Three-year-old startup Mercor has quietly turned into a $10 billion middleman in AI’s data gold rush.
Instead of hiring anonymous crowdworkers to label data, AI labs like OpenAI and Anthropic are paying Mercor’s contractors — former employees of Goldman Sachs, McKinsey and white-shoe law firms — up to $200 an hour to share hard-won expertise and guide models through high-stakes tasks.
On TechCrunch’s Equity podcast, recorded live at Disrupt, Mercor CEO Brendan Foody lays out what that shift means for the future of work — and who gets access to this new kind of job.
From high-school AWS hustles to a $10B valuation
Foody traces an unlikely arc: he started out in high school consulting on AWS credits. A few years later, he’s running a company investors now value at $10 billion.
That leap is powered by a simple bet: as AI systems get more capable, the real bottleneck isn’t GPUs — it’s specialized human knowledge. Mercor’s marketplace exists to package and sell that knowledge to the labs training frontier models.
Why AI labs want experts, not crowds
The conversation walks through why AI labs are moving away from traditional crowdsourced labor toward smaller pools of high-skilled contractors.
According to Foody, Mercor sees a power-law dynamic: the top 10–20% of contractors drive the majority of model improvement. These are people who’ve already operated at the top of fields like finance, consulting and law. Their judgment and intuition become training data.
That changes the profile of who gets paid in the AI economy. Instead of gig workers doing low-paid annotation, you have elite professionals billing hourly rates that rival top-tier consulting — to help improve the very systems that might one day automate pieces of their former jobs.
Scale AI’s stumbles, Mercor’s opening
Foody also talks about how problems at Scale AI, one of the best-known data-labeling companies, created a window for Mercor.
As he tells it, those troubles pushed AI labs to look for alternatives. Mercor’s pitch — fewer workers, more expertise, better results — landed at a moment when major labs were already rethinking how they source and manage data work.
The result: a three-year-old startup suddenly sitting in the middle of billion-dollar training runs.
When employee knowledge becomes training data
One of the thorniest parts of the discussion is the gray area between an individual’s experience and a company’s trade secrets.
Mercor’s contractors are ex-employees of firms like Goldman Sachs and McKinsey. They’re being paid to encode what they know into AI systems: how a deal gets structured, how a client memo is framed, how a regulatory risk is evaluated.
Where does personal know-how end and proprietary information begin? And should a bank like Goldman Sachs be worried that its former staff are effectively teaching AI agents how to do their old jobs?
Foody and the Equity team dig into that tension without pretending there are easy answers. The incentives are clear: contractors want to monetize their skills; AI labs want the sharpest possible training data. Legal and ethical norms are still catching up.
All knowledge work as training data
Foody doesn’t hedge on where he thinks this is going. In his view, all knowledge work is on a path to becoming training data for AI agents.
Decks, memos, models, strategy docs — anything a knowledge worker touches is potential input for systems that will increasingly assist, augment or automate that same work.
If he’s right, the economy doesn’t just adopt AI tools; it reorganizes around feeding them. Mercor’s contractors are an early, well-paid slice of that future, but the logic extends far beyond finance and consulting.
Whether that ends up empowering workers or hollowing out careers depends on who captures the value of all that expertise as it turns into data.
Hear the full conversation
The Equity episode goes deeper into:
- How Foody thinks about finding and vetting the top 10–20% of contractors
- What Mercor’s rise says about where the AI labor market is headed
- How he expects other industries to plug into the AI training supply chain
You can listen to the full episode of Equity, TechCrunch’s flagship podcast, on YouTube, Apple Podcasts, Overcast, Spotify and other major platforms.
