Scale AI is worth about $29 billion — the number implied by Meta's $14.3 billion purchase of a 49% stake in June 2025, more than double the company's $13.8 billion valuation just thirteen months earlier.
That's the short answer. The longer answer is that the Meta deal did two things at once: it handed Scale AI's early investors and founder Alexandr Wang a massive markup, and it triggered an exodus of the very customers — OpenAI, Google, Microsoft — whose training-data business built that valuation in the first place. Fourteen months later, no new round has re-priced Scale, while two of its closest rivals have doubled their own valuations chasing the clients Scale lost.

What Is Scale AI's Valuation in 2026?
Scale AI's valuation is roughly $29 billion, a figure derived from Meta paying $14.3 billion for a 49% non-voting stake in June 2025 — not from a traditional priced venture round. That $29 billion mark is more than double the $13.8 billion valuation Scale carried after its $1 billion Series F in May 2024, and as of mid-2026 it remains the company's most recent market-implied valuation, since no independent round or secondary sale has priced the company since the Meta transaction closed.
The structure of the deal matters as much as the number. Meta's stake is non-voting, which let Scale AI continue operating as a legally independent company rather than becoming a subsidiary — a distinction that mattered enormously to the customers who, until that point, trusted Scale with sensitive model-training data. As part of the deal, Scale AI founder and CEO Alexandr Wang joined Meta's superintelligence research team directly, with Jason Droege stepping in as interim CEO to run day-to-day operations.
Scale AI Valuation vs. the Rest of the Data-Labeling Market
Scale AI still holds the top valuation in the AI training-data sector, but the field closed in fast during 2026. Ranked by most recent known or in-talks valuation:
The Full Comparison Table: Valuations, Revenue, and Last Round
| # | Company | Valuation | Last Priced | Revenue / ARR |
|---|---|---|---|---|
| 1 | Scale AI | $29B | Jun 2025 (Meta stake) | ~$1B (2025) |
| 2 | Mercor | $20B (in talks) | Jul 2026 talks; $10B Oct 2025 | ~$2B ARR run rate |
| 3 | Surge AI | $15B–$25B (in talks) | 2025 raise in progress | $1B+ (2024) |
| 4 | Turing | $2.2B | Mar 2025 Series E | ~$300M ARR |
| 5 | Snorkel AI | $1.3B | May 2025 Series D | ~$148M ARR |
| 6 | Appen (ASX: APX) | ~$0.2B–$0.4B mkt cap | Public (ASX) | Declining since 2020 peak |
Figures are 2025–2026 estimates blended from Meta's disclosed deal terms, TechCrunch, Forbes, PitchBook, and companiesmarketcap.com. Mercor and Surge AI figures reflect reported in-talks valuations as of July 2026, not closed rounds. Appen's market cap is converted from AUD and fluctuates with ASX trading.
What Scale AI's $29 Billion Valuation Cost the Company
The markup came with a real business cost. Within weeks of the Meta announcement, OpenAI dropped Scale AI as a data provider, and Google — which had reportedly planned to spend around $200 million on Scale's services — sharply scaled back its engagement. Microsoft also pulled back. The concern was straightforward: with Meta holding a 49% stake and a board-adjacent relationship through Wang, competitors worried their proprietary training data and model-development plans could leak into Meta's own AI roadmap, even with contractual firewalls in place. Scale's general counsel publicly denied that Meta would receive preferential access to other customers' data, but the reassurance didn't stop the walkouts.
The company responded with layoffs — roughly 200 employees and 500 contractors cut in July 2025 as it restructured its generative-AI data unit — and a pivot toward government and enterprise AI tooling rather than pure data-labeling volume. Revenue kept growing through the disruption, from about $870 million in 2024 to just under $1 billion in 2025, but that's a materially slower growth curve than Mercor's, whose run rate roughly doubled to $2 billion in four months over the same stretch by directly absorbing the clients Scale AI lost.
What the headline misses
A $29 billion price tag reads as validation, but it's a single data point from a single, unusual transaction — a 49% non-voting stake sale, not a competitive priced round with multiple bidders setting the market price. Fourteen months on, no other investor has stepped in to confirm or challenge that number, which is itself a signal: the companies best positioned to know Scale AI's real trajectory (its own former hyperscaler customers) have been the ones walking away, not writing new checks. Appen's collapse from a roughly $4 billion valuation in 2020 to under $400 million today is the sector's live reminder that data-labeling valuations can compress fast once growth slows or a structural conflict of interest scares off the biggest buyers.
Scale AI's $29 billion valuation is real, but it's a 14-month-old number set by one unusual deal — not a market price confirmed by the customers who actually use the product.
Mercor and Surge AI are both closing the gap by absorbing exactly the clients that deal cost Scale.
Originally published in the Trace Cohen newsletter at Value Add VC.
Get VC data most people never see
— 100% free
Weekly benchmarks, valuations, and fund data. Join 5,000+ investors. No spam.