Snorkel AI Raises $350M as AI Training Data Demand Soars 18x

If you ever wanted a single data point that shows how serious the enterprise AI race has become, look at Snorkel AI’s week.

The San Francisco-based company announced a $350 million Series E on September 22, 2026, tripling its valuation to $3.5 billion. More striking than the number is the trajectory behind it: Snorkel grew its annualized revenue from roughly $20 million to more than $375 million in a single year. That is an 18x increase in twelve months, driven almost entirely by a pivot that most software companies would be too cautious to attempt.

Insight Partners and S32 co-led the round, joined by new backers including March Capital, Third Point Ventures, Blumberg Capital, Allegis Capital, Frontline, and Standard, alongside existing investors Lightspeed, Greylock, and GV. The company expects to reach profitability in 2026.

From Software to Finished Data

Snorkel started life as a research project out of Stanford’s NLP group, built around the idea that you could automate data labeling with programmatic rules rather than manual annotation. For years that was the product: software that helped engineers label training data faster.

In September 2025, CEO Alex Ratner made a bet that changed the business entirely. Instead of selling tools, Snorkel started delivering the finished product — completed datasets and training environments — as a service. Customers stopped buying software licenses and started buying the data itself.

The market responded immediately. Revenue exploded as frontier AI labs, hyperscalers, government agencies, and enterprises all found they needed the same thing: high-quality, domain-specific data that could actually improve model performance in specialized fields.

To deliver that data, Snorkel built a network of tens of thousands of domain specialists — coders, lawyers, physicians, researchers — who work to create, verify, and label the complex examples that generic data pipelines cannot produce.

Why This Matters Now

The AI model market is crowded in a way it has never been before. GPT-6 Astra, Claude Opus 5.5, Gemini 3.8 Flash, and a dozen other frontier models all launched within weeks of each other in September 2026 alone. The raw model capability gap between top providers is narrowing.

That shifts competitive advantage to something more durable: the data those models are trained and fine-tuned on.

A customer service agent trained on generic web data will answer generic customer service questions. One fine-tuned on your specific product documentation, your actual support transcripts, and your proprietary edge cases will perform entirely differently. The model architecture is close to the same. The data is not.

Snorkel’s growth reflects an industry-wide recognition that investing in the data layer is no longer optional for enterprises that want AI to do meaningful work. The cost of getting data wrong is now visible in production failures, not just benchmark scores.

The Domain Specialist Advantage

There is something worth noting about how Snorkel builds its datasets: it relies on human expertise at scale.

The company does not claim that AI can write its own training data without help. Instead, it combines programmatic automation with a vetted network of specialists who understand the specific domain — whether that is medical diagnosis, contract law, software engineering, or scientific research.

This is a useful model for any business thinking about AI deployment. The highest-value AI applications are not the ones running on the best general-purpose model. They are the ones running on models trained with the best domain-specific data, produced by people who actually understand that domain.

If your business has proprietary data — customer interactions, operational records, specialist knowledge — that data is now a genuine competitive asset. Not because AI makes it easy to use, but because building the right training and evaluation sets from it requires real expertise and real investment.

What This Means for Business

The data advantage is real. AI models with access to high-quality, domain-specific training data outperform generic deployments on specialized tasks. This is not marketing — it is what is driving Snorkel’s 18x revenue growth.

The floor for AI data investment is rising. As enterprises get more serious about production AI, they are spending more on data quality, not less. Companies that treat data as an afterthought will find themselves running expensive AI infrastructure that underdelivers.

Domain expertise is not being replaced by AI — it is being packaged into AI. Snorkel’s model is built around human specialists whose knowledge gets embedded into training data. The specialists matter. The expertise matters. AI amplifies it.

Data-as-a-service is emerging as a real category. Snorkel’s pivot from software to finished datasets points to a broader shift: the infrastructure layer of AI is maturing to the point where specialized data production becomes a distinct, outsourceable function.

For businesses building with AI — whether through platforms like Omni Ops, custom development, or first-party deployments — this is the period to get serious about what data you have, what data you need, and how you are going to produce training and evaluation sets that reflect your actual operating environment.

The companies getting results from AI right now are not the ones who found the best model. They are the ones who built the best data pipeline to support it.

---

Enterprise DNA helps businesses build AI systems that deliver real results. If you are thinking about how your organization’s data can become a competitive advantage, talk to the team.