- Mercor is in talks to raise a new round at a $20 billion valuation, double what it was worth 9 months ago, while announcing the acquisition of AI training environment startup Deeptune.
- Three founders who were 22-year-old high school debate teammates when they started the company in 2023 are now reportedly the world’s youngest self-made billionaires.
- The $2 billion annualised revenue run rate announced this week by CEO Brendan Foody reflects gross billings: contractors who do the work take home 60-70% of that figure, per Bloomberg.
In March 2026, a supply chain attack on LiteLLM, an open-source Python library, exposed up to four terabytes of Mercor’s internal data and contractor records. Meta, one of its biggest clients, paused all work with the startup indefinitely. Class action lawsuits followed. Forbes reported that former employees suspected North Korean operatives had infiltrated the contractor network using stolen credentials. For most startups, any one of those events would be existential.
Mercor is now reportedly in talks to raise at a $20 billion valuation.
Bloomberg reported that Mercor, the San Francisco-based AI training marketplace, has told investors it already received at least one term sheet at the new figure. The conversations are early and may not result in a round, the outlet cautioned.
But the mere existence of the talks, nine months after a $10 billion Series C in October 2025 and roughly four months after the breach, signals something remarkable about how quickly the AI training market has moved on.
Who built this, and how fast
Brendan Foody, CEO, Adarsh Hiremath, CTO, and Surya Midha, board chairman, founded Mercor in 2023 after dropping out of college. They were 20. All three had been high school debate teammates at Bellarmine College Preparatory in San Jose and later received Thiel Fellowships. As Tech Funding News reported when Mercor raised its $350 million Series C at a $10 billion valuation, the company went from $1 million to $1 billion in annualised revenue in 20 months, making all three founders, at 22, the world’s youngest self-made billionaires, according to Forbes.
Mercor’s model is straightforward: it recruits domain experts, such as engineers, lawyers, doctors, bankers, journalists, and places them with AI labs that need human intelligence to train frontier models. Clients have included OpenAI, Anthropic, Meta, and Google.
The company manages approximately 30,000 contractors and has vetted more than five million candidates through its proprietary AI interviewing platform. Total funding stands at roughly $492 million across four rounds, the most recent being the Series B and Series C led by Felicis Ventures, with Benchmark and General Catalyst participating. Mercor’s own figures put full-time headcount at around 400; the 30,000-plus figure refers to its rotating contractor pool, not employees.
Now, alongside the valuation talks, Mercor announced it is acquiring Deeptune, a startup that builds software environments where AI agents practise real workflows before deployment. Deeptune has recreated hundreds of enterprise applications, from spreadsheets to Salesforce, for use in reinforcement learning.
Andreessen Horowitz led Deeptune’s $43 million Series A in March 2026 — a round Foody himself invested in as an angel before Mercor acquired the company three months later. The entire Deeptune team is joining Mercor in New York. Public reporting hasn’t established whether Mercor’s board or outside investors reviewed Foody’s personal stake ahead of the acquisition, or how any gain to him was handled.
The revenue number needs context
Foody posted on X this month that Mercor’s annualised revenue run rate has crossed $2 billion, up 100% in four months.
That figure is real but requires a qualifier Bloomberg buried: contractors take home 60-70% of everything billed. Mercor’s net revenue after paying the experts who do the work is closer to $600- $800 million. At a $20 billion valuation, that implies a multiple of roughly 25-33x net revenue — aggressive, but not exceptional for a company growing at this rate in this market.
The competitive set includes Scale AI, which reached a $13.8 billion valuation in 2024 and now sits at $29 billion following Meta’s stake purchase, and Outlier AI, Scale’s own generative-AI contributor arm, which raised $200 million in 2024.
As TFN has reported on the AI agent infrastructure boom, the global AI agents market is projected to reach $52.62 billion by 2030, and training those agents at scale is becoming one of the largest categories of enterprise spending.
Mercor’s stated differentiators are the seniority of its expert network and its APEX benchmark, which measures AI agent performance against real-world professional tasks, positioning the company as an evaluator of AI capabilities rather than just a data supplier.
The unanswered question
The Deeptune acquisition makes Mercor’s ambition explicit: it wants to own the full stack of AI agent training — the environments where agents practise, the experts who grade them, and the benchmarks that define what good looks like.
Whether that vertical integration creates a durable moat, or simply makes Mercor a more attractive acquisition target for one of the frontier labs it currently supplies, is a question a $20 billion valuation asks investors to answer. Foody’s undisclosed stake in the company he just had Mercor buy is the question a $20 billion price tag asks the board to answer first.
