The Three Capitals

As of this month there are three capitals swirling around the AI economy, and the temptation is to line them up like three flavors of the same thing. Satya Nadella named the newest one on June 14, in an essay that has now passed forty million views: token capital, a firm’s own accumulated machine learning, hoarded as its new IP. The oldest one has been working for eighty years: venture capital, the engine that backs a thing before it has earned it. And the third is the word we use at the Capability Capital Institute, the one this whole project is named for: capability capital. Three capitals on the same table. The interesting fact about them is the one nobody is saying out loud. They are not parallel.

It is an easy mistake to make, because the word capital is doing three different jobs across the three terms and quietly hoping you won’t notice. In token capital it means an accumulated asset — a stock you build up. In venture capital it means a financing mechanism — money that goes looking for a return. In capability capital it means something older than both: a factor of production, the human source from which the other two are ultimately drawn. Stack three different meanings under one word and of course they look like siblings. They are not siblings. The way to tell them apart is to put the same two questions to each one. Who does it recognize? And what does it make?

Start with token capital, since it is the one in the headlines. Nadella is right about what it is. A company that runs its own evaluations, its own training loops, its own knowledge base builds an asset that compounds — every workflow improves the next one, and the advantage grows harder to copy over time. He even uses the word compounding, and he has earned it. But put the two questions to it. Who does it recognize? No one. It is a machine asset; recognition is not in its vocabulary. What does it make? More of itself. Token capital compounds, and it recognizes nobody. That is not a flaw in Nadella’s description. It is simply what the thing is — accumulation without acknowledgment.

Now venture capital, and here the picture flips. Georges Doriot built the first true venture firm in 1946 on a single radical bet: that you could recognize potential before it had proven itself, write a check on the strength of what a person might become, and be right often enough to change the world. He backed a young engineer named Ken Olsen, and Digital Equipment Corporation was the return. That is the genius of the form, and I do not say it lightly — venture capital is a recognition engine, and recognition of the unproven is one of the most generous things capital can do. But put it to the second question. What does it make? It makes a bet and then it waits. It does not form the founder; it funds him and watches. And it aims that recognition at exactly one tier of the human pyramid — the person already standing near the top, already legible as a founder. Venture capital recognizes, brilliantly. It does not form. And it has eyes for only the few.

Which brings us to the third capital, and to the only one of the three that answers both questions at once. Capability capital does what venture capital does — it recognizes a human being’s unproven potential and backs it before the market has — and it does what token capital does — it accumulates that potential into a durable, compounding asset. The mechanism is the capability account: a place where a person’s formation is recognized, recorded, and grown, the way a venture portfolio backs a founder and a learning loop accumulates IP. It recognizes a human, the way Doriot recognized Olsen. And then, unlike Doriot, it stays — it forms the person it recognized. It is the only one of the three that recognizes a human and makes one.

Make that concrete, because it is easy to wave past. Somewhere on a floor right now there is a worker who has quietly carried responsibility two levels above his title for a decade — real capability, formed by the work itself, invisible to every ledger that counts. Venture capital will never see him; he is not a founder. Token capital will see only the judgment it can harvest from him into the loop, and once it has the judgment it owes him nothing. A capability account does the thing neither will: it recognizes that decade as the asset it already is, records it in a form the worker owns and can carry, and backs the next stage of his formation against it — the way a venture firm backs a second round on the strength of the first. Recognized. Formed. Compounding. In the person.

And it runs in the other direction too, toward the worker who has not proven anything yet — which is where the venture parallel bites hardest. Doriot did not back Ken Olsen because Olsen had already built Digital Equipment; he backed him on the strength of what Olsen might build. A capability account lets you make that same wager on a twenty-three-year-old who walked onto the floor last month: recognize the potential before the résumé confirms it, stake real formation against it, and let the proof arrive later. Venture capital has always reserved that bet for the founder near the top. Token capital cannot make it at all — a machine asset has no way to believe in a person. The capability account is the first instrument that extends Doriot’s wager downward, to the floor, to the unproven many instead of the unproven few. That is what it means to recognize a human and make one.

Put the three side by side and the shape is unmistakable. Token capital compounds without recognizing anyone. Venture capital recognizes without forming anyone. Capability capital is the whole of what the other two each hold only half of — token has the compounding and not the recognition, venture has the recognition and not the formation, and capability has all of it: recognition, formation, and the compounding that follows. And it aims that whole at the place neither of the others ca