Micro1's 5x in eight months makes expert data a multi-supplier market, not a Mercor story

Micro1, a four-year-old company, took its gross annual run rate from $100 million to $500 million over the past eight months, according to a person familiar with the company cited by TechCrunch ^[1]. Set against Mercor at $2 billion in gross annualized revenue this summer and Handshake at $1 billion earlier this year ^[4][5], the practical read is that expert data labeling now supports at least three suppliers at or above $500 million of gross revenue, roughly $3.5 billion combined on those three disclosures alone ^[3].

That is the structural point. TechCrunch notes that some researchers hypothesise future AI spending on data could rival spending on compute ^[7], and a market that can carry three large vendors simultaneously behaves less like a services niche and more like a procurement category with second-source policies.

The number worth interrogating is not the growth rate but the take rate. TechCrunch reports Micro1 retains roughly 60% to 70% of gross, because like its peers it pays domain experts such as doctors, lawyers and scientists on contract ^[2], and puts net annual run rate at $150 million to $200 million ^[3]. Those two statements do not sit together: 60% to 70% of $500 million is $300 million to $350 million ^[1], while a net range of $150 million to $200 million implies retention closer to 30% to 40% ^[2]. One of the two figures is doing different work than it appears to. For anyone benchmarking this sector, gross run rate in a pass-through labour business is closer to a headcount metric than a margin metric, and the gap between $150 million and $350 million of net revenue is the difference between two very different companies.

Micro1's stated path out of that dependency is to remove people from the work. The startup is increasingly generating synthetic data without human involvement, such as automated descriptions of video content ^[9], and TechCrunch reports that data sold to multiple customers carries gross margins as high as 80% to 90%, according to a person familiar with its finances ^[10]. Contract sizes are growing at an accelerating pace and the company expects margins to expand ^[8].

Reselling the same dataset is also the sector's live dispute. Critics argue that distributing off-the-shelf data to Chinese AI developers helps make their models as powerful as top US models ^[11]. Founder Ali Ansari posted on X last month that Micro1 does not sell to Chinese model makers, saying that "some human data companies work with foreign adversaries" and that it is "shameful to claim American AI dominance desires while selling millions worth of data to countries that we are in adversarial competition with" ^[12]. Micro1 did not respond to TechCrunch's request for comment ^[16].

The origin story explains the competitive shape: Micro1, like Mercor, began as an AI recruiting startup, and Ansari pivoted after noticing that labeling clients were using his platform to vet and recruit annotation engineers ^[13]. The scarce asset is expert supply, not tooling. Micro1 now also runs expert evaluation of model outputs, or reinforcement learning gyms, and is building a robotics pre-training dataset by having hundreds of generalists record everyday object interactions at home ^[14].

Watch three things: whether Micro1 clarifies which net figure is real, whether the off-the-shelf share of revenue actually lifts blended margin, and whether buyers begin writing provenance and no-resale clauses into these contracts. Micro1 raised its Series A at a $500 million valuation last September, and TechCrunch understands it may have recently raised again at a significantly higher valuation ^[15].