Mercor in Talks for $20B Round, Double October Valuation

Nine months ago Mercor closed a $350 million Series C at a $10 billion valuation. This week, per Bloomberg reporting picked up by TechCrunch, the AI training startup is in early discussions to raise again at roughly $20 billion, and has told investors it already received at least one term sheet at that number.

The story hinges on one figure Mercor itself put in public. Founder and CEO Brendan Foody posted on X that annualized revenue run rate has surpassed $2 billion, which he framed as a 100% increase from four months earlier. If that number holds, the round would price the company at roughly ten times its run rate. On the same day the valuation talk surfaced, Mercor announced it was buying Deeptune, a company that helps train AI agents, with the entire team folding into Mercor. The acquisition matters here because it points at where Mercor is trying to be next, not just the human labelers behind reinforcement learning, but the humans who train agents.

Why this matters if you are not writing checks: the market is pricing the human infrastructure behind frontier model training as durable, not as a stepping stone that gets automated away. That is a real bet, and it has knock-on effects for anyone who sources labeling work, hires domain experts as contract raters, or plans around the assumption that model providers will bring this in house.

The honest caveats are that the talks are early and the company could end up not raising at all or on different terms, the $2 billion figure comes from Foody's own X post rather than an audited disclosure, and Business Insider previously reported that Mercor had a data breach in April and lawsuits filed by contract workers earlier in the year, neither of which this round of reporting revisits. What the reporting does not give you is which fund wrote the term sheet, what Mercor paid for Deeptune, or how the lawsuits are progressing.

For competitors in the AI training market, a doubled valuation on this timeline is a signal to accelerate their own agent-training moves, or to prepare for talent poaching in the specialized labeler pool.