Mercor’s $10 Billion Bet on the Last White-Collar Paycheck

The LinkedIn message arrives with the polite, hollow cadence of a corporate recruiter, but the job description contains a structural glitch. It invites you to a high-paying, short-term project—coding, legal analysis, or financial modeling—provided you are willing to have your every keystroke, correction, and creative pivot recorded. You are not being hired to work; you are being hired to act as a biological training set for a machine that will eventually render your role a legacy cost. This is the core engine of Mercor, a San Francisco startup that recently cleared a $10 billion valuation despite being led by two founders who have never actually held a traditional white-collar job.

The data harvest hidden in a job post

The founders, Adarsh Hiremath and Brendan Foody, represent a specific Silicon Valley archetype: the pure-play engineer who views the complexities of professional life as a series of optimization problems. Their $10 billion valuation, backed by the wealth-management elite at Iconiq, reflects a gamble that the next leap in productivity won't come from better algorithms, but from the ownership of professional 'expert' data. It is a cynical loop. You pay a human $100 an hour to train a model for forty hours; that model then performs the human’s job for cents on the dollar for the next decade. The unit economics are devastating for anyone currently holding a desk and a pension plan.

Can AI find work for those it displaces?

As Mercor accelerates the automation of the mid-level professional, a secondary market is emerging to manage the fallout. Pelgo, another startup surfacing in the wake of the current AI boom, aims to use artificial intelligence to find new jobs for the very workers displaced by it. It is a symmetry that only a venture capitalist could find comforting. Pelgo promises to provide outplacement services—resume optimization, interview coaching, and automated application filing—cheaper and faster than human-led career coaching. We are entering an era where an AI fires you, and another AI helps you apply for a job that may or may not be managed by a third AI.

This cycle reveals a mounting pressure within the tech industry to justify its own disruption. If the 'myth' of AI, as Anthropic’s new Mythos tool was recently presented at IMF meetings, is one of total efficiency, the reality is a messy, high-friction transition. In Washington D.C., financial chiefs are no longer debating whether AI will impact the workforce; they are trying to calculate how much the resulting social safety nets will cost. The conversation has shifted from the laboratory to the treasury, and the numbers are not particularly flattering for the human element.

The European pivot to hardware and light

While the United States dominates the software layer of this upheaval, Europe is attempting to secure the foundation. In a year where software-as-a-service stocks have faced skepticism over their actual utility, Europe’s best-performing stock of 2026 isn't a generative model provider. It is a French photonics firm riding the AI infrastructure wave. Photonics—the use of light rather than electricity to move data—has become the critical bottleneck for companies like Mercor and Anthropic. As models grow, the energy cost of moving electrons through copper wires is becoming unsustainable.

The French surge in photonics stocks highlights a divergence in industrial strategy. Brussels has largely accepted that it lost the race for the dominant large language model to San Francisco and Seattle. Instead, the EU is doubling down on the supply chain. Through the EU Chips Act and various sovereign investment vehicles, the focus has shifted to the specialized hardware that makes AI physically possible. If the US owns the mind of the machine, Europe wants to own the nervous system. This is a pragmatic, if less glamorous, bet on the physics of the data center rather than the whims of the LinkedIn algorithm.

The regulatory wall in Brussels

Mercor’s strategy of using professionals to train their own replacements is likely to run into the sharp edges of the EU AI Act. The regulation, which has moved from theory to enforcement, includes specific provisions regarding transparency and the rights of workers. Under current EU law, a platform that uses professional output to train a commercial model must be explicit about that intent. There is a growing legal debate in Bonn and Paris over whether 'data harvesting' disguised as 'employment' constitutes a breach of labor rights or a violation of intellectual property.

German industrial leaders, particularly in the Mittelstand, view these developments with a mixture of technical curiosity and institutional horror. The German model of 'Mitbestimmung' (co-determination) does not easily accommodate a software system that treats an engineer’s expertise as a disposable dataset. While a San Francisco startup can pivot into a $10 billion valuation overnight by bypassing traditional employment structures, a European firm attempting the same would face a mountain of litigation from works councils and data protection authorities. This regulatory friction is often cited as a weakness, but in the context of Mercor, it may serve as a critical brake on a process that treats human experience as a raw commodity.

The infrastructure of obsolescence

The capital flight toward AI is not just about the code; it is about the wealth management behind it. Iconiq, the advisor to the tech industry’s billionaire class, is shifting billions into AI infrastructure. This is not speculative retail money; this is the capital of the people who built the social media platforms and cloud providers of the last decade. They are signaling that the era of the 'human-in-the-loop' is a temporary transition phase. Once the data from Mercor’s high-paid gig workers is sufficiently ingested, the loop will close.

This creates a supply-chain tension that is often overlooked. To sustain a $10 billion valuation, Mercor needs a constant stream of high-quality professional data. But as the machines get better at the tasks, the pool of human experts shrinks. We are witnessing the cannibalization of professional expertise. The irony of the situation is that the more successful Mercor is at replacing the white-collar workforce, the more expensive its own training data becomes. It is an extractive industry that, like oil, faces a peak-resource problem. Once you have automated the junior lawyer, who is left to become the senior lawyer whose expertise you need for the next model update?

Silicon Valley has spent the last decade building tools to manage the world's information. Now, it is building tools to manage the world's absence from the workforce. Europe has the photonics and the regulations, but it hasn't yet decided if it wants to protect the workers or just provide the light for their replacement. For now, the professional class is being paid to document its own obsolescence, one LinkedIn message at a time. It is a remarkable feat of engineering, provided you aren't the one holding the keyboard.