$1.45 billion. That is the post-money valuation of Eka Robotics after a $150 million Series A led by a16z and F-Prime, TechCrunch reported on 20 October.
Raghu Raghuram, a general partner at a16z, co-led the round and says the model already exceeds human performance on some tasks. TechCrunch does not test that claim.
The company was founded by MIT professor Pulkit Agrawal and Tuomas Haarnoja, formerly of Google DeepMind. Its model, which it calls vision-force-action, weights force sensing alongside vision and language, so vision finds the object and force handles the final grasp.
It is trained mainly by reinforcement learning in physics simulation, not human demonstrations. Eka says the robot picks about 1,500 objects an hour, up to a kilogram each, against about 600 for a person, and succeeds more than 95 percent of the time.
Those figures come from the company. Kris Hauser, a professor at the University of Illinois Urbana-Champaign, said it handled many object types without retraining, and an unnamed frontier-lab roboticist compared the result to early reliable legged locomotion.
Hauser’s caution is durability. A system must run unbroken for a year on a real assembly line, and Eka’s gripper is rated for 5 million duty cycles, against the 20 million that standard industrial arms target. Most of the demonstrated work is pick-and-place.
Eka is running pilots with industrial customers on pick-and-place and wire handling. No results from them are public, and the founders have been secretive about their methods.
