Particle.news

Mercor Acquires Deeptune to Add Simulation Sandboxes to Its AI Training Stack

The purchase gives Mercor virtual environments where AI agents can practice real workflows, concentrating control of both expert labeling and simulated training.

Overview

  • The deal, announced Thursday, July 9, brings Deeptune’s simulation technology into Mercor and relocates the Deeptune team to New York as part of the acquisition.
  • Mercor CEO Brendan Foody confirmed he had made an angel investment in Deeptune during its March Series A and said the check was written with the acquisition in mind, a fact that raises governance questions about insider-funded deals.
  • The company reports roughly $2 billion in annualized revenue as of June and is in talks to raise about $500 million at a possible $20 billion valuation, a fundraising process that sources say is still developing.
  • Mercor’s move closes a technical gap for agent training by pairing human expert labeling with realistic simulated sandboxes where agents can learn to use enterprise apps without touching production systems.
  • Operational and due-diligence risks remain: Mercor disclosed a March supply-chain breach tied to LiteLLM that exfiltrated large volumes of contractor and internal data and prompted a class-action suit, and investors will likely scrutinize security and deal governance as the company pursues higher valuation rounds.