Eka Robotics raises $150m to scale its physical-AI robot model

Eka Robotics has announced a US$150 million Series A to develop robots that can handle a wider range of physical tasks. The company says its vision-force-action model has reached 1,500 operations an hour, a striking claim that now needs to hold up beyond the pitch deck.

Watch Desk analysis

What happened

AI Magazine reports that Andreessen Horowitz and F-Prime co-led the round, citing announcements by Eka co-founders Pulkit Agrawal and Tuomas Haarnoja. Eka says its model combines vision, force and action, using reinforcement learning to help robots learn how to interact with physical objects.

The company says its robots can pick up objects outside their training distribution and perform tasks including screwing in a light bulb and inserting wires. It puts their speed at 1,500 operations an hour, compared with a human baseline of about 600. Those are Eka’s performance claims, not independently established results.

Why it matters

Robotics has long had a knack for impressive demonstrations that struggle when the object, setting or task changes. Eka’s emphasis on handling unfamiliar objects gets at a real challenge: whether a robot can generalise beyond the carefully prepared demo.

The funding is substantial, but the useful test is what the robots can do reliably in customer settings. Eka says its grippers cost less than US$1,000 and are rated for five million duty cycles; the company’s stated goal is to reach the 20 million cycles it says are standard for industrial arms.

Our read

This is a notable bet on physical AI, with a more concrete proposition than “robots, but smarter”: learn to manipulate objects across tasks, then make the hardware durable enough for deployment. The reported throughput figures are exciting if they survive real-world testing. For now, treat them as the company’s claims, not a settled benchmark.

What to watch

• Whether Eka publishes independent performance results across different objects and environments.
• How quickly the company moves from demonstrations to customer deployments.
• Whether gripper durability approaches the company’s stated industrial target.

Discussion spark: Would you trust a robot handling unfamiliar objects based on a headline throughput figure, or should proven reliability in customer settings come first?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.