Snorkel AI completed a Series E funding round of $350 million on September 22, 2026, achieving a valuation of $3.5 billion – nearly three times the value from 17 months ago. The company provides training data and simulated test environments for large AI models. The annual revenue rate has grown 18-fold to $375 million within twelve months.
Insight Partners and S32 lead the round
The funding is led by investment firms Insight Partners and S32, with participation from Alphabet’s venture capital arm GV as well as existing investors Greylock, Lightspeed, and Wells Fargo. The previous valuation of $1.3 billion came from a Series D round 17 months earlier. The company, spun out of the Stanford AI Lab in 2019, has since transformed from a software provider for automated data annotation to a Data-as-a-Service provider selling complete datasets rather than just tools.
CEO Alex Ratner describes Snorkel in the official statement as “the frontier lab for agentic data.” According to the company, its customers include leading AI labs and corporations, but Snorkel does not disclose specific names. The research department has published over 250 peer-reviewed articles, which have been cited more than 25,000 times.
The round is part of a series of large financings for AI infrastructure this year. Just in July, Fireworks AI reached a valuation of $17.5 billion, and a few weeks later, chip provider SambaNova followed with $1 billion in fresh capital. Unlike these companies, Snorkel does not sell computing power or models, but the raw materials behind them: curated data and test environments.
Data business shifts to reinforcement learning
Snorkel is expanding its offerings from traditionally labeled datasets for supervised learning to materials for reinforcement learning, where models solve open tasks and receive feedback. To do this, the company employs several tens of thousands of human experts who create multi-page evaluation grids for complex tasks and run training scenarios in simulated work environments – such as replicated developer machines – where AI agents can be tested.
Traditional providers like Scale AI primarily delivered raw data and simple labels. In contrast, Snorkel positions itself as a provider of complete, vetted training and evaluation packages for specific fields, from software development to regulated industries. This shift also explains the rapid revenue growth, as customers increasingly purchase ready-made solutions instead of individual tools.
As TechCrunch reports, Snorkel accounts for the payment of the employed experts as production costs and not as a deduction from revenue – unlike some competitors. The annual revenue rate of $375 million and the 18-fold growth are thus independent, unverified company statements, and independent audit reports are not available.
New capital also flows into security research
Snorkel aims to use the fresh capital to expand its “agentic data factory,” hire additional engineers, and deepen investments in AI security research. Part of the money is also intended for open, freely available evaluation benchmarks that will allow the capabilities of AI agents to be compared across industries. The company has not yet provided specific timelines for individual projects.
Additionally, Snorkel announces plans to expand its research program into new industries and to collaborate more closely with professionals from regulated areas such as law or healthcare. The company did not specify how many of the several tens of thousands of external experts will be added in the future.
For companies looking to train or fine-tune their own AI agents, Snorkel promises ready-made test scenarios instead of labor-intensive in-house development. For example, those wanting to build an assistant for contract review can rely on pre-made evaluation grids from the legal field, rather than assigning their own legal experts to create test data. Snorkel has not yet provided a specific date for when new industry packages will be available, nor specific prices for individual fields.
Whether the influx of capital primarily reflects the growing demand for high-quality, human-reviewed training data for increasingly sophisticated AI agents or is part of the generally heated funding climate for AI infrastructure in 2026 cannot be answered based on a single round. What will be crucial is whether the self-reported revenue figures are confirmed in the next independently audited funding round.
