Eka Robotics Raises $150M for Force-Based Robot Learning at $1.45B Valuation

Eka Robotics, a U.S. startup focused on dexterous robotic manipulation, has raised $150 million in a Series A funding round led by Andreessen Horowitz and F-Prime Capital, with participation from Samsung Next. The round values the company at approximately $1.45 billion, reflecting investor confidence in a new approach to enabling robots to handle delicate and unstructured physical tasks. The company was founded by Pulkit Agrawal, a professor at MIT’s Computer Science and Artificial Intelligence Laboratory, and Tuomas Haarnoja, a former researcher at Google DeepMind. Their work centers on combining computer vision, force feedback and reinforcement learning so robots can perform fine motor tasks such as grasping fragile objects, inserting connectors and managing irregularly shaped items. Eka is betting on a Vision-Force-Action model, or VFA, rather than the widely used Vision-Language-Action approach. In this framework, robots use cameras to identify objects and then rely on force sensors to adjust grip strength, orientation and motion in real time. This is intended to improve performance in tasks where contact forces are critical, such as picking up a raspberry without crushing it or inserting a plug into a socket. The startup’s training method emphasizes large-scale simulation and reinforcement learning. Robots practice millions of attempts in virtual environments under varied physical conditions, then transfer learned skills to real hardware. Eka says this reduces the need for expensive human demonstrations and allows faster development of robust manipulation policies. In recent demonstrations and lab tests, Eka’s robots showed the ability to grasp a wide range of items, including keys, combs, earphone cases and food items. The company reported that its system can operate at up to 1,500 grasps per hour, roughly 2.5 times the benchmark human rate for the tested tasks, with a success rate above 95% for objects weighing up to 1 kilogram. Eka also highlighted practical hardware considerations. Its gripper costs less than $1,000 and is rated for about 5 million operation cycles, although industrial robotic arms typically aim for 20 million cycles. The company is working with select industrial customers on pilot applications involving picking, placing and cable handling. The funding reflects growing investor interest in embodied AI and physical robotics. The key challenge ahead is proving that simulation-trained models can maintain speed, reliability and safety in complex production environments over long periods.