Key Takeaways
- • Mercor acquires Deeptune in an undisclosed all-cash deal, consolidating AI training infrastructure as Mercor CEO Brendan Foody had angel-invested in Deeptune's $43 million Series A just three months prior, explicitly stating acquisition was his motivation.
- • Deeptune builds reinforcement learning simulation environments where AI agents practice real-world enterprise software tasks (Excel, Salesforce, Slack) before deployment; Mercor operates a $10 billion platform of 5+ million domain experts who create training tasks and scoring rubrics, combining to cover the full AI agent training stack.
- • Mercor's valuation and growth trajectory accelerated significantly, reaching $2 billion in annual recurring revenue (ARR) in June 2026, up from $1 billion a year prior—representing the company's characterization of achieving "$1 million to $2 billion ARR in 24 months" as industry-leading velocity with customers including Anthropic, OpenAI, and six of the Magnificent Seven tech companies (excluding Tesla).
- • Series A funding round led by Andreessen Horowitz with participation from 776, Abstract Ventures, and Inspired Capital; Deeptune's founding team, led by CEO Tim Lupo, will relocate to Mercor's New York headquarters post-acquisition.
- • Operational synergy addresses frontier AI lab demand for full digital replicas of enterprise software environments where large language models and AI agents can iterate and learn at scale before production deployment.
Summary
Mercor, a $10 billion AI infrastructure platform, has acquired Deeptune, an Andreessen Horowitz-backed startup specializing in reinforcement learning simulation environments. The transaction closed without disclosed financial terms, though the deal represents a strategic vertical integration within the rapidly expanding AI agent training ecosystem.
Deeptune's core technology creates high-fidelity simulation environments that replicate enterprise software workflows, allowing AI agents to practice multi-step tasks across platforms including Slack, Salesforce, Excel, and other workplace tools before execution in production systems. Mercor operates a distributed network of over five million domain experts who design training tasks and establish scoring rubrics used to train foundational AI models. The acquisition consolidates both components—simulation infrastructure and expert-generated training data—under a single platform, addressing demonstrated demand from frontier AI labs including Anthropic and OpenAI.
The transaction's timing reveals internal conviction: Mercor founder and CEO Brendan Foody served as an angel investor in Deeptune's $43 million Series A in March 2026, three months before the acquisition closed. Foody disclosed that the initial investment was structured explicitly with acquisition in mind, stating the check "was in a lot of ways the main motivation." The Series A, led by Andreessen Horowitz with co-investment from 776, Abstract Ventures, and Inspired Capital, established early-stage valuation metrics before the subsequent acquisition.
Mercor's financial performance has accelerated substantially. The platform achieved $2 billion in annual recurring revenue in June 2026, doubling from $1 billion approximately 12 months prior. Foody characterized this 24-month trajectory from $1 million to $2 billion ARR as "the fastest growth trajectory ever" in the category and claims Mercor maintains market leadership in building reinforcement learning environments, with six of the Magnificent Seven technology companies (all except Tesla) as customers. Deeptune's team, led by founder and CEO Tim Lupo, will integrate into Mercor's New York operations following closing.
Notably, Mercor experienced a significant security incident in March 2026 when hackers exploited a supply chain vulnerability in LiteLLM and exfiltrated approximately four terabytes of data, including contractor identifying information and internal records on Anthropic's safety labeling processes. The group Lapsus$ claimed responsibility and subsequently offered the data for sale, triggering class action litigation in April 2026. This incident precedes the Deeptune acquisition by several months.
Deeptune's core technology creates high-fidelity simulation environments that replicate enterprise software workflows, allowing AI agents to practice multi-step tasks across platforms including Slack, Salesforce, Excel, and other workplace tools before execution in production systems. Mercor operates a distributed network of over five million domain experts who design training tasks and establish scoring rubrics used to train foundational AI models. The acquisition consolidates both components—simulation infrastructure and expert-generated training data—under a single platform, addressing demonstrated demand from frontier AI labs including Anthropic and OpenAI.
The transaction's timing reveals internal conviction: Mercor founder and CEO Brendan Foody served as an angel investor in Deeptune's $43 million Series A in March 2026, three months before the acquisition closed. Foody disclosed that the initial investment was structured explicitly with acquisition in mind, stating the check "was in a lot of ways the main motivation." The Series A, led by Andreessen Horowitz with co-investment from 776, Abstract Ventures, and Inspired Capital, established early-stage valuation metrics before the subsequent acquisition.
Mercor's financial performance has accelerated substantially. The platform achieved $2 billion in annual recurring revenue in June 2026, doubling from $1 billion approximately 12 months prior. Foody characterized this 24-month trajectory from $1 million to $2 billion ARR as "the fastest growth trajectory ever" in the category and claims Mercor maintains market leadership in building reinforcement learning environments, with six of the Magnificent Seven technology companies (all except Tesla) as customers. Deeptune's team, led by founder and CEO Tim Lupo, will integrate into Mercor's New York operations following closing.
Notably, Mercor experienced a significant security incident in March 2026 when hackers exploited a supply chain vulnerability in LiteLLM and exfiltrated approximately four terabytes of data, including contractor identifying information and internal records on Anthropic's safety labeling processes. The group Lapsus$ claimed responsibility and subsequently offered the data for sale, triggering class action litigation in April 2026. This incident precedes the Deeptune acquisition by several months.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.