Snorkel books a $375m run rate a year after switching from tooling to finished datasets
The $350m Series E values Snorkel AI at $3.5b, roughly nine times the annualized revenue it reports. What a customer buys now is a written evaluation rubric, a graded dataset and a sandbox to train in.
The Product Desk
What happened
- Snorkel AI disclosed a $350 million Series E at a $3.5 billion valuation, led by Insight Partners and S32, with GV, Addition, Lightspeed, Greylock and Wells Fargo among the participants.
- The price is nearly triple the $1.3 billion valuation the company carried 17 months ago, when it raised $100 million in a Series D.
- TechCrunch places peer data suppliers on a similar curve, with Mercor at $2 billion gross annualized revenue, Handshake past $1 billion this year and Micro1 at $500 million.
Why it matters
- What it costs cost A customer buying graded data is paying for expert time it no longer manages, and because that time is booked as cost of goods sold, the cost per verified task sits outside the reported run rate.
- Who has to decide decision Teams that bought labeling automation to keep data work in-house now have to decide whether to keep running that pipeline or take delivery of finished datasets and rubrics written elsewhere.
- Who's exposed exposure When the vendor owns the rubric and resolves reviewer disagreements, the customer's working definition of a correct answer lives outside the customer's own building.
- What doesn't add up contradiction TechCrunch treats 60-70% specialist payouts as the reason peer revenue figures overstate the business; Snorkel's answer moves the same payments to a different line of the income statement without changing who gets paid.
Next story · Invest · Sep 22
Snorkel sells about a tenth of itself for $350 million at a $3.5 billion valuationConfirmed3 publishers
Two human reviewers score the same model answer differently. In Snorkel AI's account of its reinforcement learning work, that gap usually means the written evaluation criteria are inconsistent, so the criteria get rewritten [10]. The rubrics can run several pages. For a programming task, the guidance has to cover every cybersecurity and performance requirement the generated code must meet [9].
That document is now part of what the customer buys. Snorkel's first product was Snorkel Flow, software that cut the labeling work in supervised learning projects using statistical methods its founders developed at the Stanford AI Lab [4]. Last year the company began selling ready-to-use datasets instead, alongside the rubrics and the virtual environments a model trains in, such as a simulated developer workstation for a code generation model [5][11]. Tens of thousands of human experts generate the reinforcement learning tasks [8]. TechCrunch reports the approach is hybrid, with Snorkel's own software and models producing data synthetically next to the subject matter experts [16].
Investors put $3.5b on a $375m annualized run rate, about 9.3 times revenue [1][18]. "Since launching our new data-as-a-service offering nearly a year ago, we've grown over 18 times, and this week crossed an annualized revenue run rate of $375 million," co-founder and chief executive Alex Ratner said in a blog post [6]. Seventeen months ago, a $100m Series D valued the company at $1.3b [12].
Anyone pricing this category has to ask where the labor shows up. Mercor's gross annualized revenue has reached $2b, Handshake passed $1b earlier this year, and TechCrunch reported Micro1 at $500m [13]. Those companies hand roughly 60% to 70% of top-line income straight to the domain specialists. Net revenue lands well below the headline [14]. Run that range against $375m and $225m to $263m would go out to experts, leaving $113m to $150m [19]. Snorkel told TechCrunch that it sells environments and completed datasets, and that it does not sell human labor, so expert payments sit in cost of goods sold instead of coming off the reported run rate [15]. TechCrunch's report does not include the size of those payments [7].
Teams buying this tell themselves they are buying data, and data is inputs. They are also moving the definition of a correct answer to a supplier. Does the evaluation rubric leave with the vendor when the contract ends, and can your own staff reproduce a score on a sample the vendor already graded? If the answer to both is no, the rubric is the product you rented, and next year's renewal is priced against a standard you cannot rebuild in-house.
Snorkel says the new money goes to hiring engineers, AI safety initiatives and support for open-source model evaluation benchmarks [17].
What to watch
- Whether Snorkel publishes a gross margin now that expert payments sit in cost of goods sold.
- Whether the open-source evaluation benchmarks Snorkel says it will fund overlap with the rubrics it charges customers for.
- Whether Mercor, Handshake or Micro1 begin reporting net rather than gross annualized revenue.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence52
- Adoption48
- Hype gap+24
- Incentives80
- Confidence60
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Snorkel AI Inc. disclosed a $350 million Series E funding round at a $3.5 billion valuation.
- [2]
Insight Partners and S32 led the Series E; existing investors including Addition, Lightspeed, Greylock, GV and Wells Fargo also participated, with more than a half-dozen other backers joining.
- [3]
Snorkel AI was founded in 2019 by researchers from the Stanford AI Lab, and launched commercially in 2019 after four years of research by co-founder and CEO Alex Ratner and his team.
- [4]
Snorkel AI's first product was Snorkel Flow, a software platform that reduced the work involved in supervised learning projects by automating labeling using statistical methods developed by the company's founders at Stanford.
- [5]
Last year Snorkel AI changed its business model, pivoting from selling software that helps developers create training data to providing ready-to-use training datasets, and expanded beyond supervised learning into reinforcement learning.
- [6]
"Since launching our new data-as-a-service offering nearly a year ago, we've grown over 18 times, and this week crossed an annualized revenue run rate of $375 million," co-founder and Chief Executive Officer Alex Ratner detailed in a blog post.
- [7]
TechCrunch's report does not state the size of Snorkel's payments to its human experts or its gross margin.
- [8]
Snorkel AI relies on tens of thousands of human experts to generate reinforcement learning training tasks.
- [9]
Human reviewers check model work against predefined evaluation criteria that can span several pages; for a programming task the guidance must cover all the cybersecurity and performance requirements that AI-generated code must meet.
- [10]
Snorkel AI develops evaluation rubrics for customers and improves them using reviewer feedback; when two human reviewers score the same response differently, the difference usually points to an inconsistency in the underlying evaluation criteria.
- [11]
Snorkel AI provides training sandboxes alongside its datasets and evaluation rubrics; a code generation model might require a simulated version of a developer workstation.
- [12]
The new round values the seven-year-old startup at nearly triple the $1.3 billion valuation it had when it raised $100 million in a Series D 17 months ago.
- [13]
Mercor's gross annualized revenue has climbed to $2 billion, Handshake hit the $1 billion milestone earlier this year, and TechCrunch reported that Micro1 has scaled to $500 million.
- [14]
According to TechCrunch, these AI data companies pay out roughly 60% to 70% of their top-line income directly to the domain specialists doing the work, so their net annual revenue is substantially lower than the headline gross figures.
- [15]
Because Snorkel sells reinforcement learning environments and complete datasets rather than human labor, payments to its human experts are accounted for in cost of goods sold rather than in its headline annualized revenue number, according to the company.
- [16]
Rather than operating purely as a human expert marketplace, Snorkel uses a hybrid approach, generating data synthetically with its software and models alongside subject matter experts.
- [17]
Snorkel AI will use the new funding to hire more engineers, and plans to invest in AI safety initiatives and support development of open-source model evaluation benchmarks.
- [18]
A $3.5 billion valuation on a $375 million annualized revenue run rate is about 9.3 times revenue.
- [19]
If Snorkel paid experts at the 60% to 70% of top-line rate TechCrunch attributes to peer data companies, $225 million to $263 million of a $375 million run rate would go to experts, leaving $113 million to $150 million.
Sources
2 independent publishers whose own reporting we read for this story.
- siliconangle.comTraining data provider Snorkel AI raises $350M at $3.5B valuation
2 articles · September 22, 2026
- techcrunch.comSnorkel AI triples valuation to $3.5B as demand for AI training data booms
2 articles · September 22, 2026