The scramble for unique training data is minting fortunes in the annotation business, and Micro1 just became the latest proof. The four-year-old startup grew its gross annual run rate from $100 million to $500 million in eight months, according to a person familiar with the company. After paying its network of contract doctors, lawyers, and scientists, Micro1 keeps roughly 60% to 70% of that figure — putting net annual run rate somewhere between $150 million and $200 million. Those are serious numbers for a company most people have never heard of.
Micro1 still trails the category leaders. Mercor reportedly hit $2 billion in gross annualized revenue this summer, and Handshake crossed $1 billion earlier this year. But the fact that three players can post numbers like these simultaneously tells you something about the demand curve. Some researchers now argue AI spending on data could eventually rival spending on compute, which would make today’s annotation boom look like the early innings.
The economics get more interesting when you look at margins. Micro1 is increasingly generating synthetic data without any human in the loop — think automated video descriptions — and selling the same datasets to multiple customers. Off-the-shelf data carries gross margins of 80% to 90%, which explains why investors are paying attention. The startup raised its Series A at a $500 million valuation last September and may have closed another round since at a significantly higher price.
That resale model comes with geopolitical baggage. Selling identical datasets to multiple buyers has drawn criticism for potentially helping Chinese AI developers close the gap with U.S. models. Micro1 founder Ali Ansari drew a hard line on X last month: “Some human data companies work with foreign adversaries. [A]nd the results show today in Kimi K3.” He called it “shameful” to claim American AI dominance while selling millions in data to adversarial countries. Whether that stance costs Micro1 business or becomes a selling point with Washington depends on how much the government decides to care about the data supply chain. Either way, Ansari’s pivot from AI recruiting to data labeling — after noticing clients using his platform to vet annotators — looks like one of the better calls in a market that’s only getting hungrier.
