top of page

What Happened to Scale AI After Meta's $14.3 Billion Deal?

Writer: Page 21 Team
Page 21 Team
18 hours ago
5 min read

The short answer

In June 2025, Meta invested about $14.3 billion for a 49% stake in Scale AI, valuing the data-labeling company at roughly $29 billion, and hired its founder and CEO Alexandr Wang to lead Meta's new superintelligence effort. Within days, several of Scale's biggest customers, including Google and OpenAI, paused or wound down work with the company, because a data supplier part-owned by a rival AI lab is a hard sell. Scale named chief strategy officer Jason Droege interim CEO, and in July 2026 appointed former Google Cloud COO Francis deSouza as CEO, effective August 10, 2026. Scale says it remains an independent company.

Related on Page Twenty One: for the full story, read How Scale AI Became $29 billion AI Data Powerhouse: Business Model Breakdown.

When we broke down how Scale AI became a $29 billion data powerhouse, the story was about a neutral "picks and shovels" supplier selling to every AI lab at once. The Meta deal tested whether that neutrality could survive one of those labs owning nearly half the company.

$14.3B for a 49% non-voting stake, valuing Scale AI at about $29B (June 2025).
Meta Platforms headquarters sign in Menlo Park, California
Meta's Menlo Park headquarters. Photo: LPS.1 / Wikimedia Commons, CC0

What exactly did Meta buy?

According to Reuters, Meta took a 49% stake for $14.3 billion, a deal that valued Scale at about $29 billion. Scale's own announcement said the investment valued the company at "over $29 billion" and would "substantially expand" the commercial relationship between the two firms.

A few details matter more than the headline number:

  • It was not a full acquisition. Meta holds a minority of Scale's equity. A Scale spokesperson told CNBC that Meta would not have voting power at the company.

  • Shareholders got paid. Scale said it would distribute proceeds to shareholders and vested equity holders, giving early employees and investors liquidity while letting them stay invested.

  • The founder moved over. Alexandr Wang joined Meta to work on its AI efforts, while staying on Scale's board. A small number of Scale employees went with him.

Scale AI's valuation, round by round

Bar chart of Scale AI valuation from $1B in 2019 to $29B after the 2025 Meta deal
Scale AI valuation by funding round. Sources: Scale AI, Reuters, Crunchbase News, Dealroom. Chart: Page Twenty One

  • Aug 2019: Series C values Scale at about $1B

  • Dec 2020: Series D at $3.5B

  • Apr 2021: $325M Series E at $7.3B

  • May 2024: $1B Series F at $13.8B

  • Jun 2025: Meta invests $14.3B for 49%; Alexandr Wang joins Meta; valuation about $29B

  • Aug 10, 2026: Francis deSouza becomes CEO after Jason Droege's interim stint

Why would Meta pay this much for 49%?

A source briefed on the talks told Reuters that the main driver was bringing in Wang to lead Meta's superintelligence unit. This structure, a large investment plus hiring the founder instead of buying the whole company, has become a common pattern in AI. It gets a big tech company the talent and a close commercial partner without the regulatory scrutiny a full takeover would draw.

Why did Google and OpenAI pull back?

Scale's business depends on AI labs trusting it with sensitive information: what data they are collecting, how they are training models, and where their research is heading. Once a direct competitor owns 49%, that trust gets harder to keep, even with promises of separation.

Here is what was reported in the weeks after the deal:

  • Google halted several projects with Scale hours after the announcement, according to internal correspondence seen by Business Insider. Reuters had earlier reported Google planned to cut ties.

  • OpenAI confirmed it was phasing out its work with Scale. An OpenAI spokesperson told CNBC and Bloomberg that the company had already been winding down that relationship for six to 12 months and that the decision was not caused by the Meta deal. OpenAI said Scale was a small part of its data needs.

  • xAI paused some Scale projects, according to contractors who spoke to Business Insider.

Scale, for its part, said Meta would not get access to its business information or customer data.

Scale AI logo
Scale AI logo. Source: Wikimedia Commons (public domain, trademark of Scale AI)

What does this mean for Scale AI's business model?

Scale grew by being the default outsourced data partner for frontier AI. That model has two weak points that the Meta deal exposed:

1. Customer concentration

When a handful of labs make up much of your revenue, losing even two of them hurts. Companies building frontier models are few, and they talk to each other's former employees, so perception of a conflict spreads fast.

2. Rising demands for specialist data

OpenAI's comment is telling. It said it wanted providers that "understand what the latest models need." Early AI training relied on huge volumes of simple labeling. Newer models need expert-written data in areas like coding, math, medicine and law. That is a different kind of workforce and a more competitive market, with rivals pitching themselves as neutral alternatives.

3. The upside: one very large customer, and a pivot

Meta is one of the largest spenders on AI in the world. A deeper relationship with Meta can replace some lost revenue, and Scale also has government and enterprise work. Scale has also leaned into building AI applications for enterprises and governments. In a January 2026 post, Droege said 2025 was Scale's strongest financial year, with well over $1 billion in new business and enterprise customers including Mayo Clinic, BP and Allianz. The question is whether that mix can replace being the neutral supplier to every lab.

Aerial view of the Meta (Facebook) campus in Menlo Park, September 2019
Meta's Menlo Park campus from the air. Photo: Pi.1415926535 / Wikimedia Commons, CC BY-SA 3.0

What can founders learn from this?

Page Twenty One has covered plenty of companies that grew fast and then hit a wall, from Ginkgo Bioworks to Casper. Scale's story is different because it is not a collapse. It's a trade-off. A few lessons apply beyond AI:

  • Neutrality is an asset. If your customers compete with each other, your independence is part of what you sell. Taking money from one of them has a cost that won't show up on the term sheet.

  • Liquidity events can come from minority deals. Founders and employees don't need a full sale or IPO to cash out.

  • Your moat moves with the market. What made Scale essential in 2019 (labeling at scale) mattered less by 2025, when models needed expert data.

Frequently asked questions

Did Meta buy Scale AI?

No. Meta bought a 49% minority stake for about $14.3 billion. Scale says it is still an independent company, and Meta does not have voting control.

How much is Scale AI worth?

The Meta deal valued Scale at about $29 billion, according to Reuters and Scale's own announcement in June 2025.

Who is the CEO of Scale AI now?

Francis deSouza, formerly COO of Google Cloud and CEO of Illumina, became CEO on August 10, 2026. Jason Droege had served as interim CEO since Alexandr Wang left for Meta in June 2025.

Why did OpenAI stop working with Scale AI?

OpenAI said it had already been reducing its reliance on Scale for six to 12 months and wanted providers with more specialized data. It said the Meta deal did not drive the decision.

What does Alexandr Wang do at Meta?

He joined Meta to lead its superintelligence efforts and remains a director on Scale's board.

Related reading on Page Twenty One

Sources

Image credits: cover background photo of the Meta headquarters sign by Nokia621, CC BY-SA 4.0, via Wikimedia Commons. Scale AI logo: CC BY-SA 4.0, via Wikimedia Commons.

Comments

News worthy of your time.

Thanks for submitting!

  • Instagram
  • Twitter
  • YouTube
  • Pinterest
  • LinkedIn

© 2026 by Page Twenty One

bottom of page