What is Scale AI?

What is Scale AI?

Scale AI is the largest independent 'data foundry' for artificial intelligence — it labels, generates, and curates the training data, human feedback, and model evaluations used to build and grade frontier AI models. Founded in 2016, it grew from autonomous-vehicle image labeling into a core data supplier for OpenAI, Google, Microsoft, Meta, and the U.S. government, then in June 2025 sold a roughly 49% non-voting stake to Meta for $14.3 billion at a ~$29 billion valuation.

Scale runs a vertically integrated platform plus a managed network of human annotators and subject-matter experts, organized into three lines: Build AI (the Scale Data Engine for generative AI, government, and automotive), Apply AI (the Scale GenAI Platform and the Donovan defense product), and Evaluate AI (Scale Evaluation for benchmarking and red-teaming models). In practice, customers pay Scale to produce the high-quality labeled and human-feedback data that makes large language models accurate, safe, and aligned, and increasingly to deploy and evaluate those models in production.

The business scaled with the LLM boom and then accelerated through the Meta deal. Revenue rose from roughly $870 million in 2024 to about $2 billion in 2025 — more than doubling — on the back of over $1 billion in new bookings, nearly half of which closed in Q4 2025. Customers cluster into three segments: frontier model labs, the U.S. government (including a $249 million Department of Defense engagement and two nine-figure awards in Q3 2025), and enterprises, with new 2025 logos including Mayo Clinic, BP, and Allianz.

Meta's June 2025 investment reshaped the company. Founder-CEO Alexandr Wang left to lead Meta's Superintelligence Labs, Jason Droege stepped in as interim CEO, and several rival labs — Google, OpenAI, Microsoft, and xAI — began pulling work over data-leakage and neutrality concerns, prompting a 14% workforce cut in July 2025. Scale remains independent and is repositioning around enterprise, government, and model-evaluation demand under a 'reliability' mantra.

Sources:Sacra — Scale AI revenue & valuationScale AI — building for 2026 (Droege)TechCrunch — Meta confirms Scale investment, Wang departs

What does Scale AI offer?

Scale sells training-data generation, human feedback (RLHF), model evaluation, and applied/government AI products, grouped into Build AI, Apply AI, and Evaluate AI.

Sources:Contrary Research — Scale business breakdownScale AI — generative AI data engine

How does Scale AI make money?

Scale makes money by charging for data work and AI services — per-task/per-unit pricing for self-serve labeling and custom enterprise and government contracts for large training-data, RLHF, evaluation, and applied-AI programs. It marks up a managed network of human labor, yielding an estimated 50%+ gross margin.

At the low end, Scale runs a pay-as-you-go, per-unit model: historically about 2¢ per image and 6¢ per annotation, with free starter tiers (the first ~1,000 labeling units or ~10,000 images free) to seed self-serve usage. The economics come from arbitrage — Scale programmatically routes tasks to a global pool of annotators and experts, marks up the labor, and expands margin as pre-labeling AI lets each contractor label more per hour — producing roughly 50%+ gross margins (below pure-software norms because of the heavy human-service component).

The real revenue is enterprise and government: multi-million-dollar, volume-committed contracts for frontier-model training data, human feedback, evaluation, and increasingly applied-AI deployments. Google was reportedly Scale's largest customer at roughly $200 million per year, and Scale holds a $249 million U.S. Department of Defense engagement plus two nine-figure government awards booked in Q3 2025 — a sign that large, recurring, services-heavy deals (not credit-card-style usage) drive the bulk of the top line.

Growth historically tracked the frontier-lab spend race on data and human feedback. After Meta's stake triggered rival labs to pull work, Scale's growth thesis shifted toward government, enterprise 'Apply AI' deployments (Mayo Clinic, BP, Allianz), and model evaluation — markets where neutrality, reliability, and security clearances matter more than a competitor's ownership. 2025 revenue still roughly doubled to ~$2 billion on more than $1 billion in new bookings, showing the rebuild outran the customer churn.

Sources:Sacra — Scale AI business model & marginsScale AI — record 2025 bookings (TipRanks)CNBC — Google plans split after Meta deal

Who leads Scale AI?

Scale was founded by Alexandr Wang and Lucy Guo in 2016. Since June 2025 it has been led by interim CEO Jason Droege, after Wang left to run Meta's Superintelligence Labs (he remains a Scale board director).

Sources:Scale AI — next phase of company evolutionAxios — Droege interview (May 2026)CNBC — Wang exit for Meta, Droege named interim CEO

How do you contact Scale AI's leadership?

Scale AI's email pattern is First.Last@scale.com (e.g. jane.doe@scale.com), the dominant format on its primary domain. The addresses below follow that verified pattern; they are not individually published, so treat them as format-derived rather than confirmed personal inboxes. For media, Scale routes through press@scale.com.

• Jason DroegeInterim Chief Executive Officerjason.droege@scale.com
• Dennis CinelliChief Financial Officerdennis.cinelli@scale.com
• Press / MediaCommunicationspress@scale.com

Sources:LeadIQ — Scale AI email format (First.Last dominant)Scale AI — contact / about

How much funding has Scale AI raised?

Scale AI raised about $1.6 billion in venture equity across roughly six priced rounds from 2016 to 2024, reaching a ~$13.8 billion valuation. In June 2025 Meta invested $14.3 billion for a ~49% non-voting stake, valuing Scale at about $29 billion.

The early rounds were classic Silicon Valley: a 2016 Y Combinator seed, a $4.5 million Series A in May 2017 led by Accel, and an $18 million Series B in August 2018 led by Index Ventures. Scale became a unicorn with its $100 million Series C in August 2019, led by Peter Thiel's Founders Fund at a $1 billion valuation, with Coatue, Index, Spark, and Thrive participating.

The LLM era accelerated everything. A $155 million Series D in December 2020, led by Tiger Global, valued Scale at about $3.5 billion; a $325 million Series E in April 2021 (Dragoneer and Greenoaks, with Tiger) roughly doubled that to ~$7 billion. In May 2024, Accel led a $1 billion Series F that nearly doubled the valuation again to ~$13.8 billion, with NVIDIA, Amazon, Meta, Y Combinator, Thrive, Coatue, and Spark participating. In early 2025 the company also floated a tender offer near a $25 billion valuation to give shareholders liquidity.

The defining event was June 2025: Meta paid $14.3 billion for a ~49% non-voting stake, valuing Scale at roughly $29 billion and giving early backers like Accel a large cash return. The deal was structured as an investment rather than an acquisition so Scale could remain independent; founder Alexandr Wang and senior researchers moved to Meta as part of the arrangement.

Sources:Tracxn — Scale AI funding & investorsTechCrunch — Series F, $13.8B valuationYahoo Finance — Meta pays ~$15B for Scale stake

How did Scale AI get here?

From a 2016 autonomous-vehicle labeling startup to a ~$29B AI data foundry doing ~$2B in 2025 revenue, capped by Meta's $14.3B stake and a leadership change in 2025.

Sources:Wikipedia — Scale AI historyScale AI — building for 2026 (Droege)TechCrunch — Scale lays off 14% of staff

Who are Scale AI's competitors?

Scale competes with a new wave of expert-data and RLHF providers (Surge AI, Mercor, Micro1) and with annotation/data-platform players (Labelbox, Snorkel AI, Appen) — many of which gained share after Meta's stake pushed labs to diversify suppliers.

Sources:CNBC — Mercor raises $350M at $10B valuationComputerworld — Meta stake triggers Scale customer exodusSuperAnnotate — Scale AI alternatives after the Meta deal