Mercor Doubles to $20B With $750M Series D
By @sharedot · · 7 pages
- Startups
- Venture Capital
- AI Training Data
- Mercor
AI training startup Mercor raised $750 million at a $20 billion valuation, doubling its worth in thirteen months as it moves beyond frontier labs.
What happened: a $750M round at $20B
Mercor, the San Francisco company that pays human experts to train and evaluate AI models, announced a $750 million Series D on November 10 at a $20 billion valuation. The round doubles the company's worth in just thirteen months: its October 2025 raise valued it at $10 billion on $350 million.
Investors include Benchmark, Felicis, General Catalyst, Sapphire Ventures, NVIDIA, Salesforce and Zoom, with additional participation from funds managed by affiliates of Apollo, G Squared, StepStone Group and XTX Ventures, part of which came through an employee tender offer.
No lead investor was disclosed.
Why the doubling is surprising
Thirteen months is an unusually short span for a valuation to double at this scale, and Mercor's trajectory tracks the explosive demand for expert-driven training data.
The company claims a network of more than five million experts across over 800 occupations, supplying the lawyers, doctors, bankers and engineers whose judgment frontier labs need to make models useful.
According to Ventureburn, Mercor reports an annualized revenue run rate of more than $2 billion as of July 2026, though that figure is gross billings and contractors take a substantial margin. Co-founder Brendan Foody's framing is blunt: 'AI can only be as useful as the human brain behind it.'
The pivot: from frontier labs to every enterprise
The strategic shift explains the new capital. Until recently, frontier AI labs were Mercor's primary buyers; now the company is courting enterprises that want to evaluate and post-train existing models on their own workflows.
Co-CEO Adarsh Hiremath argues a model can score well on a public benchmark and still fail on 'the systems, policies, and edge cases that define how a specific business runs.'
The playbook includes custom benchmarks built from each company's data, post-training against them, and the Deeptune acquisition earlier in 2026, whose team spent roughly two years recreating hundreds of enterprise apps as simulated practice environments for AI agents.
The evidence and the wider capital wave
The round lands amid an extraordinary stretch of AI fundraising.
Ventureburn notes OpenAI's $11 billion raise last week at a $1.25 trillion valuation, and the same week tech.eu reports that Lightspeed and Alphabet's CapitalG backed UK chip startup Fractile in a £500 million-plus round valuing it at over £5bn ahead of its inference chips reaching the market.
Sapphire Ventures' Jai Das told Pulse 2.0 that frontier labs' dominance of expert-driven training data buying is about to change as open models become good enough for enterprises to build on, calling Mercor's expert network and evaluation environments hard to replicate.
What comes next
The $750 million will fund research, new evaluation methodologies and expansion of post-training infrastructure as Mercor chases demand from finance, healthcare, law and engineering, where AI agents must navigate specialized terminology, compliance requirements and proprietary edge cases.
The company's APEX benchmark series, designed to measure AI performance on economically valuable professional tasks, anchors that push.
The open question, as Ventureburn puts it, is whether AI agents will actually work in real business workflows — but investors have just doubled their bet that Mercor will be the front seat in the market for training AI on every job.