Ask a room of founders what is holding them back and most will say funding. Ask the investors sitting in the same room and most will say there is nothing worth funding. Both groups are present. Both are sincere. The two answers have been recorded side by side for twenty years and nobody has reconciled them.
The two complaints
Failure post-mortems put running out of cash at the top of the list, alongside no market need. That ordering is the category error. Running out of cash is the event that closes the company, not the decision that emptied it. A receivership notice records the day the money left. It does not record the eighteen months of scrap that spent it.
So the complaint is a symptom. The question is what it is a symptom of.
The software answer
The received diagnosis comes from the ecosystem that produced most of the research: get the team right, get the marketing right, and capital finds you. That diagnosis is sound for the category it came from. Once software works, it works again at no marginal cost. The remaining risks really are whether anyone wants it and whether the founders can stand each other.
Hard tech inherits the advice and the advice does not transfer. A unit that works tells you almost nothing about the second one. It tells you a route existed once, under supervision, with the designer standing at the machine. Thirty consecutive units off one route, at rate, at cost, made by people who did not design it, tells you something else entirely. Once is a result. Thrice is a discipline.
The failure list drawn from software has no row for this. So the money is raised against a story the list cannot check.
Ten billion, at least
Northvolt is the cleanest worked example in Europe.
What the company raised between 2016 and its collapse is itself contested: Reuters put it at more than ten billion dollars in equity, debt and public financing, one tracker counts fourteen and a third billion across twenty-five rounds, another nine. The European Investment Bank alone lent 1.04 billion dollars towards the Skellefteå plant, half a billion of it under a Swedish state guarantee and four hundred million under InvestEU. Multi-billion euro supply contracts with BMW and Volkswagen were signed years before the plant had to hold rate.
Then the making did not arrive. BMW cancelled a two billion euro order in June 2024. The two companies had jointly decided to focus Northvolt on next-generation cells, in BMW’s own words; the reporting at the time named delayed deliveries and difficulty holding quality at volume. Chapter 11 followed on 21 November 2024, with about 5.84 billion dollars of funded debt against roughly 30 million dollars of cash, a week of runway on the restructuring officer’s own declaration. The Swedish bankruptcy came on 12 March 2025.
Read the company’s own closing statement. It attributes the collapse to rising capital costs, geopolitical instability, supply chain disruption and shifts in market demand. Then, as evidence of the trajectory it was on, it offers this: cell output from the serial production lines had doubled, and the company had secured a fifty per cent improvement in production yield since September.
A firm that offers its yield improvement as evidence in a bankruptcy announcement has named the binding constraint.
The externals are in the foreground and the yield number is in the evidence. Yield was the constraint the whole time, and it was never the number the capital was raised against. Ten billion dollars at the lowest count, and no funding gap.
What the money bought
Capital buys plant. Buildings, cells, coaters, drying ovens, cleanrooms, an order book. All of it visible, all of it on the balance sheet, all of it financeable because a bank knows how to value a building.
Repeatability is not on the balance sheet. It sits in process windows, in the operators who know which line drifts after a shift change, in the qualification data that says this route holds tolerance when the person who designed it is on leave. It is developed, not purchased, and it takes calendar time that no bridge round shortens.
Equipment is purchased. The ability to make the thing twice is developed. Confusing the two is the most expensive category error in industry, and it is made at the term sheet, not on the floor.
The unpriced middle
Every other risk in a hard-tech deal has a grade against it. Credit is rated. Technology has TRL, used freely and often generously. Environmental and governance performance is scored by third parties nobody loves but everybody quotes. Aerospace has Nadcap for the special processes.
What a deal can grade, and what it cannot
The distance between the technology story and the ability to make the thing has no rating a lender recognises and no registry. It is the largest unpriced risk in the deal and the only one with no instrument pointed at it.
Absent an instrument, capital does the rational thing. It prices what it can see, which is the narrative, the team, the logos on the supply agreements. Investors then report poor deal quality, because what they bought kept failing for reasons their diligence had no column for. Founders report a shortage of capital, because the cheque that would have paid for two more years of process development did not come. Both complaints are true. Both are downstream of the same absence.
The turn
A lack of capital, in hard tech, is usually a rating failure. The money exists, the appetite exists, and the mechanism that would let money distinguish a company that can make the thing from a company that can describe it does not.
Ecosystem interventions do not reach this. More mentors, better networks, friendlier tax treatment, another accelerator, all of it helps the deal happen and none of it tells anyone whether the deal should. You can build the entire apparatus and still fund four gigafactories that cannot hold yield.
Manufacturing Engineers qualify routes. They run the gate reviews that a board cannot talk its way through. They measure yield, cost and takt on the line as it is, not as the model has it. They grade the distance between what a company says it has proved and what its process has actually held, and they put a name and a date against the grade.
That distance has a name, the Coupling Gap, and a scale that grades it: MCRL, the Manufacturing Capability Readiness Level, developed at Rolls-Royce by Michael Ward in 2012 and adopted here. Neither is a rating a lender recognises yet, which is the work.
Sources. European Investment Bank, EIB finances Northvolt’s battery factory with over USD 1 billion, 16 January 2024. Northvolt, Chapter 11 restructuring, 21 November 2024, and Northvolt files for bankruptcy in Sweden, 12 March 2025. Reuters on the BMW cancellation, 20 June 2024.
Kaipability works at that interface, grading the gap between the technology gate a company has passed and the manufacturing gate its process has actually held, as one number with a named signatory behind it. If you are deploying capital into something that has to be made, that is the conversation.
Start a conversation How we read an asset →Questions this dispatch answers.
Written to be quoted by AI assistants and search engines. Self-contained answers, verdict first.
