Snorkel AI has closed a $350M Series E at a $3.5B valuation, a little over 17 months after its previous round priced the company at $1.3B.
Insight Partners and S32 led the raise, with Addition, Lightspeed, Greylock, GV and Wells Fargo returning. The startup sells finished datasets and simulated reinforcement learning environments to AI labs and large enterprises, an offering it calls data-as-a-service.
The shift mattered. Labelling automation was the original product. Delivering finished datasets came later, and that business blends machine generation with human specialists. Annualised revenue now runs at $375M, an eighteenfold jump in a year.
Rivals in the same trade are growing just as fast. Mercor reports $2B in gross annualised revenue. Handshake crossed $1B earlier this year. Micro1 has reached $500M.
Those headline figures deserve context. Human-data marketplaces hand 60 to 70 percent of top-line income to the specialists doing the work, so net revenue runs far lower. Snorkel argues its numbers are cleaner because expert payments sit in cost of goods sold.
Chief executive Alex Ratner started the company in 2019 out of research at a Stanford AI lab.
