Snorkel AI's $3.5 Billion Valuation Outruns Its Own Revenue Math

Snorkel AI raised $350 million in a Series E round announced Sept. 22, co-led by Insight Partners and S32, the company said in its own release. The round values the seven-year-old data-labeling company at $3.5 billion, nearly triple the $1.3 billion it commanded when it raised its Series D.

That prior round, a $100 million Series D led by Addition, closed May 29, 2025, according to a Business Wire release. Announced and closed are the same event here: both filings describe a completed raise, not a committed-but-unfunded tranche.

What Insight Partners and S32 are actually buying

Snorkel's release names its buyers precisely. Lonne Jaffe, a managing director at Insight Partners, said the company's data is becoming important to building capable AI systems. Andy Harrison, chief executive and general partner of S32, said Snorkel's setup creates a loop between human experts and AI output. Existing backers Addition, Greylock, Lightspeed, GV, Prosperity7 and Wells Fargo returned; new money came from March Capital, Blumberg Capital, Allegis Capital, Frontline, Standard and Third Point Ventures.

None of that names a number the money is chasing. Founder and chief executive Alex Ratner's quote in the release talks about frontier labs wanting a research partner. The release runs eleven paragraphs of investor and platform description and does not once state Snorkel's revenue.

Where the $375 million figure actually comes from

The number that makes the valuation legible at all appeared only in Ratner's answers to TechCrunch, not in Snorkel's own filing: an annualized revenue run rate of $375 million, up eighteenfold over the prior 12 months. Divide $3.5 billion by $375 million and the round prices Snorkel at just over nine times the revenue it reported to TechCrunch, a multiple nobody in the announcement itself works out.

That 18-fold growth rate traces to a specific pivot. Snorkel launched Snorkel Expert Data-as-a-Service in September 2025, shifting from selling data-labeling software to selling finished training and evaluation datasets built with subject-matter experts, plus reinforcement-learning environments, according to the company's own Series D materials. The Series E money is earmarked for the same shift: expanding what Snorkel calls its agentic data factory and deepening an Open Benchmarks Grants program.

The revenue number nobody defines the same way

TechCrunch's reporting sets Snorkel's figure against three declared peers in the same expert-data business: Mercor, at $2 billion in gross annualized revenue; Handshake, which has crossed $1 billion; and Micro1, at $500 million in gross run rate. Measured against those, Snorkel's $375 million looks like the smallest number in the room.

That comparison does not hold up cleanly. TechCrunch's reporting notes that Mercor, Handshake and Micro1 typically pass 60 to 70 percent of their gross revenue straight through to the domain experts who do the labeling, and count that pass-through as revenue on the way in. Snorkel instead treats those expert payments as cost of goods sold, which nets them out before the top line is even reported. A company using the gross method with Snorkel's same expert costs would report a materially larger revenue figure without processing a single additional dollar of business.

Round	Date	Amount	Post-money valuation
Series D	May 29, 2025	$100 million	$1.3 billion
Series E	Sept. 22, 2026	$350 million	$3.5 billion

1. Mercor (gross)2000 $ million
2. Handshake (gross)1000 $ million
3. Micro1 (gross)500 $ million
4. Snorkel AI (net of expert pay)375 $ million

Source: TechCrunch, accessed Sept. 24, 2026

The chart above states figures exactly as each company reports them, not on a common accounting basis, which is the point: none of these four numbers was produced the same way, and a reader comparing them at face value is comparing a gross figure to a net one three separate times.

What Snorkel didn't say

Snorkel's release does not disclose what share of its own $375 million goes to the specialists who do the underlying data work, so there is no way to convert its net figure to the same gross basis as Mercor, Handshake or Micro1 to check whether the pool of underlying business Snorkel is running is actually smaller than theirs, or merely booked differently. The company also did not restate its Series D valuation multiple against the revenue it was reporting at that time, which would show whether the multiple itself expanded or only the revenue did.

Other enterprise AI vendors have run into the same scrutiny over how a headline metric is built: Stripe's internal AI adoption figure drew the same kind of questioning over what, exactly, was being measured. Snorkel's own round follows Arcee AI's Series B, another AI infrastructure round priced this month above the revenue disclosed in its own announcement.

What would settle the multiple

The next number that would clarify this is Snorkel's gross bookings figure, the total dollar volume flowing through its data factory before expert payments are subtracted, which neither the company nor its investors have published. Until that appears in a future filing or investor deck, the nine-times multiple implied by Tuesday's round rests on a revenue definition of Snorkel's own choosing, compared against three competitors using a different one. A Series F priced on the same net basis, rather than a restated gross figure, would at least confirm the multiple is not quietly expanding twice.