Pittsburgh-based Skild AI has raised close to $1.4 billion to build a single, universal “brain” for all robots. Valued at over $14 billion, the startup is developing the Skild Brain, an omni-bodied foundation model designed to control any robot—from humanoids to quadrupeds to industrial arms—without prior training on that specific hardware, aiming to solve a core bottleneck in general-purpose robotics.
The dream of a truly adaptable, multi-purpose robot has long been hampered by a fundamental limitation: artificial intelligence is typically custom-built for a single machine’s body and task. Skild AI, founded in 2023 by pioneers in adaptive robotics, is betting that the path to ubiquity requires decoupling intelligence from hardware. Their proposition is a unified model that can command a warehouse manipulator to sort packages one moment and guide a humanoid to cook an egg the next, simply by adapting in real-time. “We believe that a unified, omni-bodied brain is the fastest way to establish a continuous data flywheel,” stated Deepak Pathak, co-founder and CEO of Skild AI.
Building this kind of generalized physical intelligence presents a monumental data challenge. Unlike large language models that feast on the entire internet, there is no equivalent “Internet of robotics.” Skild AI tackles this by pre-training its model on diverse, alternative data streams: learning from vast libraries of human activity videos and practicing in endless physics-based simulations. This approach allows the Skild Brain to develop a fundamental understanding of physical concepts—like grip, force, and balance—that transfer across wildly different robot forms. The company claims this enables adaptation to severe, unforeseen scenarios like a lost limb, a jammed wheel, or an entirely new body without any retraining.
READ ALSO: https://modernmechanics24.com/post/worlds-first-20mw-offshore-wind-turbine/
Central to this capability is what the company calls “in-context learning.” When a robot powered by the Skild Brain enters a new environment and its initial actions fail, the model doesn’t shut down; it uses that live failure as immediate feedback to adjust its behavior on the fly. “This enables robots to operate dynamically in complex environments, without requiring preprogrammed instructions for each scenario,” explained Abhinav Gupta, co-founder and president of Skild AI. This research-backed method is key to moving beyond robots that operate only in carefully controlled cages to machines that can navigate the unpredictable chaos of real-world homes and job sites.
The staggering funding round, led by SoftBank Group with participation from heavyweights like NVentures (NVIDIA’s venture arm) and Jeff Bezos through Bezos Expeditions, signals immense confidence in this approach. The capital injection follows explosive reported growth, with Skild AI claiming it went from zero to $30 million in revenue within just a few months in 2025. Early applications are targeting enterprise domains where labor shortages are acute: security, warehouse logistics, manufacturing, and data center maintenance. Strategic investors like Samsung, LG, and Schneider point to future integration across consumer electronics and industrial equipment.
WATCH ALSO: https://modernmechanics24.com/post/china-humanoid-robot-combat-moves/
“Skild AI is building foundational technology for physical AI across robots, tasks, and environments,” said Dennis Chang, managing partner at SoftBank Investment Advisers, highlighting the broad vision. The ultimate goal, as stated by the company, is deployment in consumer homes, with enterprise tasks serving as the necessary proving ground. By creating a shared intelligence that improves with every single robot deployment worldwide—regardless of the maker or model—Skild AI isn’t just building another robot controller. It’s attempting to build the foundational operating system for the physical world, positioning itself as the critical layer upon which a new generation of automated industry will run.
