Skild AI hits a $100M run rate 10 months into selling robot brains

Why it matters

Skild AI's $100M run rate gives Pathak's hardware-agnostic strategy a commercial proof point. The harder test is whether deployments repeat without costly customization.

Deepak Pathak (@pathak2206) has pushed Skild AI to a $100 million recurring-revenue run rate roughly 10 months after the robotics software maker began commercial use, Bloomberg reported on Thursday.

The figure is the strongest commercial marker yet for Pathak's bet that a shared AI model can control many kinds of robots, rather than tying intelligence to one machine or one narrowly programmed job. Pathak told Bloomberg that Skild AI was "moving from an era of demos to an era of deployments."

Run rate annualizes revenue at its current pace. It does not establish how much revenue Skild AI has recognized, how long its contracts run or how much comes from production deployments rather than pilots. Bloomberg did not attach customer names, contract values or robot counts to the figure. Those details will determine how durable the milestone proves to be.

The speed still stands out. Skild AI began generating commercial revenue around November 2025, based on Bloomberg's 10-month timeline. In its January Series C announcement, Skild AI said live revenue had grown from zero to about $30 million during 2025. Skild AI described that earlier number as live revenue and the new figure as a recurring-revenue run rate, making a direct growth calculation unreliable. Together, the two figures show Pathak has moved quickly from research demonstrations into paid work.

Two professors and one robot brain

Pathak founded Skild AI in 2023 with Abhinav Gupta, now Skild AI's president. Both founders came from Carnegie Mellon University's robotics research community, where the problem they chose had frustrated researchers and commercial operators for years: software trained for one robot or task rarely transfers cleanly to another.

Pathak is a Raj Reddy Associate Professor in Carnegie Mellon's School of Computer Science and a member of its Robotics Institute. According to his university biography, he earned a computer science degree from IIT Kanpur and a Ph.D. from the University of California, Berkeley, before working at Meta AI Research and as a visiting postdoctoral researcher at Berkeley. His research has focused on computer vision, machine learning and robots that learn from limited supervision.

Gupta is a professor at Carnegie Mellon's Robotics Institute whose work spans computer vision, self-supervised learning and robot manipulation. He also helped establish Facebook AI Research's robotics group before returning full-time to Carnegie Mellon in 2022.

Their commercial thesis is concise: "any robot, any task, one brain." Skild AI's model, called Skild Brain, is designed to work across quadrupeds, humanoids, tabletop arms and mobile manipulators. Skild AI trains it with a mix of simulation, internet video, teleoperation and data gathered from machines in use.

That hardware-agnostic approach separates Skild AI from robotics developers building a proprietary humanoid and the software that runs it. Pathak and Gupta can sell an intelligence layer into existing fleets and robot manufacturers without carrying the full cost of designing and producing every machine.

Deployment is also the training strategy

Skild AI's sales effort can also produce data for future model training. Each deployment can generate examples from real factories, warehouses and other physical environments, though its usefulness depends on customers permitting operational data to be used and on that data being varied and clean enough to improve the model.

In March, Skild AI announced work with ABB Robotics, Universal Robots and Nvidia, expanding its relationships with robotics and industrial technology companies.

Skild AI also announced the acquisition of Zebra Technologies' robotics arm, adding a robotics business to its software strategy.

Skild AI later published S1, a robotics foundation model focused on in-context learning. The company's public performance claims remain largely based on its own materials, and independent testing across customer sites remains the harder measure.

A nine-figure run rate after a 10-figure round

Skild AI has had ample capital to make the transition. The founders took Skild AI out of stealth in July 2024 with a $300 million Series A at a $1.5 billion valuation. The round included Lightspeed Venture Partners, Coatue, SoftBank and Bezos Expeditions, alongside Sequoia Capital, Felicis Ventures, Menlo Ventures, General Catalyst, CRV, SV Angel, Carnegie Mellon University, the Amazon Industrial Innovation Fund and the Alexa Fund.

In January 2026, SoftBank led a $1.4 billion Series C that valued Skild AI above $14 billion. NVentures, Macquarie Capital and Bezos Expeditions participated alongside strategic investors including LG, Schneider Electric, CommonSpirit and Salesforce Ventures. Pathak told Bloomberg at the time that Skild AI had raised more than $2 billion in total, according to TechCrunch.

The $100 million run rate gives that valuation a revenue reference point. It also raises the next set of operating questions: how much of the revenue repeats after initial deployments, how expensive each installation is to support and whether one model can deliver consistent performance across the range of machines Skild AI wants to control.

Pathak's strategy depends on deployments improving both sides of the business at once. More customers produce revenue and more physical-world training data. The reported run rate suggests that customers are willing to pay before general-purpose robotics is fully solved. Skild AI now has to show that each deployment makes the next one faster, cheaper and more reliable.