Snorkel AI Raises $350 Million at $3.5 Billion on a $375 Million Run Rate

Snorkel AI announced a $350 million Series E on Sept. 22 at a $3.5 billion valuation, with Insight Partners and S32 leading. The figure that will travel is the valuation, and the more useful figure is the $375 million annualized run rate that Snorkel says supports it. Neither the company's post nor the coverage fetched labels the valuation pre-money or post-money.

The valuation also appears two ways. Snorkel's own post, written by co-founder and CEO Alex Ratner, says $3.5 billion. Crunchbase News listed the round at $3.2 billion in its weekly roundup for Sept. 19 to 25. The company's figure is used here, and the gap is unexplained.

The multiple, in words

1. May 2025 round1.3 $ billion
2. Crunchbase, Sept. 20263.2 $ billion
3. Snorkel post, Sept. 20263.5 $ billion

Source: Snorkel AI post; Implicator; Crunchbase News, accessed 2026-10-01

At $3.5 billion against $375 million, the round prices Snorkel at about 9.3 times annualized revenue. That is a multiple on a run rate, a projection from recent weeks rather than audited sales. Against the $1.3 billion valuation Implicator reports from May 2025, the step-up is about 2.7 times.

Ratner wrote "Since launching our new data-as-a-service offering nearly a year ago, we've grown over 18x." If that multiple applies to the $375 million figure, the starting point was about $21 million. The post does not say which base the 18x refers to.

What changed in the business

Snorkel began as labeling-automation software. It now sells finished training and evaluation datasets built with subject experts and software, plus reinforcement-learning environments. Implicator lists frontier AI labs, hyperscalers, enterprises and U.S. government agencies as customers and notes that no source breaks revenue out by category.

That matters because a concentrated customer base changes what a multiple means. If a handful of labs account for most of the run rate, the figure is closer to a contract book than to recurring software revenue. The sources do not say either way.

Investor group	Names
Leads	Insight Partners, S32
New	Third Point, March Capital, Blumberg Capital, Allegis, Standard VC, Frontline
Returning	Addition, Lightspeed, Greylock, GV, P7, Wells Fargo, Walden Catalyst, Factory

What is not in the announcement

No named customer, no gross margin and no independent comparison of dataset quality appear in any source. The site's earlier GPT-Synopsys coverage hit the same wall: a vendor claim with no customer willing to be named. The Maven Robotics round at least came with a unit count.

Expert-labeled data carries labor cost that software does not, so margin is the number that decides whether 9.3 times is high or low. The $350 million raised is also about 93 percent of the $375 million run rate, a ratio that shows how much capital arrived relative to sales already booked. The next observable event is any disclosure of margin or customer concentration, which the company has so far not offered.