Every frontier AI model that exists today was trained on data. Every improvement to those models — every new capability, every reduction in hallucination, every increase in reasoning accuracy — required more data, better data, and more precisely constructed training environments. The companies building those frontier models do not build their own training data. They buy it, commission it, and increasingly depend on a small number of specialised suppliers who can deliver the quality and scale they require.
Snorkel AI is the dominant player in that supply chain. On September 22 2026, the San Francisco company raised $350 million in a Series E at a $3.5 billion valuation, co-led by Insight Partners and S32. Follow every US startup funding story at BestStartup US.
The Revenue Number That Explains Everything
Snorkel AI’s annualized revenue run rate reached $375 million in September 2026. Twelve months earlier, that number was approximately $20 million. That is an eighteenfold increase in twelve months — a growth rate that almost never occurs in enterprise software, and almost certainly never occurs in a company that builds physical infrastructure for AI training rather than a pure software product.
The valuation of $3.5 billion represents a 9.3 times revenue multiple — modest by frontier AI standards, aggressive by traditional enterprise software benchmarks. The investor base reflects that positioning. New investors in this round include March Capital, Blumberg Capital, Allegis Capital, Frontline, Standard and Third Point Ventures. Existing investors Greylock, Lightspeed, GV, Addition, Wells Fargo and Walden Catalyst also participated. Read more US AI startup funding stories at BestStartup US.
What Snorkel AI Actually Builds
Snorkel AI describes itself as the frontier AI data lab. That label is precise. The company builds training datasets, evaluation frameworks, synthetic data environments and reinforcement learning setups for AI systems that require more than generic publicly available data to reach frontier capability.
The shift from its original positioning as a labelling automation tool to its current positioning as a finished data products company is significant. Labelling automation helped companies create their own training data faster. Finished data products means Snorkel creates the data itself — expert-constructed datasets covering specific domains, capabilities and evaluation benchmarks that frontier AI labs cannot easily produce internally at the speed and quality they require.
Today’s frontier and agentic AI systems demand what Snorkel calls Data 2.0 — expert agentic tasks, environments and evaluation rubrics that take even the most qualified humans hours or days to construct individually. Snorkel has built the factory that produces these at scale. Its customers include frontier AI labs, hyperscalers, US government agencies and large enterprises. Follow every AI infrastructure story at BestStartup US.
Why This Round Is Different From Every Other AI Round This Week
The week of September 21-25 2026 produced dozens of large AI funding rounds. Most of them were bets on AI applications — companies using AI to build products that serve end customers in specific verticals. Snorkel AI is a bet on AI infrastructure — the layer underneath applications, underneath models, underneath everything visible in the AI stack.
Infrastructure bets are structurally different from application bets. Applications compete on product quality, distribution and brand. Infrastructure competes on reliability, scale and switching costs. Once a frontier AI lab is using Snorkel’s data factory for its training runs, switching to a different supplier means rebuilding the data pipelines, re-establishing quality benchmarks and absorbing months of transition costs. That switching cost is a structural moat that application companies rarely achieve.
Snorkel’s 18x revenue growth is the commercial proof that frontier AI labs have decided it is the supplier they want. The $350 million round funds the capacity expansion required to serve the next wave of demand as model training cycles accelerate and the number of frontier labs globally continues to grow. Also read: Crusoe raises $3.9B AI infrastructure Series F and Profound raises $180M for AI search visibility. Follow every US startup story at BestStartup US.
Key Takeaways
Snorkel AI raised $350 million Series E at $3.5 billion valuation on September 22 2026. Co-led by Insight Partners and S32. Annualized revenue run rate $375 million — eighteenfold increase in 12 months. Builds training datasets, evaluation frameworks and RL environments for frontier AI labs, hyperscalers and US government. Shift from labelling automation to finished data products. Customers include frontier AI labs, hyperscalers and large enterprises. New investors include March Capital, Third Point Ventures, Blumberg Capital. Existing investors Greylock, Lightspeed, GV, Addition and Wells Fargo participated.
Frequently Asked Questions
What is Snorkel AI and what does it do?
Snorkel AI is a frontier AI data lab that builds training datasets, evaluation frameworks, synthetic data environments and reinforcement learning setups for AI systems. Its customers include frontier AI labs, hyperscalers and US government agencies.
How much did Snorkel AI raise in September 2026?
Snorkel AI raised $350 million in a Series E at a $3.5 billion valuation on September 22 2026, co-led by Insight Partners and S32, with participation from March Capital, Greylock, Lightspeed, GV, Third Point Ventures and others.
How fast is Snorkel AI growing?
Snorkel AI’s annualized revenue run rate reached $375 million in September 2026, an eighteenfold increase over the previous twelve months — one of the fastest revenue growth rates recorded in enterprise AI infrastructure.
Where can I follow US AI startup funding news?
Follow every US startup funding round and AI ecosystem story at BestStartup US — updated every week.
