On June 12, 2025, Meta announced a $14.3 billion investment for a 49 percent, non-voting stake in Scale AI, and folded founder and CEO Alexandr Wang into its superintelligence effort. Wang confirmed his exit in a memo to staff, and the company handed the CEO seat to Jason Droege, previously its chief strategy officer. The deal was shocking enough on its own; the real aftershock hit the customer list.
On June 18 and 19, OpenAI confirmed it was phasing out its data work with Scale AI, while Reuters reported that Scale’s largest customer, Google, also planned to cut ties. Within a single week, one large check redrew the competitive map of the AI data supply chain — and turned “neutrality” into a new line item on the industry’s pricing sheet.
OpenAI confirms the wind-down
An OpenAI spokesperson told CNBC that the company had been pulling back from Scale over the previous six to twelve months, and said the decision “wasn’t influenced by the Meta partnership.” Scale, the spokesperson noted, accounted for only a small fraction of OpenAI’s overall data needs, and the company had been looking for providers that kept pace with innovation and understood what its latest models required.
Fortune’s report, published with Bloomberg, filled in the rest: OpenAI had concluded Scale was no longer the best fit for the more specialized data that advanced AI models demand, and had turned to other providers such as Mercor. The about-face is telling because OpenAI CFO Sarah Friar had said only days earlier at VivaTech in Paris that “we don’t want to ice the ecosystem because acquisitions are going to happen.” The gap between the public reassurance and the internal calculus is the most honest annotation on this whole episode.
Google reportedly next, and the neutrality problem
Reuters, citing people familiar with the matter, reported that Google — Scale’s largest customer — planned to drop Scale as a data provider after the Meta deal. Google declined to comment. The shared worry among customers was simple: with Meta as a major investor, could Scale give Meta visibility into rivals’ AI development efforts? A data vendor’s value rests on serving everyone at once, and once that premise loosens, the partnership loosens with it.
Scale’s general counsel published a blog post to calm the waters, saying Meta would get no preferential treatment, would not see other customers’ confidential information, and that Wang would stay out of day-to-day operations. TechCrunch spotted the flip side of the panic: Scale’s competitors saw an influx of interest, as AI model providers scrambled to line up “neutral” data partners. Commercial terms can be written into contracts; trust cannot. And the biggest customers pivot fastest, because they are the ones who can afford alternatives.
Scale’s response and its new direction
Facing the churn, Scale’s interim CEO Jason Droege told employees and customers the company was not winding down, and said it would double down on its applications business, building custom AI for governments and enterprises. The statement amounts to a positioning call: beyond the data-labeling core, applications are now the growth story Scale wants to tell.
For Scale, this is a re-selection of identity: remain the data backbone for everyone, or become a solutions vendor for governments and enterprises. What Wang took with him was not just himself but a group of employees — and the market’s imagination about the company’s independence. Droege’s first task is to convince the remaining customers to stay while two of the biggest walk out the door.
What it means for the AI data supply chain
The episode exposes a structural issue in the AI data industry: labeling firms serve many frontier labs at once, and when one client binds the company and its founder with a giant investment, every other client’s assumption of neutrality starts to crack. The money from an equity stake arrives once; customer churn compounds. Scale likely understood that math — it simply had little choice once Meta came calling.
For developers and enterprise buyers, the June 2025 shakeout offers a concrete lesson: when evaluating data vendors, ownership structure now matters as much as data quality and delivery capacity. In one week, Scale taught the industry a crash course in pricing neutrality risk — and the tuition was paid out of its customer roster.
Sources
- CNBC: OpenAI is winding down its work with Scale AI
- Fortune: OpenAI is phasing out Scale AI work
- TechCrunch: OpenAI drops Scale AI as a data provider
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.