Mercor, a startup specializing in sourcing and hiring "AI trainers," is targeting a $20 billion valuation. According to Bloomberg, the company has already received a term sheet at this figure, despite its capitalization sitting at a "modest" $10 billion as recently as October following a $350 million Series C round. Doubling in value in just a few months might seem like madness, were it not for the underlying revenue metrics.
Explosive Growth and Aggressive Consolidation
Brendan Foudy, Mercor’s founder and CEO, officially confirmed on X that the company’s Annual Recurring Revenue (ARR) has surpassed $2 billion. This represents 100% growth in just four months. Against this backdrop, a $20 billion valuation looks less like a venture capital bubble and more like a pragmatic 10x revenue multiple. To solidify its lead in the AI agent race, Mercor has also acquired Deeptune. As Foudy explained, the entire Deeptune team is joining Mercor to scale technical capabilities in model training.
Brendan Foudy reported that Mercor's annual revenue run rate has crossed $2 billion—the company doubled its key metrics in the last quarter.
Industrializing Data Labeling: Human Brains as a Scarce Resource
The market is willing to pay massive premiums for infrastructure that hires high-level specialists. The era of cheap outsourcing for simple image labeling is fading. Today, OpenAI, Google, and Anthropic are battling for elite AI trainers—scientists, developers, and linguists capable of complex data labeling and Reinforcement Learning from Human Feedback (RLHF). Mercor is essentially building a "digital mine" where human expertise is extracted like a finite and expensive fossil fuel for neural networks.
Judging by the momentum, investors are quick to forgive the company's past operational missteps. In early 2026 (per the company's fiscal timeline), Mercor faced data leaks and contractor lawsuits, as reported by Business Insider. However, the market’s appetite for "AI oil" has proven stronger than reputational risks. This is a classic arms race: as long as synthetic data and self-learning methods cannot fully replace a living expert, the valuation of such intermediaries will continue to grow exponentially. Nevertheless, the business model carries the inherent risk of a sudden collapse the moment technology learns to operate without the "crutch" of the human brain.