A billion dollars, and who wrote the checks
Scale AI announced on May 21, 2024 that it had raised $1 billion in a Series F round at a $13.8 billion valuation, led by Accel, an existing investor. The round nearly doubled Scale's prior valuation of $7 billion, set in its 2021 Series E. [1][2]
Alongside Accel and a long list of returning venture investors (Y Combinator, Index Ventures, Founders Fund, Coatue, Thrive Capital, Tiger Global and others), the round brought in a group of new corporate investors: Amazon, Meta, Cisco, Intel, AMD, Qualcomm and ServiceNow, each investing through its own venture or strategic arm. Scale did not disclose revenue figures alongside the announcement. [1][2]
The investors were already customers
Scale AI's business is labeling and evaluating the data that trains other companies' models: image and text annotation, human review of model outputs, red-teaming and benchmark scoring, sold to AI labs, automakers and the U.S. government. By the time of the Series F, its customer list already included Microsoft, Toyota, General Motors, the Department of Defense and, notably, Meta and OpenAI. Amazon and Meta joining the round as investors put two of Scale's own customers on its cap table. [2]
Atlas interpretation: Buying a stake in a vendor a company already depends on is a common enough move: it can secure supply, buy a board seat's worth of visibility, or just ride a bet the buyer likes. But it also means the investor's interest in the vendor staying independent and well-run is no longer purely commercial. That tension was mostly theoretical in May 2024, when Meta held a passive minority stake alongside a dozen other investors. It stopped being theoretical thirteen months later. [2]
The bet the round was funding
Scale pitched the raise around a thesis it called data abundance: that frontier labs would keep needing more, and better, labeled data to keep scaling models, and that whoever supplied it at the largest scale would keep winning the business. Founder Alexandr Wang was quoted describing the goal as making sure frontier labs are not "data-constrained in getting to GPT-10." [2]
Atlas interpretation: The wager was really two bets stacked on each other: that data would remain a bottleneck worth paying a premium for, and that Scale specifically, rather than a rival or a customer's own in-house team, would keep capturing that spend. The first bet held up well enough. The second did not resolve the way $1 billion of new capital would suggest. Reinforcement learning from human feedback, the product line built on exactly this thesis, is also the product line Meta bought half of a year later, and OpenAI and Google were both reported to have pulled back from Scale as a vendor once a competitor held a stake in it. [2][3]
Thirteen months later, one of the investors bought in
On June 12, 2025, Meta invested $14.3 billion for 49 percent of Scale AI, valuing the company above $29 billion, roughly double the Series F price a year earlier, and Alexandr Wang left to lead a new superintelligence lab inside Meta while remaining a director on Scale's board. [3]
Atlas interpretation: Read next to each other, the two rounds tell a tighter story than either does alone. The Series F recruited Meta as a minority financial investor in a vendor it already used. The 2025 deal made Meta Scale's largest shareholder and its founder a Meta employee, while OpenAI and Google were the customers reported to be walking away. The data abundance thesis was correct that labeled data stayed valuable. It did not anticipate that the company selling it would become harder for competing labs to trust as a neutral vendor once one of its biggest customers effectively owned it. [3]
Sources
- Scale's Series F: Expanding the Data Foundry for AI
Scale AI · May 21, 2024
- Data-labeling startup Scale AI raises $1B as valuation doubles to $13.8B
TechCrunch · May 21, 2024
- Scale AI Announces Next Phase of Company's Evolution
Scale AI · Jun 12, 2025