Skild AI: One Brain for Every Robot

Skild AI: The Robotics Foundation Model Built to Run Any Robot

Every robot today needs its own brain. A warehouse arm can't share intelligence with a delivery drone. A factory manipulator can't borrow what a security robot learned yesterday. Each machine gets trained in isolation, for one body, one task, one environment. That's not a hardware problem. It's a software one, and Skild AI is working to solve it.

Founded in 2023 by Carnegie Mellon professors Deepak Pathak and Abhinav Gupta, Skild AI is building a robotics foundation model designed to work across every robot type and every task. The company calls it the Skild Brain. Unlike traditional models built for specific robot designs, the Skild Brain is omni-bodied, meaning it can control quadrupeds, humanoids, tabletop arms, and mobile manipulators without needing prior knowledge of each machine's exact body form.

The analogy that keeps coming up is language models. Before large language models arrived, you needed a different AI for every text task. After, one model handled most of them. Skild is pursuing the same shift for robots in the physical world.

How the Skild Robotics Foundation Model Learns Without Robot Data

The hardest part of building a general robotics AI isn't the algorithm. It's the data. There is no internet of robotics the way there's an internet of text and images, so most approaches hit a ceiling quickly. Skild addresses this by training on two sources at scale. First, it watches human videos on the internet, learning physical tasks the same way a person does, through observation. Second, it trains across physics-based simulations covering 100,000 different robot configurations, building adaptability before the model ever touches real hardware.

The outcome is a model that handles situations it wasn't specifically trained for. Pathak described the principle directly: the model adapts rather than memorizes, which is closer to how biological intelligence actually works. Whether that holds across every environment at full commercial scale is still being tested in the real world.

Where the Skild Robotics Foundation Model Is Running Today

Skild's technology is already deployed across security inspection, last-mile delivery, warehouse operations, manufacturing, data centers, and construction. The company is also working with ABB Robotics and Universal Robots to bring the Skild Brain into industrial settings alongside existing equipment.

Revenue grew from zero to around $30 million in a few months during 2025, which reflects genuine customer adoption rather than pilot-stage interest. In January 2026, Skild raised close to $1.4 billion in Series C funding led by SoftBank Group, with participation from NVIDIA's NVentures, Jeff Bezos through Bezos Expeditions, Lightspeed, Sequoia Capital, Felicis, and Coatue. The company's valuation reached above $14 billion, up from $1.5 billion just eighteen months earlier.

Abhinav Gupta described the company's position in straightforward terms: Skild is not building the robot. It's building the brain that can go into any robot. That distinction matters because a software platform scales differently than a hardware company. The question the market is now answering is whether one brain can actually work well enough across enough bodies to make that model stick.