Eka is worth $1.45B thanks to a physical AI model that has experts raving

Robots have been lousy at the simplest-looking part of the job: grabbing things without mangling them. Eka Robotics thinks it has a better answer than the usual flood of video data and visual training, and investors just paid a steep price to back that claim. The startup closed a $150 million Series A at a $1.45 billion post-money valuation, with a16z and F-Prime leading the round.

The company was founded by MIT professor Pulkit Agrawal and Tuomas Haarnoja, who previously worked at Google DeepMind. And the pitch is not just another robot demo with shiny motions and a nice camera angle. People who have seen it say the system feels different. Kris Hauser, a robotics professor at the University of Illinois Urbana-Champaign, said it had a “wow factor” compared with recent demos. Another roboticist told TechCrunch the experience felt emotionally striking, like the period when legged robots first became reliably mobile.

Eka’s system is built around a force-sensitive gripper and a deep-learning model that tracks vision, force and action together. That matters because, in Agrawal’s view, vision only gets you close. The last bit of the task — the actual search, squeeze and adjustment — comes from force. The company says the robot can pick up about 1,500 objects an hour, up to 1 kg each, while a human manages about 600. Eka also says it gets the object more than 95% of the time.

The hardware is part of the story too. Eka designed a gripper with more flexibility and more force sensors, so it can react when it touches something instead of crushing it. The training process leans on physics simulation and reinforcement learning, which the founders say lets them generate thousands of hours of robot experience in minutes. That is the bet: not more human demonstrations, but a model that can learn in simulation and then transfer that skill to real objects.

There’s still a gap between a lab win and a machine you can trust on a factory line for a year. Eka says it is already running secretive pilots with industrial customers on pick-and-place and wire manipulation, and Agrawal says the current gripper costs less than $1,000 and is rated for 5 million duty cycles. The long-term target is the 20 million cycle bar set by standard industrial arms. That is a tall order, but at this valuation, the market is already buying the argument that force beats pretty demos.