the windfall exists. That means legislation requiring a supermajority to redirect Jar 2 money, or an independent body like the OBR with real sign-off power. A ministerial promise is exactly what a future Chancellor under spending pressure will break.
How Jar 2 could work in practice
- The Regulated Asset Base (RAB) model used for Sizewell C is a live example. Consumers start contributing during construction instead of only after completion. That lowers the risk for investors and cuts the interest that builds up during a long construction period, a cost that Hinkley Point C's developer had to carry under its contract-for-difference model.
- RAB turns a decade or more of construction risk into something closer to a regulated utility return. That is the kind of steady asset that patient, state-linked capital, such as sovereign wealth funds and pension infrastructure vehicles, is suited to hold.
Uncertainty about which technology wins (a separate question from financing)
- Once a technology is proven, such as the Rolls-Royce SMR, a pressurised water reactor using well-understood physics, financing is a straightforward discounted cash flow problem: find the cheapest capital and choose a repayment schedule.
- Where the winning technology is not yet known, as with fusion (HTS tokamaks, field-reversed configurations and laser inertial confinement are all in contention), the right approach is a portfolio. Fund several different bets in parallel, release money only as each hits technical milestones (as ARPA-E and the US Department of Energy's Reactor Pilot Program do), and move to single-commitment, RAB-style financing only once the question has narrowed from "does this work at all" to "can it be built on time and budget".
- In practice, a state should not pick a single national fusion champion while the physics is still open, even though the eventual numbers for the winning design will look very attractive once it is known.
The classification test (version 2): two separate tests
Early drafts mixed up "is this capital?" with "is this strategic?". That reopens exactly the loophole that killed the Brown-era golden rule, because any current spending can be dressed up as "creating jobs in a strategic sector". The two questions are now separate tests.
Test 1: the universal capital test (every pound, every sector, no exceptions). Does this spending create a lasting asset (physical infrastructure, IP or R&D output, or measurable human capital) that keeps producing a return in productivity, income or avoided costs for 10 years or more after the spending stops? If yes, it is eligible for Jar 2. If the benefit disappears when the funding stops, it belongs in Jar 1, whatever the sector.
Test 2: the 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 gets ring-fenced status and is defended first when trade-offs bite. Capital spending outside the key sectors, such as a new hospital wing, still belongs in Jar 2. It just doesn't get the extra protection.
This closes the loophole, since a wage subsidy in a favoured sector still fails Test 1, and it avoids wrongly excluding real capital projects that sit outside the strategic list.
Governance requirement: Test 1 needs an independent arbiter, like the OBR, rather than self-certification by the department doing the spending. Departments marking their own homework on "does this count as investment" is how the original golden rule was gamed. (Open tension: this moves real power from elected government to an unelected body. Noted, not resolved.)
Keep the strategic sector list short, on purpose
Once Test 2 exists, there will be pressure to keep adding sectors to it, spreading the money thinly across all of them. Resist this. Concentration beats spread, for two separate reasons.
- Critical mass. Strategic sectors need enough firms, suppliers and skilled workers in one place before the spillovers sustain themselves. Split the same money across ten sectors and you will probably get ten clusters that never reach that point, instead of two or three that do. South Korea's concentrated bets, set against Latin American import substitution spread across too many sectors at once, is the textbook contrast.
- Government attention is scarcer than money. A government can plausibly keep two or three sectors as cross-party priorities for 20 years. It can't do that for ten. Industrial policy fails far more often through inconsistency across elections than through a lack of initial funding.
The counter-risk: concentration cuts both ways. Fewer bets means each wrong bet costs more (British Leyland, Concorde). A short list chosen arbitrarily is as fragile as a long, diffuse one. The discipline has to be in the selection criteria (a real or credible UK edge, a strategic or security dimension beyond commercial return, and the ability to pull other sectors up with it), not only in the length of the list.
Two different kinds of "diversification". Diversifying how to win once you have chosen where to compete, such as funding several fusion designs in parallel (see the technology section above), is cheap insurance against backing the wrong approach. Diversifying where to compete in the first place, by spreading the sector choice itself thin, is the failure described above. Don't confuse the two.
Evidence for candidate key industries: UK university spinout data
Rather than picking sectors in the abstract, start from where UK research has already produced commercial results that others have paid for. Academic subject rankings such as QS and THE are a different and mostly unhelpful signal here. Oxford topping world rankings in anatomy and physiology, anthropology, geography and modern languages, or Cambridge in archaeology and English literature, says nothing about exportable industrial technology. Spinout commercial value is the measure that answers the sector question.
The data. The Royal Academy of Engineering's Spotlight on Spinouts 2026 puts the combined value of UK university spinouts at £49bn, 2.8 times its 2020 level. Venture capital raised by spinouts in 2025 fell to £1.3bn, the lowest since 2021, and most spinout value sits in three areas: life sciences (biotech, pharma and medtech), AI, and quantum and semiconductors. Two of Europe's six spinout exits worth over $1bn in 2025 came from Oxford: OrganOx, a medical device company, at about $1.5bn, and Oxford Ionics, a quantum computing company, at about $1.1bn.3
This overlaps with the government's Modern Industrial Strategy, but that strategy names eight priority sectors (the "IS-8"): advanced manufacturing, clean energy, creative industries, defence, digital and technologies (which covers AI and quantum), financial services, life sciences, and professional and business services.4 Eight sectors covering a large share of the economy is already the kind of thin spread warned against above. On this paper's logic, the protected Jar 2 list should be much shorter than the IS-8.
A regional note: this is not only a story about Oxford, Cambridge and London. There is real depth in Bristol (AI chips, where Graphcore was founded), Manchester (engineering biology and enzyme design), Newcastle (synthetic DNA manufacturing) and Glasgow (quantum devices).
Additionality: why "follow the money" isn't enough
Commercial success is a necessary check but not enough on its own to decide where state capital belongs. The question that matters is additionality. Does state money change the outcome, or does it subsidise something private investors would have funded anyway, and possibly crowd them out?
There are two different cases, needing two different roles for state capital, not one flat sector list:
- Tier A: proven winners (AI, quantum, life sciences). Private capital already works well here. Venture capital needs the chance of very large, fast-growing returns within a fund life of about 7 to 10 years, and these sectors fit that pattern: Oxford Ionics and OrganOx both sold within a normal VC timeframe. State money chasing the same deals adds little and risks crowding out a working market. The state's job here is narrower. It is to fix the UK's specific weak point, the scale-up gap, where promising spinouts are bought early or list abroad instead of growing into large UK companies. (UK spinout IPOs have slowed since their 2021-22 peak, while acquisitions continue.) That calls for patient growth-stage capital and incentives to list at home, not funding for the underlying science.
- Tier B: additionality bets (energy, agriculture). These fail the VC test on every count. Payback takes decades, far beyond any private fund's life (the same cost-of-capital logic set out for fusion above). Returns are steady and modest rather than a few huge winners. Most importantly, much of the value is an externality: energy security, food security, and resilience against coercion or shocks, rather than profit a private investor can capture. A rational private investor will give that externality too little weight because they can't collect it. The state can, and should. This is the tier the RAB and patient sovereign capital case earlier in this paper was built for. Here the state needs to be the main source of capital, not a gap-filler.
What this means for the strategic sector list (Test 2): don't build one undifferentiated list. Tag each candidate sector by tier. The tier decides both the type of intervention (main funder or gap-filler) and the argument you would make to a sceptical Treasury (additionality and externalities for Tier B; scale-up and retention for Tier A). These are different arguments, and mixing them weakens both.
Each tier needs its own tools
Because the market failure differs by tier, the policy tools have to differ too. A simple check: move a tool to the other tier and see if it still makes sense. If it doesn't, the two tiers really are different interventions, not two versions of the same one. A clawback on companies that move abroad does nothing in Tier B, where private capital never shows up. A guaranteed revenue stream is meaningless for a Tier A software company being sold.
Tier A (AI, quantum, life sciences): the problem is keeping companies and helping them grow, not a shortage of capital. - Growth-stage co-investment alongside private VC and private equity at Series C and later (through vehicles like the British Growth Partnership), so companies aren't pushed into an early sale just to get cash. - Tax treatment tied to staying in the UK: R&D credits or capital gains relief clawed back if the headquarters or IP moves abroad within a set period. - Public procurement as anchor demand in defence, the NHS and security-cleared work. Domestic contracts give companies a route to scale without needing a foreign buyer for market access, and cost little compared with direct investment. - Pension investment reform along the lines of the Mansion House Accord (see below), directing a small share of private long-term capital to the growth-stage gap rather than the state having to be the investor itself. The UK gap is large. Onward found in 2023 that pensions supply about 10% of the UK venture capital pool, against 72% in the US, and that UK pension funds invest about 15 times less in start-ups than Canada's, 9 times less than America's and 4 times less than Australia's.5 - Listing reform to make a London listing cost- and tax-competitive with Nasdaq, so listing abroad isn't the default.
Tier B (energy, agriculture): the problem is a mismatch of time horizon and externalities, not retention. - Revenue certainty: RAB or contract-for-difference price guarantees that turn an uncertain, externality-heavy return into the steady, utility-like cashflow that long-term capital such as pension infrastructure funds wants to hold. Structured well, this is a fair exchange, not charity. - First-loss or blended capital: the state takes the riskiest slice so that private investors will join first-of-a-kind projects they would otherwise avoid. This crowds private money in, the opposite of the Tier A crowding-out risk. - Pricing the externality directly: a formal payment for security of supply, using the same logic as the UK's electricity capacity market, so food and energy security are paid for rather than left unpriced. - Long-dated infrastructure bonds designed to match pension liabilities, as a separate instrument from ordinary government debt.
The VC funding gap: investment has recovered, domestic sourcing hasn't
UK venture capital investment reached $23.6bn in 2025, up 35% on 2024 and the first annual rise in four years.6 The total amount invested is no longer the main bottleneck it was.
Where the money comes from is the bigger problem. New UK VC funds raised only about £2bn in 2025.7 Set against the roughly £18bn of the $23.6bn invested, that means much of UK deal flow depends on foreign money rather than domestic institutions. This makes the retention and scale-up problem worse, because foreign-funded rounds bring foreign expectations about where a company should eventually list or be sold.
This is already an active area of policy. Any Generational Reset proposal should be positioned as speeding up what is under way rather than inventing it: - The Mansion House Accord of May 2025: 17 large workplace pension providers aim to put at least 10% of their main default funds into private markets, including venture capital, by 2030, with at least half of that in the UK. The pledge is voluntary and doubles the 5% target of the 2023 Mansion House Compact.8 - The British Business Bank's financial capacity has been raised to £25.6bn, and its British Growth Partnership has attracted commitments from Aegon, NatWest Cushon and London CIV.9 - Nest, the UK's largest pension scheme by membership (£62.9bn of assets in March 2026), has committed up to £1bn to VC by 2030, with £200m already invested through Schroders Capital.10
This supports the additionality case above. The British Business Bank is already the largest investor in UK venture and venture growth funds.11 The state is not deciding whether to be involved; it already is. The real question is whether today's support is designed to draw in a self-sustaining private market backed by pensions, as in the US, or will stay as permanent life support.
The US comparison needs care. American venture capital wasn't private from the start. The Small Business Investment Company programme of 1958 gave early venture funds government-backed leverage, and much of the technology VC later commercialised, including the internet, GPS and semiconductors, came out of DARPA and defence research decades earlier. The model the UK holds up as its target was itself heavily seeded by the state, which strengthens the case for the UK doing this deliberately now.
One limit money can't fix. Investing pension capital well needs experienced fund managers whose judgement and networks are built over several fund cycles. It is the same learning-curve argument made for fusion manufacturing and China's rail programme elsewhere in this paper. Fixing allocation rules doesn't create that capacity overnight. It takes time whatever the policy.
Refinement: equity versus royalties, and against a sector-specific equity ceiling
The TenU USIT Guide sets university equity at 10% or less for less IP-intensive spinouts and up to 25% for IP-intensive ones such as life sciences, and the average stake has already fallen from 21.5% to 16.1% in a year.12 That is the right direction, but the sector-specific ceiling deserves a harder look, not just a lower number.
The mechanism. A large university equity stake doesn't only collect value after success. It lowers the chance of success. Later-stage investors mark down a company with a big non-operating shareholder: it complicates option pools, dilutes future rounds and signals extraction rather than partnership. So a 25% ceiling for IP-intensive sectors isn't neutral cost recovery. It is a drag on the very outcome it claims to serve, and arguably a bigger drag in life sciences than in software, given how many funding rounds a biotech company needs before it earns anything.
The case against cutting it further (the "entrepreneurial state" argument associated with Mariana Mazzucato): the state and universities funded the riskiest early research that nobody else would. Taking almost no equity in the winners socialises the risk and privatises the reward, short-changing the public that took the risk.
Resolution: separate the instrument from the return. Equity is a poor way for universities to take a return because it is the thing that damages the company's chances. Push equity towards a flat, low level in every sector and drop the 25% sector ceiling. Let royalties on revenue carry whatever return the Mazzucato argument justifies. Royalties don't sit on the cap table, don't dilute future rounds, and grow with actual success rather than being fixed up front regardless of outcome. Some technology transfer offices, including Cambridge's, are already moving towards this kind of mixed model.
Fund technology transfer offices and incubators directly, not by taking a cut from the companies they support. They are a Tier B-type public good: the benefits spread across the whole regional pipeline and no single spinout captures them. At present they are funded with a Tier A-type tool, equity. Fund them through direct UKRI or government grants, the same way the underlying research is funded, and let each spinout's equity and royalty terms be set purely to maximise the number of spinouts and their chances of success.
Refinement: patents and IP. The problem is speed, not strength of protection
Two claims need separating. "IP protection slows spinouts down" is largely false as stated. "The process of securing IP and agreeing terms slows spinouts down" is true, and is the recognised problem: negotiating terms commonly holds spinouts up for six to 12 months.13 The fixes under way, the USIT Guide's standard "landing zones" and a separate lighter-touch guide for software, speed up the process without weakening protection. That is the right call. Weakening protection would hurt Tier B sectors most (life sciences, materials, energy technology, engineering biology), where an unpatented discovery can simply be copied by a better-funded, often foreign, incumbent. That would undercut the sovereign capability Tier B exists to build. Software and AI are the exception, which is why they already have their own lighter guide. Keep the distinction by IP intensity, and push hard on speed, for example with non-dilutive funding to cover patent costs up front and faster decisions by technology transfer offices, rather than weakening protection across the board.
Links to other pillars
This spinout, incubator and technology-transfer work isn't only Economy material. The overlaps need flagging before any pillar draft, to avoid duplication or contradiction:
- Education: academic incentives (promotion and tenure still reward publications and grant income over commercialisation) decide whether spinout activity happens at all, so equity and royalty reform is necessary but not enough on its own. PhD entrepreneurship training and technology-transfer capacity are themselves a skills pipeline, and this paper already contains a warning from that area: the collapse of the National College for High Speed Rail (see the HS2 case study below).
- Healthcare: life sciences and medtech spinouts such as OrganOx feed directly into NHS adoption of new technology. The Tier A "procurement as anchor demand" tool is largely an NHS commitment.
- Defence: AI, quantum, robotics, space and photonics spinouts are largely dual-use, so the same procurement tool is also a Defence commitment. The National Security and Investment Act 202114 already lets government review or block foreign takeovers in sensitive sectors, including AI, quantum, synthetic biology, satellites, advanced materials and semiconductors. That is a stronger, statutory version of the retention clawback proposed above. The reform question is whether it is applied firmly enough to takeovers of early-stage spinouts, not whether a new mechanism is needed.
- Agriculture: UKRI's strategy explicitly targets robotics, AI-enabled systems and engineering biology for farming, so Tier B agritech spinouts belong partly to the Agriculture pillar. Check for overlap with Agriculture's existing work on robotics and precision farming.
- Public Office Covenant: the governance questions running through this paper (an OBR-style independent arbiter, funding directly rather than by extraction, and a lock on continuity) are as much institutional design questions for the Public Office Covenant as economic ones. This paper may need joint ownership rather than sitting under Economy alone.
An unplaced retail idea: "UK Growth Investments" (UGI)
Not yet placed within the Tier A/B structure. Needs its own placement decision later.
Where the idea came from, and a correction. It was prompted by a belief that Chancellor Reeves had tied ISA tax relief to London-listed shares. That isn't what happened. The Autumn 2025 Budget cut the Cash ISA allowance for under-65s from £20,000 to £12,000 from April 2027, and HMRC proposed in June 2026 a 22% charge on interest earned on cash held inside Stocks and Shares ISAs.15 Both are meant to push savers from cash into investments in general, not into UK-listed assets in particular. The belief may have come from the "British ISA" proposed in 2024 and dropped later that year. Check again before citing any of this as a precedent in a pillar draft.
The core design tension. Growth and VC returns are variable and dominated by a few big winners: most investments fail and a few pay for the rest. Promising a fixed return on a vehicle that invests in growth companies is a contradiction unless someone else absorbs the swings. There's no way round this. It is the same "who holds the risk" question as the RAB and first-loss discussions above.
Two workable designs, not a combination of fixed return and growth exposure: 1. A state-guaranteed floor, run by NS&I. National Savings & Investments is already the UK's trusted, state-backed home for guaranteed-return savings (Premium Bonds, Guaranteed Growth Bonds), so it is the natural place for this rather than a new institution. The state absorbs the downside to offer a real floor, and invests the pooled money through British Business Bank and British Growth Partnership channels (see the VC section above). This is a sizeable contingent liability and has to be disclosed and sized openly, not hidden behind the word "fixed". 2. A tiered, first-loss structure for retail savers. This reuses the Tier B blended-capital tool: the state or the British Business Bank takes the first-loss slice, and retail savers hold a senior slice that behaves close to a fixed return in exchange for giving up most of the upside. The new element here is taking a tool used only by institutions (first-loss capital) and opening it to ordinary savers.
Existing UK precedent, in part. Venture Capital Trusts and the Enterprise Investment Schemes already give retail investors large tax reliefs for backing early-stage UK companies: 30% income tax relief for EIS, 50% for SEIS and, from April 2026, 20% for VCTs (down from 30%), plus capital gains benefits.15 These carry real risk and no fixed return. The relief compensates for the risk rather than disguising it. They are also fairly niche and complex compared with an ISA. If UGI goes ahead, its main contribution may be a simple mass-market version of something that already exists rather than a wholly new product.
Governance warning, in line with this paper's HS2 and North Sea lessons. A "fixed return" that later proves unaffordable and is quietly cut under fiscal pressure is exactly the broken-promise pattern warned against above. Any guaranteed floor needs the same continuity lock as Jar 2: written into law, not just promised.
Worked case study: HS2
Test 1: passes easily. Physical rail infrastructure that lasts for decades is clearly Jar 2. Classification was never HS2's problem.
Test 2: doesn't automatically earn top-tier protection. HS2 is connectivity infrastructure that helps whichever sectors sit along its route. "It enables growth generally" is exactly the vague, sector-free justification the short list is meant to screen out. It should stand on its own transport business case, not borrow standing from a protected sector list.
The real lesson is a failure of governance and continuity, not of classification. In the early 2010s the full London–Manchester–Leeds network was estimated at around £33bn. In May 2026 the Department for Transport put the cost of completing London to Birmingham alone at £87.7bn to £102.7bn, with first services between Old Oak Common and Birmingham in 2036-39 and the link to Euston in 2040-43.16 That is roughly three times the cost for a much smaller railway. Most of the eastern leg to Leeds was cut in 2021 and the western leg to Manchester in 2023. The May 2026 reset cut the design speed from 360 to 320 km/h and dropped automatic train operation.16 Each government found it politically survivable to cut off another piece. That is the failure the Jar 2 lock (a supermajority or independent sign-off, see the North Sea section above) is meant to prevent, applied here to scope rather than to raiding the fund.
Three contributing causes are worth carrying into the design of any future fusion or strategic-asset programme: - Building before the design was finished. Main construction began in 2020 before the design and consents were complete, which HS2's chief executive has since blamed for inefficient working. - Optimism bias in the original business case. This is the general megaproject pattern of over budget, over time and under benefits, not a peculiarly British problem. - Defensive contractor pricing. Risk was pushed onto contractors, who then priced their bids defensively once they held cost-overrun risk they hadn't caused.
The NAO has found that most of the cost increases come from changes of scope, inefficient delivery and underestimating how long the work would take.16
Skills and IP legacy: the clearest example of the continuity problem. The National College for High Speed Rail opened in 2017 with room for over 3,000 learners. Its Doncaster site cost about £26m, and the two sites together about £40m. It was explicitly meant to leave "a lasting skills legacy". By the time it stopped teaching in July 2023 it had 49 students and 119 apprentices, about 2% of capacity, and a £2.8m loan from HS2 Ltd was written off.17 Doncaster's Chamber of Commerce blamed the government's "flip-flopping" on scope, which meant employers couldn't commit apprentices with any confidence that contracts would still exist when they qualified.18 The one part of the project built specifically to create an exportable human-capital return collapsed because of the same continuity failure that cut the physical scope.
Three national models for getting technology and skills out of big projects
This bears on whether a future strategic bet, fusion or anything else, should be designed from the start to leave an exportable technology and skills legacy, not just the physical asset.
- China (forced transfer, then home-grown designs, then exports). Technology transfer was made a condition of market access through joint ventures with Kawasaki, Alstom, Siemens and Bombardier. China then developed its own designs (the Fuxing CR400) and exported them, for example on the Jakarta–Bandung high-speed line. This needs a home market big enough to force concessions from suppliers, and a programme big enough to drive a real manufacturing learning curve: China's high-speed network reached about 50,000 km by the end of 2025.19 The UK has neither lever at HS2's scale.
- Japan (home-grown, with a patient payoff over decades). The Shinkansen opened in 1964, 17 years before France's TGV. It was built largely from domestic engineering capacity, notably aircraft and precision engineers redirected after the post-war limits on Japanese aircraft manufacturing, as an entirely new standard-gauge network rather than an upgrade of existing lines. It was framed from the start as building national technological capability, not just a transport service. The export return was real but took decades: Taiwan's high-speed line (2007) runs Shinkansen-derived trains, and India's Mumbai–Ahmedabad line is being built with Shinkansen technology. What made it work was the consistency of Japanese governments over decades, the same continuity HS2 lacked.
- UK (HS2). Neither the leverage to force technology transfer nor sustained continuity over decades. It treated the skills and IP legacy as a side benefit (a training college) rather than a protected core objective, and that is why it collapsed.
What this means for any future strategic bet: if an export-grade technology and skills return is wanted, not just the physical asset, it has to be an explicit objective, locked into contracts from the start and protected by the same continuity mechanism proposed for Jar 2. It won't simply fall out of the main project.
Open question: still needs dedicated work
What goes in Jar 1 and what goes in 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: school buildings are clearly Jar 2, but teacher salaries produce the actual human capital while looking like Jar 1. - Defence: equipment and infrastructure versus personnel. A defence hawk treats readiness itself as the asset; a fiscal conservative counts only hardware as capital. - NHS: the capital estate versus staffing. Preventive care has a stronger investment case than most current spending, because it reduces future acute costs. - R&D: current spending in form, capital spending in function. - Welfare versus active labour market programmes: pure income support is clearly Jar 1, while retraining and childcare that enables work have a real investment logic.
The pattern across all of these is that the argument is rarely about the technical definition of investment. It is about whose favoured spending gets to claim Jar 2's protected status. This needs its own piece of work: map the plausible classification schemes for each case, identify the real party-political dividing lines, and test whether Test 1's independent arbiter 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