Mercor buys Deeptune to build training environments for AI agents

Artificial intelligence training data company Mercor.io Corp. announced today that it has acquired Deeptune Inc., a startup that builds simulated software environments used to train AI agents. Financial terms were not disclosed.

The deal closed nearly four months after Mercor Chief Executive Brendan Foody wrote a personal angel check into Deeptune’s $43 million Series A round, which Andreessen Horowitz led in March. Foody told Fortune that the investment was made with a purchase already in mind. “It was in a lot of ways the main motivation, actually,” he said.

Deeptune is based in New York and builds what Chief Executive Tim Lupo calls “training gyms,” reinforcement learning environments that copy the software knowledge workers use every day. An agent drops into a simulated spreadsheet or a Salesforce queue, gets the task wrong and tries again. Nothing it touches is real.

Lupo has likened the setup to a flight simulator. Mercor said the team has recreated hundreds of enterprise applications over the past two years and was already supplying environments to frontier labs, with Mercor among them as a customer.

Foody framed the acquisition as a bet on where the bottleneck in model training now sits.

“Reinforcement learning has reached the point where a model can learn almost any task that can be clearly defined and scored,” he wrote in a blog post announcing the deal. “The constraint has shifted to the environments themselves: the places where models practice the work and get measured on whether they did it well.”

Mercor supplies the human layer of that system. Its network of more than 5 million domain experts writes the tasks and the verifiers whether an agent completed them correctly and its APEX benchmarks score model performance across real-world workflows. Deeptune supplies the software those tasks run inside.

The purchase lands in the middle of a fundraising push. Bloomberg reports that Mercor is in early talks to raise at a valuation of about $20 billion, double the $10 billion it commanded in October when it closed a $350 million Series C. Foody said on X that annualized revenue crossed $2 billion in June, up 100% in four months.

Mercor’s year has not been uniformly good. Attackers using malware planted in the open-source LiteLLM library reached the company’s systems in March. Lapsus$ later claimed to hold four terabytes of Mercor data, including source code, a user database and a bucket of video and identity-verification files. The stolen data reportedly included contractor Social Security numbers, passport scans, interview recordings and facial biometrics. Contract workers have since filed suit.

Foody told Fortune the incident has not cost Mercor business. “Every frontier lab has expanded their relationship with us since the data breach,” he said.

Lupo and the rest of Deeptune’s staff are joining Mercor in New York.

Image: Mercor

A message from John Furrier, co-founder of SiliconANGLE:

Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities.

15M+ viewers of theCUBE videos, powering conversations across AI, cloud, cybersecurity and more

11.4k+ theCUBE alumni — Connect with more than 11,400 tech and business leaders shaping the future through a unique trusted-based network.

About SiliconANGLE Media