Micro1 hits $500M run rate: expertise is now inventory

Micro1 grew its gross annual run rate from $100 million to $500 million in eight months by paying contract doctors, lawyers, and scientists to write down how they think. It is not even the biggest company doing this. That number is worth sitting with, because it is a price tag on the exact thing most small businesses assume is not for sale: specialized domain expertise.

What actually happened

TechCrunch reported the run-rate jump at the four-year-old startup, which began life as an AI recruiting platform before pivoting into training data. Most of that gross figure goes straight back out to the contractors — TechCrunch puts the net annual run rate at $150 million to $200 million.

The margin story is in the other product line. Data generated once and sold to multiple labs carries gross margins of 80% to 90%, per the same report. Bespoke expert work pays the people; the resale catalog pays the company.

Micro1 is third in a crowded field. Mercor hit $2 billion in gross annualized revenue this summer. Handshake crossed $1 billion earlier this year. Three companies, all growing, none of them running out of buyers. The labs are not short on compute. They are short on people who can explain a diagnosis, a contract clause, or a failed experiment well enough to train against.

Why the training-data boom matters for your business

Two things follow, and they point in opposite directions.

The models are getting better at your vertical on purpose. If you run a niche business — specialty insurance, industrial parts, regulated services — you have probably told yourself the general-purpose models will never understand your corner well enough to matter. That was true when the constraint was data nobody had bothered to collect. It is less true when three companies with a combined multi-billion-dollar run rate are collecting it deliberately, by hiring people who work in your corner. Recheck the model against your hardest real cases every quarter. The answer changes.

Your process knowledge is an asset with a market rate. The judgment your senior estimator applies to a bid, the reason your ops lead reroutes a shipment — that is the same category of thing these companies buy. Most small businesses have never written it down, which means it is not documented, not transferable, and walks out the door at retirement. It also means your contractors and staff can sell their version of it on the side, and some of them already do.

The move is not to panic about either. It is to write the knowledge down for yourself first: the decision rules, the exceptions, the cases where the obvious answer is wrong. That document is what you automate against, what you train new hires on, and what you would need before you could ever license it. Encoded expertise is the asset. Everything downstream of it — the agent, the workflow, the internal tool — is just plumbing.

Source: TechCrunch — AI data startup Micro1 reaches $500M gross run rate amid AI training boom.