In February 2001, seventeen people met at a ski lodge in Utah and wrote a manifesto. It was a good manifesto. For software.
Working code over documentation. Responding to change over following a plan. Ship, learn, ship again. It suited a world where a mistake costs a rebuild and a deploy, where iteration is nearly free and almost everything is reversible. In that world the doctrine was not just sensible. It was correct.
Then it got a certification industry. Scrum masters, story points, a fortnightly liturgy of stand-ups and retrospectives. By 2014 one of the original signatories, Dave Thomas, stood up and declared the thing dead, killed by the consultants who had turned a way of working into a product to be sold.
The cheque version
Venture capital learned the same tempo and added money.
Fail fast. Pivot. Race to product-market fit. Spray small bets, double down on the survivors, mark the book up every eighteen months. This is Agile with a cheque attached, and on bits it works beautifully. Software has near-zero marginal cost, reversible decisions, and a feedback loop measured in days. Capital priced for that velocity is well matched to it.
The trouble starts when the same money, moving at the same speed, points itself at things made of metal.
The atoms
An actuator does not iterate weekly. A certification cycle does not fit a fortnight. A supply chain does not pivot on a retrospective. The physical world has its own clock, and it does not check whether your fund needs a markup this year.
We already ran the experiment. Autonomous vehicles took more than a decade and tens of billions of dollars to produce driverless service in a handful of cities. That was not a shortage of capital and it was not a shortage of cleverness. It was atoms keeping their own time. Sensors, weather, edge cases, liability, the long unglamorous grind of making a thing safe enough to leave alone. None of it sprints.
So watch what happens now in physical AI. Early robotics teams quietly rewrite the deck to sound like a foundation-model company, because that is the story the fast money understands. The gripper becomes a platform. The perception stack becomes a moat. The pitch gets tuned to the tempo of the capital rather than the tempo of the work. That pivot might close the round. It rarely builds the machine.
The structure that never left
Here is the part the current excitement keeps forgetting. The structure that actually builds physical things never adopted Agile. It did not need to.
Industrialists leading Manufacturing Engineers. Long commitment horizons. Capital that is held as patient ownership rather than rented for a fund cycle. The Mittelstand ran on it for a century. The great corporate labs planned in atom-time before venture existed as an asset class. The Wallenberg shape held assets for generations because some things only mature on a generational clock.
That structure did not die. It got drowned out by twenty years of software confidence, when the people who could exit a company in three years started to believe everyone should. What is now being marketed as patient specialist capital is a faint echo of what the family firm and the corporate lab always were. The rediscovery is welcome. The amnesia that made it necessary is the interesting bit.
The turn
So the question is not whether physical AI needs more money. It is awash in money. The question is whether the money and the method are running on the right clock.
Agile capital and Agile method are the same category error wearing two costumes. Both were tuned for a world where iteration is cheap and failure is reversible. Atoms are neither.
You cannot retro your way past a six-month tooling lead time. And you cannot LP your way to more of the rare people who can hold the whole system in their head while it misbehaves at two in the morning.
Manufacturing Engineers do not sprint. They qualify supply chains. They validate production processes. They design the feedback loops that turn first-of-a-kind data into nth-of-a-kind confidence, and they set that tempo by the physics, not the funding round.
Kaipability works at this interface, where the money meets the metal and the method meets the machine. The firms that win the physical AI cycle will be the ones that match the clock of their capital and the clock of their method to the clock of their atoms. That is a conversation worth having before the round closes, not after the prototype dies.