Eka Robotics Hits $1.45B Valuation with Physical AI Model Demonstrating Superhuman Manipulation

Technical breakthrough in physical AI

Eka Robotics, founded by MIT’s Pulkit Agrawal and former DeepMind researcher Tuomas Haarnoja, has just secured a $150 million Series A round, bringing its post‑money valuation to $1.45 billion.

The company’s flagship system combines a force‑sensitive gripper with a novel deep‑learning architecture that treats language, force inputs and motion as a unified schema, rather than relying on vision‑language‑action models.

In benchmark tests the robot can handle up to 1,500 items weighing a kilogram each within an hour, achieving a 95 % success rate and demonstrating near‑human speed while avoiding damage to delicate objects such as a raspberry.

Industry experts have praised the system for its “wow factor” and its ability to adapt to new objects without retraining, a capability that could reshape dexterous manipulation in manufacturing, logistics and service robotics.

For GPU and AI infrastructure providers, the breakthrough underscores the growing demand for high‑throughput compute that can execute real‑time force‑feedback models, signalling a new market opportunity for high‑performance AI accelerators.