When AI Costs More Than People: Mercor's Bold New Reality
A seismic shift is quietly unfolding inside one of Silicon Valley's fastest-growing startups. Mercor, the $10 billion AI labor marketplace, has crossed a threshold that few companies have publicly acknowledged: it now spends more money on AI tokens — the computational currency that powers large language models — than it does on paying its human employees. For anyone tracking the intersection of artificial intelligence and the future of work, this is not just a headline. It is a warning shot.
Mercor CEO Brendan Foody made the revelation during an appearance on the 20VC podcast, hosted by Harry Stebbings. When Stebbings pressed him on whether AI token expenditure had genuinely surpassed payroll costs, Foody confirmed it without hesitation: "That's correct. It's pretty incredible."
And incredible it is — both as a business milestone and as a cultural signal about where enterprise technology is headed.
What Is Mercor and Why Does It Matter?
Founded in 2023, Mercor operates as a high-tech matchmaker between human experts and AI companies that need quality training data. Its client list reads like an AI industry hall of fame, including OpenAI and Anthropic — two of the most influential organizations in the global AI race. Mercor connects these companies with skilled human annotators, evaluators, and specialists who help train and fine-tune AI models.
As of October 2025, the company employed roughly 300 people, according to PitchBook data. That figure makes the token-versus-salary revelation all the more striking: a company with a relatively lean headcount has still managed to push its AI compute bill above its total payroll. The math suggests that Mercor's AI agents are, in a very tangible sense, doing the work of a large additional workforce — one that is paid not in dollars per hour, but in API calls per task.
How Mercor Deploys AI Agents Internally
Foody described a wide range of internal functions now handled — at least in part — by AI agents. These include:
- Project management: AI agents help coordinate workflows, timelines, and deliverables across the organization, reducing the administrative burden on human managers.
- Recruiting and candidate evaluation: Mercor has conducted more than 5 million AI-assisted interviews, using agents to screen, assess, and rank candidates at a scale that would be impossible with a human-only team.
- Accounting and finance: Routine financial tasks, reconciliations, and reporting are increasingly automated through AI-driven processes.
- Fraud detection: AI models monitor transactions and behavioral patterns to flag anomalies in real time, a task that traditionally required dedicated human analysts.
This breadth of deployment illustrates something important: Mercor is not using AI agents as a novelty or a pilot program. It has embedded them deeply into core operational functions. The result is a company that runs leaner on human capital while running heavier on computational capital — and whose cost structure reflects that trade-off in a very direct way.
A Prediction That Should Make Every Executive Pay Attention
Foody did not frame Mercor's situation as a one-off curiosity. He suggested it is a preview of where most companies are headed. Before long, he argued, the average organization could find itself spending more on AI compute than on employee salaries. That is a prediction with enormous implications for corporate budgeting, workforce planning, and the broader labor market.
If AI compute becomes a line item that rivals or exceeds payroll — historically the largest single operating expense for most service businesses — then the entire logic of how companies are structured begins to change. CFOs will need new frameworks for evaluating AI spend. HR departments will need to rethink headcount models. And workers will need to understand that their competition is no longer just other humans applying for the same jobs.
The ROI Question: Are Soaring AI Costs Worth It?
Not everyone in the technology world is convinced that escalating AI budgets are delivering proportionate returns. A growing chorus of tech leaders and investors has begun questioning whether the enormous capital being poured into AI compute is actually translating into measurable productivity gains or revenue growth. The debate has intensified as companies report ballooning infrastructure costs without always being able to point to clear, quantifiable outcomes.
For Mercor, the calculus appears favorable — at least from Foody's perspective. The company's rapid growth since its 2023 launch, its $10 billion valuation, and its ability to serve marquee clients like OpenAI and Anthropic suggest that its AI-heavy operating model is working. But Mercor also occupies a uniquely advantageous position: it exists at the very center of the AI economy, building infrastructure that AI companies themselves depend on. That context may make its aggressive internal AI adoption more sustainable than it would be for companies in less AI-native industries.
What This Means for the Future of Work
The Mercor story is a concrete data point in what has largely been an abstract conversation about AI's impact on employment. For years, researchers, economists, and technologists have debated whether AI would replace jobs, augment them, or create new categories of work entirely. Mercor's cost structure suggests a fourth possibility: that AI will simply become a dominant factor of production, sitting alongside — and in some cases above — human labor in the organizational budget.
This does not necessarily mean mass unemployment is imminent. Mercor still employs around 300 people, and those people are clearly doing work that AI agents cannot fully replicate. But it does mean that companies will increasingly make marginal hiring decisions by asking not just "should we hire another person?" but "should we spend those dollars on people or on tokens?" That is a fundamentally different calculus than the one that governed business for the past century.
A Turning Point Worth Watching
Mercor's CEO may be stating something that many executives are quietly observing but not yet willing to say publicly: the age of AI compute costs exceeding human labor costs has arrived, at least for some organizations. Whether that transition accelerates across industries or remains confined to AI-native companies will be one of the defining business stories of the next decade. For now, Brendan Foody and Mercor have put a number — and a name — to a trend that the rest of the corporate world will need to reckon with sooner than most expect.