Introduction
Scale AI builds data-labeling and curation infrastructure that AI labs and large enterprises use to prepare the training and evaluation data underlying machine-learning models. Founded in 2016 by Alexandr Wang and Lucy Guo, the company grew into one of the most important, if largely behind-the-scenes, suppliers to the frontier AI industry, providing labeled data to companies including OpenAI and Google before its business relationship with the AI industry changed dramatically in mid-2025.
In June 2025, Meta announced a $14.3 billion investment for a 49% stake in Scale AI, a deal that valued the company at $29 billion and came paired with an unusual condition: Scale AI’s young CEO, Alexandr Wang, would leave to lead a new AI research division inside Meta, while Scale AI itself continued operating as a separate, non-controlled company.
What the company does
Scale AI provides the data infrastructure that sits underneath model training: recruiting and managing large workforces to label images, transcribe and annotate text, and generate or refine examples used to train and fine-tune machine-learning models, alongside evaluation tools that help AI labs measure how well a model performs on specific tasks. Because large language models depend heavily on the quality and diversity of their training and reinforcement-learning data, Scale AI’s role — though rarely visible to end users of AI products — has been structurally important to the pace of progress across the AI industry.
Who it serves
Scale AI’s customers are AI labs and large enterprises that need substantial volumes of accurately labeled or curated data, including frontier model developers such as OpenAI and Google, which reportedly relied on Scale AI’s data-labeling operations before Meta’s investment, alongside government and defense customers using its infrastructure for specialized applications. Following Meta’s investment, Scale AI has continued to operate as an independent supplier to the broader AI industry rather than becoming an exclusive Meta subsidiary.
Company background
Alexandr Wang and Lucy Guo founded Scale AI in 2016 in San Francisco, building a business around the labor-intensive but critical task of labeling data for machine-learning models at a time when few investors viewed data infrastructure as a particularly glamorous category. The company grew steadily through the 2010s and into the generative AI boom of the 2020s, raising a $1 billion round in 2024 from investors including Amazon and Meta at a $13.8 billion valuation. In June 2025, Meta announced it would invest $14.3 billion for a 49% stake in Scale AI, valuing the company at $29 billion in what The New York Times described as Meta’s second-largest deal ever, after its $19 billion acquisition of WhatsApp. As a condition of the investment, Wang, then 28, left his CEO role to join Meta and lead a new internal division, initially called Meta Superintelligence Labs, while remaining on Scale AI’s board; Jason Droege, Scale AI’s chief strategy officer, became interim CEO of the company, which Meta said would remain an independent entity notwithstanding its large, non-controlling stake.
Product and AI capabilities
Scale AI’s technology centers on human-in-the-loop data pipelines: recruiting, training, and managing distributed workforces that label, annotate, and evaluate the raw material — text, images, and other data — that machine-learning models are trained on, alongside reinforcement-learning-from-human-feedback data used to align model behavior with human preferences. Following Meta’s investment, Wang’s new Meta Superintelligence Labs released Muse Spark in April 2026, its first model since the deal, a natively multimodal reasoning model that Meta has said is the first step in a broader scaling program rather than an endpoint. While Muse Spark is a Meta product rather than a Scale AI product, its release is widely viewed as an early proof point for whether the unusual Meta–Scale AI arrangement — investment and talent acquisition without a full takeover — accelerates Meta’s underlying AI research the way the deal was designed to.
Key developments
Meta announced its $14.3 billion investment for a 49% stake in Scale AI on June 12, 2025, with the deal confirmed and CEO Alexandr Wang’s departure to Meta announced the following day; Jason Droege became Scale AI’s interim CEO. On April 8, 2026, Meta Superintelligence Labs, led by Wang, released Muse Spark, its first model developed since the investment closed. Scale AI has continued operating as an independent company through 2026, supplying data infrastructure to AI labs and enterprises beyond Meta under the terms of Meta’s non-controlling stake.
Why it matters
Scale AI’s role in the AI industry is a useful reminder that model quality depends heavily on data quality, and that the companies producing that training data can be just as strategically important as the labs building the models themselves — a dynamic Meta’s unusually structured investment made explicit. The deal’s design, buying influence and talent without full ownership, is also a notable template that other large technology companies may look to as they try to secure access to critical AI suppliers without the regulatory scrutiny or cultural risk of an outright acquisition.
Sector context
Within Brel’s enterprise software coverage, Scale AI sits further upstream in the AI supply chain than most other companies profiled here, supplying the training-data infrastructure that underlies model development rather than selling finished enterprise applications. Its data-quality focus is nonetheless closely tied to model-building platforms such as Databricks and Hugging Face, both of which depend on well-curated training and fine-tuning data to support the custom and open-source models their customers build.
Sources and references
This profile draws on contemporaneous reporting from The New York Times and TechCrunch on Meta’s investment and Alexandr Wang’s departure, supplemented by Forbes’ later analysis of the deal’s strategic rationale.
- The New York Times — Meta–Scale AI investment coverage (2025)
- TechCrunch — Scale AI confirms Meta investment (2025)
- Forbes — Meta–Scale AI deal analysis (2026)