Eka Robotics Reaches $1.45B Valuation With Force-Sensing Manipulation Model
MIT spinout Eka Robotics closed a $150M Series A, introducing a vision-force-action model that achieves superhuman dexterity in object manipulation tasks.
Last verified:
Eka’s Breakthrough in Physical Manipulation
Eka Robotics, the MIT-backed startup founded by MIT professor Pulkit Agrawal and former Google DeepMind researcher Tuomas Haarnoja, has achieved a $1.45 billion post-money valuation following a $150 million Series A round led by Andreessen Horowitz (a16z) and F-Prime Capital, according to TechCrunch AI. The valuation reflects growing investor confidence in the company’s novel approach to robotic dexterity—a domain where despite substantial capital deployment, manipulation tasks have remained stubbornly difficult to automate at scale.
Force-Sensing Architecture Sets Eka Apart
Rather than adopting the vision-language-action (VLA) paradigm dominating generative AI robotics, Eka deployed a vision-force-action model that prioritizes tactile feedback alongside visual perception. According to TechCrunch AI, this design choice directly addresses a long-standing critique by robotics pioneer Rodney Brooks, who has argued that purely visual-semantic models lack the grounding to handle real-world contact dynamics. By modeling the relationship between visual inputs, force sensors in the gripper, and motor commands, Eka’s system can learn manipulation primitives that generalize across diverse objects without catastrophic failures.
Performance Metrics Exceed Human Operators
The robot arm achieves approximately 1,500 object picks per hour with 95%+ success rate, compared to roughly 600 picks per hour for human workers, according to TechCrunch AI. Objects tested weigh up to 1 kilogram. The speed and force profile—aggressive enough that the robot could injure itself if the control algorithm failed—suggests high confidence in the underlying model’s reliability. Kris Hauser, a roboticist at the University of Illinois Urbana-Champaign, told TechCrunch AI that the robot demonstrated a “wow factor” and notably adapted to different objects without retraining.
Expert Reception and Market Significance
Reactions from robotics researchers underscore the technical achievement. One anonymous researcher at a frontier robotics lab told TechCrunch AI that witnessing Eka’s robot execute a gentle touch sequence felt comparable to the emotional impact of witnessing the first reliable legged locomotion demonstrations five to ten years prior. Raghu Raguram, the a16z general partner co-leading the fundraise, predicted to TechCrunch AI that Eka will achieve superhuman performance across many dexterous manipulation tasks within months.
Why This Matters
Eka’s valuation and technical reception signal a shift in robotics AI from vision-centric to multimodal sensorimotor models. Manufacturing, logistics, and fulfillment operations have long sought reliable autonomous picking solutions; if Eka’s architecture scales and generalizes as claimed, it could unlock automation for warehouse and assembly tasks that have remained economically unfeasible. The force-action design also validates an alternative trajectory away from pure foundation-model scaling, suggesting that domain-specific inductive biases—here, tactile grounding—remain critical for embodied AI. Follow-up questions: whether the 95% success rate holds under real-world contamination (wet surfaces, deformable items), and whether the model’s adaptability carries to tasks outside the picking domain.
Frequently Asked Questions
What makes Eka's model different from other robotic AI systems?
Eka uses a vision-force-action architecture that incorporates tactile force feedback, rather than the vision-language-action models prevalent in generative AI robotics. This design allows the robot to understand and respond to physical forces during manipulation.
How does Eka's performance compare to human workers?
Eka's robot picks approximately 1,500 objects per hour with 95%+ reliability, versus about 600 objects per hour for human operators. The robot also demonstrates adaptive behavior across different object types without retraining.
Who are the investors backing Eka?
Andreessen Horowitz (a16z) and F-Prime Capital led the $150M Series A round, valuing the company at $1.45B post-money.