Snorkel AI has raised $350 million in a Series E round at a $3.5 billion valuation, a figure nearly triple the $1.3 billion the company was valued at when it last raised capital 17 months ago. The round was led by Insight Partners and S32, with participation from existing backers including Addition, Lightspeed, Greylock, GV, and Wells Fargo.
The Business Shift: From Software to Data-as-a-Service
Snorkel originally sold software for automating data labeling. Last year, the company pivoted to a different model. Instead of selling tools that help customers label their own data, Snorkel now delivers completed datasets directly. The company calls this offering data-as-a-service.
The approach is deliberately hybrid. Snorkel uses its own software and machine learning models to generate training data synthetically, then layers in subject matter experts to refine and verify the output. This distinguishes it from a pure human expert marketplace, where the platform simply matches workers with tasks and takes a cut.
The Revenue Picture and the AI Data Boom
Snorkel says its current annualized revenue run rate has reached $375 million, representing an eighteenfold increase over the past 12 months. The company attributes the growth to the intense demand from AI labs for high-quality training data at scale.
Snorkel is not the only company in this space experiencing rapid expansion. Mercor, a competitor focused on expert networks, reported gross annualized revenue of $2 billion. Handshake crossed the $1 billion milestone earlier in the year, and Micro1 scaled to $500 million, according to reporting from TechCrunch.
There is an important distinction in how these numbers are calculated. Companies like Mercor, Handshake, and Micro1 pay out roughly 60 to 70 percent of their top-line revenue directly to the domain specialists performing the work. Their headline gross figures therefore overstate the actual net revenue the company retains. Snorkel's structure is different: the company sells reinforcement learning environments and complete datasets rather than human labor. Payments to its human experts flow through cost of goods sold rather than appearing in the revenue line, which means its $375 million run rate is closer to net revenue than the gross figures of its competitors.
Origins and the Stanford Lineage
Snorkel's roots trace back to a Stanford artificial intelligence laboratory. Co-founder and CEO Alex Ratner and his team spent four years on research before the company launched commercially in 2019. That research foundation informed the company's early focus on programmatic data labeling, and the expertise developed during that period is what allowed the transition to end-to-end dataset production.
The 17-month gap between the Series D and Series E reflects both the speed of the market and the confidence investors have in the trajectory. A $350 million raise at a nearly threefold valuation increase is not a bet on potential in the abstract. It is a valuation grounded in revenue that has already scaled by a factor of eighteen.
The broader pattern across the AI infrastructure layer is unmistakable. As foundation models grow more capable, the bottleneck has moved from the model itself to the data that trains and fine-tunes it. The companies that solve the data supply problem are attracting the kind of capital that, just a few years ago, would have flowed almost exclusively to model builders.