A drawing can be redone in eighteen months.
A works cannot be redone at all.
That is the whole argument, and most industrial strategy is written as if it were not true. A design is a file. You can send a file. If the file is wrong you open it and fix it, and the people who got it wrong the first time are still there to get it right the second. A capability is not a file. It is four hundred people across three shifts who have watched the process go bad and know, without being told, what to do in the next ten minutes. You cannot send that. You cannot buy it in a hurry. And when those people scatter, they do not come back just because the money came back.
The two readiness ladders
There are two ladders, and industry has quietly agreed to only fund one of them.
Technology readiness asks whether the thing works. It works in a cell, on a good day, with the man who designed it standing next to it and as much scrap as he cares to make. That is a real achievement. It is also where all the money goes.
Manufacturing readiness asks whether it can be made again tomorrow, at rate, at quality, by people on shift who never read the paper. Control plan. A gauge study that is not a work of fiction. First article. Capability held across a run instead of hit once and photographed. A written answer to what you do when the process drifts, because it will drift.
A demonstrator is a photograph of a capability, not a capability. Customers know this. It is why they do not buy demonstrators.
The marking
So why does the second ladder go unfunded?
Look at who is marked on what. Researchers are marked on novelty. Investors are marked on markups and the life of the fund. Programme managers are marked on programmes delivered. Ministers are marked on announcements made and ribbons cut. Every one of those lands cleanly on one bank of the river or the other.
Nobody is paid to stand in the water.
So the valley of death is not a place where something goes wrong. It is what is left over after everybody has done exactly what they were rewarded for doing. Call it a market failure and you imply a market that broke. Nothing broke. There is a hole in the organisation chart, and a hole in an organisation chart is a remarkably durable object. It has no budget line, no owner, and no annual review. Nothing erodes it.
The loop
What holds the hole open is a loop, and it is worth saying out loud, because most funding instruments are built as though it were not there.
No serious buyer designs in a part with no field data. There is no field data until somebody designs it in. Round and round. That little circuit has killed more good engineering than bad physics ever has, and nobody inside the company can break it, because the company is not the one holding the pen. Someone outside has to agree to be first.
Historically that someone was a customer. Not a grant-giver, not a fund, not a review panel. A buyer with a specification, a delivery date, and the stomach to accept that the first three units will be rubbish. That is how the crossing gets made. It is how it has been made in every industrial economy that has managed it, and there is no second example of it happening any other way.
The instrument
Which makes the choice of instrument the whole decision. It is usually made without anyone noticing a decision was available.
A grant buys a report. An offtake buys a qualified process. Same money. Different instrument. Completely different object at the end of it, because what a grant owes you is knowledge, and what an order owes you is a part that has to turn up.
Put it in accounting terms and it goes quiet very quickly. A grant creates an obligation to explain. An order creates an obligation to deliver. Only the second one drags in the qualification work, the gauge studies, the process capability evidence, the supplier development — all the boring apparatus that is the difference between an industrial capability and a good impression of one.
There is a near miss worth naming. Buying the finished part is not the same as buying the ability to make it. An order can be honoured by something designed, fabricated and assembled somewhere else entirely, and then you have bought the by-product and none of the leverage. The specification has to reach the process, not just the part.
The lanes
The choice is usually presented as a settled national characteristic. America does private enterprise. China does the state. Everyone else picks a lane and lives with it.
That is not what the record shows. It is a decision every government re-makes, and most of them have re-made it recently.
Who holds the capability, and how the state shows up
Eight positions on two questions. The vertical axis is the instrument: whether the state writes a cheque or places an order. The horizontal axis is ownership: whether the capability ends up in state hands or private ones.
Hover or tap any position to read what it commits that government to.
Dashed arrow = direction of movement since 2022 · four of eight positions have shifted
The Anchor
Buys the output
The Acquirer
Buys the asset
The Centre
Mixed instrument, mixed ownership
The Sponsor
Funds it and hopes
The Builder
Funds and owns it
Two questions decide where a country sits: who ends up holding the capability, and whether the state turns up as a funder or as a customer. All four arrows on the board were drawn in the last four years.
America was never as private as the story says. The capability sits in private hands, yes, but the crossing was paid for by a buyer. The CHIPS Act pulled that position down towards subsidy — a change of instrument, not a change of faith.
Japan held the private-capability corner for forty years. Fanuc, Keyence, Disco, Tokyo Electron. Then the state took the largest shareholding in Rapidus, with a golden share and a veto and support running into trillions of yen. That is a country changing lanes in front of everybody.
Europe has been subsidy from the beginning. Chips Act 2.0, proposed in June 2026, adds demand accelerators and innovation procurement. Brussels has read the same evidence and started climbing the vertical axis.
The Gulf is the odd one out, and the most honest about it. Sovereign wealth buys the asset instead of growing it. Capability acquired rather than accumulated. Very fast, and it leaves the tacit half behind, because the tacit half was never for sale.
Britain is in the awkward corner. The capability is assumed to be privately held; the instrument is overwhelmingly grant and equity. It is the one square on the board where nobody is confidently holding anything. The £150 million advance purchase commitment points the right way and is dwarfed by everything standing next to it.
Kaipability sits top right on purpose. Privately held capability, earned by delivering rather than allocated in a round. That is the corner the chart argues for, and the useful part is that a firm can occupy it whether or not its government ever does.
The turn
None of this is a knowledge problem. It is all written down. Most of it sits in public reviews that have been on government websites for a decade, recommending the same corrections in slightly different words, patiently, to nobody in particular.
The trouble is not that nobody knows what to do. It is that doing it means holding an obligation nobody is marked on, for longer than any fund will wait, over an asset that does not show up on a balance sheet until the day it is too late to build one.
That is the job nobody has taken. It has been open a long time. There is a name for whoever takes it: the Modern Industrialist.
Manufacturing engineers qualify processes. They write control plans and then stand on the floor and prove them while the process is being difficult. They run the first article, then the fiftieth, then the ten thousandth. They build the loop that turns scrap into knowledge instead of into a write-off. They turn an order into a capability, and a capability into a rate.
Kaipability works at that seam, where a technical claim has to turn into a delivered part and somebody has to own the ground in between. If you are carrying an obligation nobody marks you on, that is the conversation worth having.
Questions this dispatch answers.
Written to be quoted by AI assistants and search engines. Self-contained answers, verdict first.
