Snorkel AI raises $350M Series E at $3.5B valuation as training data demand surges

Snorkel AI has raised a $350 million Series E at a $3.5 billion valuation, the company said, as AI labs and corporations bid up prices for high-quality training data. Insight Partners and S32 led the round.

The new valuation is nearly triple the $1.3 billion Snorkel commanded 17 months ago, when it raised a $100 million Series D. Existing investors Addition, Lightspeed, Greylock, GV, and Wells Fargo also put money into the new round. Snorkel launched commercially in 2019, after CEO and co-founder Alex Ratner spent four years researching the problem at a Stanford AI lab.

Snorkel originally sold software that automated data labeling, the process of tagging raw data so a model can learn from it. Last year the company shifted to selling finished datasets directly, an approach it calls data-as-a-service. Rather than running a marketplace of human labelers, Snorkel pairs its own software and models with subject matter experts, generating some training data synthetically and using people to refine and check it. Part of that business now includes reinforcement learning environments, simulated tasks that train AI systems through trial and reward, sold alongside complete datasets.

Snorkel says its annualized revenue run rate has reached $375 million, up eighteenfold over the past 12 months, a jump it attributes to growing demand from AI labs for advanced training data. Other startups in the same space report similar growth: Mercor’s gross annualized revenue has reached $2 billion, Handshake passed $1 billion earlier this year, and Micro1 has scaled to $500 million, according to prior TechCrunch reporting. Those companies typically pay 60% to 70% of their top-line revenue to the domain experts doing the labeling work, so their net revenue runs well below the headline figures. Snorkel’s accounting works differently: because it sells finished datasets and RL environments rather than human labor directly, payments to its subject matter experts are booked as cost of goods sold rather than folded into revenue, the company says, which makes its $375 million figure hard to compare directly with rivals’ gross numbers.

Snorkel has not said how it plans to spend the new funding, and it has not disclosed how much of its revenue growth comes from the newer data-as-a-service line versus its original labeling software. Whether the eighteenfold revenue jump holds pace, and whether rivals like Mercor and Handshake pursue similarly large rounds at similarly steep valuation increases, is not yet known.