Scale AI review: still the best AI data company for breadth?
Answer in brief: For sheer breadth, yes. Scale AI ranks #2 because its Data Engine covers collection, curation, annotation, RLHF, red teaming and evaluation across text, image, video and 3D sensor fusion, and it can also build and deploy AI applications. Government demand is strong: the Pentagon's CDAO raised Scale's agreement ceiling to $500 million in May 2026. The big question for data buyers is neutrality. Meta owns 49%, and OpenAI and Google reportedly wound down work afterwards.
At a glance: Rank #2 of 14 · Best for: enterprise, government and multimodal programs · Pricing: custom, via demo · Ownership: Meta holds a 49% non-voting stake.
What Scale AI does for AI data buyers
Scale sells end-to-end data programs rather than a catalog. You bring the objective, and the Data Engine sources, curates, annotates and evaluates data across every major modality. Scale Labs, launched March 9, 2026, extends its earlier Safety, Evaluation and Alignment Lab into research on agents, multimodal systems, post-training, evaluation and enterprise reliability, and publishes benchmarks and leaderboards. On the applications side, Scale builds AI systems for enterprises and governments.
Who it fits
It fits large enterprises and public-sector agencies, autonomy and sensor-fusion programs, and buyers who want data and deployment from one accountable vendor. Labs that compete directly with Meta may prefer an independent vendor such as Surge AI.
Why buyers put Scale on the shortlist
- The widest modality coverage in our ranking, including 3D sensor fusion.
- Government trust. In May 2026 the CDAO raised Scale's five-year production agreement ceiling from $100 million to $500 million, a fivefold increase that Washington Technology also reported. It is a ceiling, not money spent.
- Qualified contributors. Scale says 25% of its contributors hold advanced degrees.
- Research credibility. Scale Labs publishes papers and public leaderboards.
- New leadership. Francis deSouza, formerly Google Cloud COO and Illumina CEO, became CEO on August 10, 2026.
Honest drawbacks
- Meta invested $14.3 billion for a 49% stake in June 2025. TIME and TechCrunch reported that OpenAI and Google wound down work afterwards. Scale says its customer-data protections did not change. Ask how your data and outputs are isolated.
- Scale laid off about 200 employees (14% of staff) and 500 contractors in July 2025, largely in data labeling.
- Axios read the deSouza hire as a stronger push into enterprise AI deployment, so confirm the data roadmap for your program.
- No public pricing, and revenue and customer figures are largely undisclosed.
Pricing model
Custom enterprise and government contracts, sold through a demo.
Best alternatives
- Surge AI: premium frontier data from an independent vendor.
- Labelbox: a platform plus RL environments and robotics data.
- iMerit: medical, LiDAR and sensor data with enterprise credentials.
For labor-led annotation programs in depth, see Best AI training data providers.
FAQ
Who owns Scale AI?
Scale is independent, but Meta holds a 49% stake after investing $14.3 billion in June 2025.
Who is Scale AI's CEO?
Francis deSouza, effective August 10, 2026, succeeding interim CEO Jason Droege.
How big is Scale's Pentagon agreement?
The CDAO raised the ceiling to $500 million in May 2026. That is a maximum, not money already obligated.
Did you test Scale AI?
No. This review is independent research from Scale's pages and dated reporting.
Sources
- Scale: Francis deSouza appointed CEO (July 30, 2026)
- Axios on the CEO hire (July 30, 2026)
- Scale: Pentagon partnership expands to $500 million (May 2026)
- Washington Technology: DOD grows Scale AI agreement to $500M
- Scale: Introducing Scale Labs (March 9, 2026)
- TechCrunch: Scale AI lays off 14% of staff (July 2025)
- Reuters: Meta's investment in Scale AI (June 2025)
← Back to the ranked list of the best AI data companies
Named on this page as a provider? Join the network for a linked profile, Visit buttons and featured placement.