The race to build smarter AI systems has created an insatiable appetite for one critical ingredient: high-quality training data. Snorkel AI, a seven-year-old startup born out of Stanford research, just raised $350 million in a Series E round at a $3.5 billion valuation, nearly tripling its previous valuation of $1.3 billion from just 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 numbers tell a striking story. Snorkel AI now reports an annualized revenue run-rate of $375 million, representing an 18-fold increase over the past 12 months. That kind of growth trajectory is rare even by Silicon Valley standards, and it signals just how much AI labs are willing to spend to fuel their model development pipelines.
From Labeling Software to Full-Service Data Provider
Snorkel AI did not get here by standing still. The company originally focused on software that automated the tedious process of labeling data for machine learning. Last year, it made a significant strategic pivot, shifting toward what it now calls data-as-a-service, delivering complete, ready-to-use datasets directly to clients rather than just giving them tools to build their own.
Rather than relying purely on crowdsourced human labor, Snorkel uses a hybrid model that combines its proprietary software and AI models to generate data synthetically, while still incorporating input from domain-specific subject matter experts. This approach gives it a meaningful structural advantage over competitors who rely more heavily on human-only workflows, allowing for faster turnaround and more scalable output.
A Booming Market With Real Revenue to Back It Up
Snorkel is not alone in riding this wave. The broader AI data industry has seen explosive growth across multiple players. Mercor recently reported gross annualized revenue climbing to $2 billion, Handshake crossed the $1 billion milestone earlier this year, and Micro1 has scaled to approximately $500 million. However, context matters here. Many of these companies pay out between 60% and 70% of their top-line revenue directly to the human specialists doing the work, which means their net revenue figures are substantially lower than the headline numbers suggest.
Snorkel operates differently. Because it sells reinforcement learning environments and complete datasets rather than human labor hours, payments to its experts are treated as a cost of goods sold, not subtracted from its reported revenue figures. That accounting distinction makes its $375 million run-rate a cleaner and arguably more meaningful metric than comparable figures from human-labor-dependent competitors.
What This Means for Businesses Adopting AI Tools
For enterprise buyers and technology decision-makers, Snorkel AI’s rise is a clear signal that AI training data quality is now a core competitive variable, not a back-office concern. Companies evaluating AI platforms and enterprise tools in 2025 and beyond should factor in how their vendors source, validate, and maintain training data. The valuation premium investors are placing on companies like Snorkel reflects a market consensus: the quality of the data underneath an AI product increasingly determines the value of the product itself.
