The round
Snorkel AI announced a $350m Series E on Tuesday at a $3.5bn valuation. Insight Partners and S32 led the round, with participation from Third Point Ventures, March Capital, Blumberg Capital, Allegis Capital, Standard, Frontline and existing backers including Addition, Lightspeed, Greylock, GV, Prosperity7, Walden Catalyst and Wells Fargo.
That is close to three times the $1.3bn valuation the company carried after its Series D, which TechCrunch reported was 17 months earlier.
What changed in the business
Snorkel spent most of its life selling software: tools that let a company label its own training data programmatically, out of research that began at Stanford. About a year ago it started selling the output instead, under an expert data-as-a-service offering that combines synthetic data generation with subject-matter experts.
The company says that line has grown more than eighteenfold and crossed a $375m annualised revenue run rate in the week of the announcement. Chief executive Alex Ratner framed the demand in the release: “The teams pushing the frontier want a research data partner who pioneers the science of data development.”
Why the money is here
The constraint on frontier models has moved. Compute is expensive but purchasable; the scarce input is data that is specific, hard and correctly judged — the medical reasoning, the legal argument, the failing test case, the environment an agent can be graded in.
That has made the suppliers of that data into one of the fastest-growing categories in the sector. Lonne Jaffe of Insight Partners said in the release that Snorkel’s “research-grade approach to AI data, environments, and measurement is becoming an increasingly important ingredient in building capable and reliable AI systems.” Snorkel says it now works with frontier labs, hyperscalers, enterprises and US government agencies.
The pattern in the sector
Snorkel is not alone in finding that selling the service beats selling the tool. The same move — from software licence to delivered dataset, staffed by vetted experts — has driven the growth of several data suppliers over the past two years, and it carries the same structural risk: revenue is concentrated in a small number of labs whose spending is itself dependent on the funding cycle.
What to watch
Whether that concentration shows up. A $375m run rate built on a handful of frontier customers is a different asset from the same number spread across enterprise buyers, and the difference will only be visible if one of those customers cuts a contract.
