• Scale AI raised $1B in June 2024 and a year later Meta paid $14.8B for 49%.
  • The founder left for Meta, and Scale became something entirely different.

Scale AI Raised $1 Billion to Label Data. Then Meta Paid $14.8 Billion for Half the Company.

The June 2024 Series F was the most visible moment in Scale AI's history to that point — a billion-dollar round at a $13.8 billion valuation, led by Accel with backing from Y Combinator, Founders Fund, Nvidia, Amazon, Cisco, Meta, and Intel Capital. The investor list read like a roll call of everyone with a stake in AI's future, which was exactly the point. Scale AI had positioned itself as the infrastructure layer that all of them depended on.

Twelve months later, Meta returned — and paid more than fourteen times what it had previously invested. In June 2025, Meta Platforms announced a significant new investment valuing Scale at over $29 billion, acquiring a 49% non-voting stake for approximately $14.8 billion. Former CEO Alexandr Wang departed to become Meta's Chief AI Officer, leading Meta Superintelligence Labs. Jason Droege, former Uber Eats founder and Scale's Chief Strategy Officer, assumed the Interim CEO role.

In thirteen months, Scale AI went from a billion-dollar funding round to a $29 billion acquisition target — and fundamentally changed what it is in the process.

The Founder Who Built the Data Layer for AI

Alexandr Wang's origin story is one of the more unusual in Silicon Valley. His parents were Chinese immigrant physicists who worked at the Los Alamos National Laboratory — the birthplace of the atomic bomb. Conversations at the Wang family dinner table revolved around black holes, wormholes, alien life, and supernovae. His parents taught him algebra in second grade.

Wang dropped out of MIT at eighteen, co-founded Scale AI through Y Combinator in 2016 with Lucy Guo, and built on a thesis that proved prescient: that the constraint on AI development was not models or compute, but data quality. "Data is the new code," he said, long before the phrase became a cliché. By 2021, Wang officially became the youngest self-made billionaire in the world at just 24, as Scale hit a $7.4 billion valuation.

The business model was straightforward in principle and genuinely difficult in execution: take raw, unstructured data — images, video, audio, text — and produce labelled, structured training datasets that AI models could learn from. The work required human judgment at scale, coordinated through software, quality-controlled rigorously. Scale built the software to orchestrate this; a global workforce did the labelling.

What the $1 Billion Round Was Actually Building

The June 2024 Series F arrived as Scale was generating approximately $870 million in annual revenue, with AI labs like OpenAI, Google, Meta, and Microsoft as its largest customers. Key revenue drivers included expanding contracts with major technology companies and growing government partnerships, including a $249 million contract with the US Department of Defence.

The government dimension was more significant than most AI coverage acknowledged. Scale had been selling to US federal agencies since 2020. In January 2022, Scale won a $250 million contract with federal agencies — a deal that established credibility with defence contractors, aerospace companies, banks, and healthcare providers who paid attention when the Pentagon trusted a vendor. Wang testified before Congress, met with heads of state, and positioned Scale as critical national security infrastructure.

In March 2025, Scale won the Thunderforge project with the US Department of Defence — a contract to use AI to "plan and help execute movements of ships, planes, and other assets," with the goal of speeding up military decisions in both peace and wartime. The deal was awarded alongside Anduril Industries and Microsoft by the Defence Innovation Unit. Scale later joined Palantir in a $185 billion missile defence initiative.

The Meta Deal and Its Complications

The June 2025 announcement was framed as a commercial expansion and strategic investment. The reality was more complex.

The Meta partnership created substantial customer concentration risk. Major clients including Google, OpenAI, and xAI reduced or paused engagements due to data confidentiality concerns and competitive conflicts. When the company whose data infrastructure you depend on is 49% owned by your direct competitor, the decision to continue sharing proprietary training data becomes difficult to justify internally. Several of Scale's largest non-Meta customers concluded they could not.

The leadership transition compounded the uncertainty. Wang had been Scale AI's primary relationship with every major AI lab. His departure to Meta removed the person most responsible for landing and maintaining those enterprise contracts. Droege, despite a strong operational track record from Uber Eats and Axon, inherited a company that had just lost its founder and potentially its largest non-Meta customers simultaneously.

The Pivot to Enterprise Applications and Physical AI

Scale's response has been to accelerate a transition that was already underway: moving from pure data labelling toward AI applications and evaluation services.

In March 2026, Scale launched Scale Labs, an expanded research division focused on AI model capabilities, post-training evaluation methods, enterprise deployment, and risk-oversight infrastructure. The division develops public benchmarks and evaluation frameworks including SWE-Atlas and Voice Showdown.

Scale also launched its Physical AI data collection platform for robotics and autonomy companies, leveraging over 100,000 production hours at its San Francisco prototyping laboratory. Customers include Physical Intelligence, Generalist AI, and major autonomous vehicle companies. Unlike language models that train on web-scraped text, robotics requires physical interaction data collected in real-world environments — a category where Scale's human-in-the-loop infrastructure has a structural advantage.

Revenue is projected to exceed $1 billion in 2026, with the applications business generating $200–300 million and the data business described by CEO Droege as "very, very large."

What This Means Beyond Silicon Valley

Scale's data labelling workforce has operated globally, with contributors on its Outlier platform earning over $500 million in the last year across 9,000 towns in the US alone. The global picture includes significant labelling workforces in South and Southeast Asia, where AI data annotation has become a meaningful source of digital employment.

In February 2025, Scale AI agreed to a five-year partnership with the Qatari government to improve government services via AI-based tools and training. The deal, signed at the Web Qatar 2025 Summit, signals Scale's emerging markets ambitions beyond its defence-heavy US positioning — and the willingness of Gulf state governments to invest in AI infrastructure partnerships on terms that Western tech companies can work within.

Bottom Line

Scale AI's $1 billion Series F in June 2024 was the last moment the company looked like a data labelling business with a large government contract. The fourteen months that followed — the Thunderforge deal, the Meta investment at $29 billion, Wang's departure, the client defections, and the pivot to enterprise applications and physical AI — produced a fundamentally different company.

Revenue reached $870 million by 2024 and is projected to exceed $1 billion by the end of 2026, but the revenue mix, the customer base, and the strategic direction have all shifted. The question now is whether Scale under new leadership can replace the revenue from departed AI lab clients with government contracts and enterprise applications fast enough to justify a $29 billion valuation.

Edited By Nabarun.