Snorkel AI Valuation Triples to $3.5B on Training Data Boom

Data startup Snorkel AI has raised $350 million in fresh funding at a $3.5 billion valuation, Reuters reported Tuesday, after CEO Alex Ratner disclosed the round. The financing was led by Insight Partners and S32, with existing backers Addition, Greylock and Wells Fargo also participating, according to the company. The new Snorkel AI valuation is nearly triple the $1.3 billion it commanded after raising $100 million in May 2025, a sign of how much money is now chasing the companies that supply AI labs with training data.

The fundraising follows a striking revenue run. Snorkel's annualized revenue run rate now stands at $375 million, TechCrunch reported, up from roughly $20 million a year earlier, an eighteenfold increase in about twelve months. That run is the backdrop for the Snorkel AI valuation jump. The growth is being driven by what the company calls data as a service: instead of selling software that helps customers build datasets, Snorkel now sells finished datasets and reinforcement-learning environments directly to frontier labs and other customers.

From labeling software to data as a service

Snorkel was founded in 2019 by researchers spinning out of Stanford's AI lab, and it started out as a software business. The pivot came last fall, when the company began supplying finished datasets and RL environments instead of tools. That change transformed the economics: revenue now scales with the data labs consume rather than the seats they buy. That pricing power is what underwrites the Snorkel AI valuation.

The product, in Snorkel's telling, is an "agentic data development platform" that pairs human experts with thousands of specialized AI models and agents to create and vet training data. Experts design scenarios, tasks and grading rubrics, while AI automates much of the labor-intensive quality assurance, Ratner told Reuters, adding that the company works with AI labs to design new ways of acquiring high-quality datasets. Behind the platform sits a network of tens of thousands of specialists across fields such as coding, law and medicine. Snorkel sells the data products that network generates rather than billing for human labor, a structure Ratner said lets the company pay experts more generously while protecting its margins.

"The teams pushing the frontier want a research data partner who pioneers the science of data development," said Alex Ratner, co-founder and CEO of Snorkel AI, in a company news release reported by Tech Startups. "That's what Snorkel was built to be: the frontier lab for agentic data, combining human excellence with over a decade of research and technology."

Why frontier labs are paying up

The bet investors are making is that data has become the binding constraint on AI progress. As models grow more capable, developers have moved beyond simple labeling work and started demanding harder, higher-stakes data to train and evaluate their systems: complex reasoning tasks, simulated environments, carefully graded evaluation sets. That data bottleneck is the core argument behind the Snorkel AI valuation. Snorkel said it will use the fresh capital to expand the agentic data factory that produces those assets for advanced AI labs and other customers.

The numbers from elsewhere in the sector explain the urgency. TechCrunch reported that rival data lab Mercor's gross annualized revenue has climbed to $2 billion, while Handshake hit $1 billion earlier this year. The same appetite for specialized data is showing up across the industry, from AI systems trained to test drugs in silico to the evaluation pipelines labs use to grade their models. One caveat, flagged by TechCrunch: because these companies pass most of their gross revenue straight through to the specialists doing the work, their net revenue is far lower than the headline figures. Snorkel's version sidesteps that dynamic, since it sells RL environments and complete datasets rather than human labor, so payments to experts sit in cost of goods sold rather than inflating the top line.

For the specialists, the boom is a jobs story as much as a funding story. Tens of thousands of coders, lawyers and doctors now get paid to design the tasks and rubrics that train frontier models, work that looks less like gig labeling and more like expert consulting. It is part of a broader shift in how young professionals relate to AI: a recent Deloitte survey found sixty-three percent of workers already use AI tools on the job, and the data economy is turning some of them into the people AI learns from.

Snorkel's next test is whether the factory can keep up. Demand for complex training data keeps climbing as labs push models into harder territory, and the company's expansion plans put the new money directly into production capacity. For now, the Snorkel AI valuation marks the market's latest verdict: in the AI boom, the data suppliers are getting rich alongside the model builders.