• Snorkel AI raised a confirmed $350 million Series E at a $3.5 billion valuation, up from $1.3 billion at its Series D roughly 17 months prior.
• The round was co-led by Insight Partners and S32, with Wells Fargo, Lightspeed and Greylock among the participants who were already on the cap table.
• The valuation jump rests on Snorkel's own claim of 18x revenue growth to a $375 million annualized run rate since it pivoted to a 'data-as-a-service' model — a figure that is self-reported, not independently verified.
• Snorkel folds payments to its subject-matter experts into cost of goods sold rather than netting them from revenue, a structural choice that makes its headline run rate not directly comparable to marketplace rivals that pay out 60-70% of revenue to contributors.
• No named enterprise or government customers were disclosed, leaving the demand story anecdotal even as the price tag triples.

Snorkel AI has raised a $350 million Series E at a $3.5 billion valuation, co-led by Insight Partners and S32, the company said in its own announcement — nearly tripling the $1.3 billion mark it set at its Series D roughly 17 months ago. The headline number is confirmed by a primary source. The growth case behind it is not: Snorkel's claim of eighteenfold revenue growth to a $375 million annualized run rate comes from the company alone, with no independent audit disclosed.

The round, confirmed

Snorkel AI, a seven-year-old company whose technology traces to a Stanford research project, said it closed a $350 million Series E led by Insight Partners and S32, with participation from Wells Fargo, Lightspeed, Greylock, GV and other existing backers alongside new investors including Third Point. The round values the company at $3.5 billion, up from $1.3 billion at its $100 million Series D roughly 17 months earlier — a nearly threefold markup in under a year and a half.

From labeling software to a data seller

Snorkel began as software that automated data labeling for machine-learning teams. Roughly a year ago it pivoted to what it calls data-as-a-service: selling AI developers finished training datasets, produced through a hybrid of synthetic generation and human subject-matter experts, rather than tools for building datasets themselves. Ratner framed the shift as a response to model developers needing data too complex for human labor alone to supply at scale, saying, "Frontier data is no longer solvable by the optimized supply of human hours and bodies."

The growth number is Snorkel's own

Snorkel says its data-as-a-service revenue grew 18x over the past year to a $375 million annualized run rate as of the announcement. That figure is self-reported in the company's own funding post, not filed with a regulator or confirmed by a third party — standard for a private company, but a reason to treat the number as reported rather than established. Snorkel also books payments to its subject-matter experts as cost of goods sold rather than netting them from revenue, a presentation choice that inflates the top-line figure relative to marketplace rivals that pass 60-70% of revenue through to contributors, per TechCrunch's reporting.

What's still unconfirmed

Snorkel named its customer categories — frontier labs, hyperscalers, "neolabs," vertical AI companies, enterprises and U.S. government agencies — but disclosed no specific client names, contract values or gross margin. That leaves the demand story that justifies a $3.5 billion price tag resting on category labels rather than verifiable accounts, even as the round itself, the investor list and the valuation figure are on the record from the company.

The verdict

The $350 million check and $3.5 billion valuation are real and sourced. The narrative used to justify tripling that valuation in 17 months — 18x growth, a $375 million run rate — is Snorkel marking its own homework, and the COGS treatment of expert labor means the run rate isn't apples-to-apples with rivals. Investors evidently found the story convincing enough to write the check; readers should treat the growth figures as a claim awaiting independent confirmation, not as settled fact.