Mercor.io has acquired Deeptune Inc., a startup that builds simulated training environments for AI agents, in a deal announced today with undisclosed financial terms. Mercor CEO Brendan Foody had invested in Deeptune's $43 million（約69億円） Series A round in March with the acquisition already in mind. The purchase addresses what Foody identifies as the current bottleneck in AI model training: the need for realistic simulated environments where agents can practice and be scored on real-world tasks. Mercor supplies the human layer—domain experts who write tasks and verify results—while Deeptune supplies the software environments in which those tasks run.
Summaries like this, in your inbox every morning.
Sign up free →What happened
Mercor.io Corp. has acquired Deeptune Inc., a startup that builds simulated software environments—called "training gyms"—used to train AI agents. Financial terms were not disclosed. The deal closed nearly four months after Mercor CEO Brendan Foody invested a personal angel check in Deeptune's $43 million（約69億円） Series A round in March, which he said was made with the acquisition already in mind.
Why it matters
Mercor frames the deal as addressing the current bottleneck in AI model training. While reinforcement learning has advanced to the point where models can learn nearly any clearly-defined, scorable task, the constraint has shifted to training environments themselves. Deeptune's simulated software environments—which recreate hundreds of enterprise applications like spreadsheets and Salesforce—complement Mercor's existing layer of domain experts who write tasks and verify agent performance, creating a more complete training pipeline.
What to watch
Mercor is in early talks to raise funding at a valuation of about $20 billion（約3.2兆円）, double the $10 billion（約1.6兆円） valuation from its October Series C (which raised $350 million（約560億円）). The company reported annualized revenue crossed $2 billion（約3200億円） in June, up 100% in four months. Deeptune's team is joining Mercor in New York.
The acquisition reflects a strategic shift in how the AI industry approaches model training. Mercor CEO Brendan Foody's framing—that reinforcement learning has matured to the point where the bottleneck is no longer algorithmic but environmental—suggests that frontier AI labs are moving beyond theoretical optimization into practical, task-based training. By combining Deeptune's simulated software environments with Mercor's existing network of more than 5 million domain experts (who write scoring rubrics and verify agent performance) and its APEX benchmarks, the combined entity creates a vertically integrated training platform. Foody's statement that Mercor was already a customer of Deeptune underscores how the acquisition consolidates capabilities both companies were building separately.
The timing and structure of the deal—Foody's angel investment three months prior, with acquisition already in mind—suggests a deliberate acquisition strategy rather than an opportunistic one. This approach allowed Mercor to evaluate Deeptune's technology and team before committing to a full purchase. The fact that Deeptune was already supplying environments to frontier labs (including Mercor) validates the market need for this infrastructure.
Mercor's financial momentum is notable: the company's reported 100% four-month revenue growth to $2 billion（約3200億円） annualized, combined with a pending fundraise at a $20 billion（約3.2兆円） valuation, positions it as one of the most rapidly valued infrastructure players in AI. However, the company faced a significant security incident in March when attackers using malware in the open-source LiteLLM library compromised its systems, resulting in the theft of contractor personal data and biometric information. Foody's statement that "every frontier lab has expanded their relationship with us since the data breach" suggests the incident has not undermined customer confidence, though contract workers have filed suit over the exposure.
No comments yet. Be the first to share your thoughts!
Log in to join the discussion
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
1 minute a day. The AI essentials.
200+ sources · Email / LINE / Slack
