Mercor reportedly seeks $20B valuation as AI training scales
Mercor’s reported valuation talks and revenue growth show how quickly the expert-data and agent-training layer behind frontier AI is scaling.
The infrastructure behind model training and evaluation is becoming a major business category of its own.
What happened
Mercor is reportedly in early talks to raise capital at a $20B valuation, double the $10B valuation from its October financing.
The financing is not closed.
The company has also said its annualised revenue run rate crossed $2B and announced the acquisition of agent-training startup Deeptune.
Why it matters
Frontier AI systems increasingly need specialised human expertise, evaluation and training workflows.
That creates large businesses around the model-improvement layer rather than the models themselves.
Mercor’s reported valuation talks suggest investors see expert-data and agent-training infrastructure as capable of supporting very large standalone companies.
The bigger picture
AI value is spreading across the training stack.
The market is no longer only about who builds the best model. Companies that organise human expertise, evaluation and reinforcement workflows may become equally strategic as agents grow more specialised.