Technology

A Robot That Can Feel What It Picks Up Just Raised $150 Million

Martin HollowayPublished 3w ago2 min readBased on 5 sources
Reading level
A Robot That Can Feel What It Picks Up Just Raised $150 Million
Photo by Mikhail Nilov on Pexels

Eka Robotics has raised $150 million at a $1.45 billion valuation. TechCrunch

The round was led by a16z and F-Prime. Eka was started by Pulkit Agrawal, an MIT professor, and Tuomas Haarnoja, a former Google DeepMind researcher.

Eka's robot pairs a gripping hand that can feel pressure with new AI software. The company calls it a vision-force-action model, built around language instructions, touch signals, and movement. Cameras and language make the plan, and touch corrects it on contact. Think of reaching into a laundry basket: your eyes find the shirt, but your fingers adjust the grip.

Eka says the robot can pick up about 1,500 objects weighing up to 1 kg per hour, compared with about 600 for a human. It succeeds more than 95% of the time. If those rates hold outside a careful demo, they would beat manual rates for sorting small packages.

Outside reaction has focused on handling new objects and recovering from mistakes. University of Illinois Urbana-Champaign professor Kris Hauser said the demo had a "wow factor" compared with other recent robot demos. He said the robot adapted to many different objects without retraining. He described its habit of trying again after a missed grasp until it secured the object as humanness.

Eka puts touch first. Most vision-language-action systems still rely on cameras or a simple yes-or-no signal for whether a grip worked. Eka treats force as a main input alongside sight and language. That helps with common bin problems: slip, wedging, deformation, and uneven weight.

The broader context here is strong investor interest in physical AI, or AI for the physical world. SiMa.ai reached a $1.45 billion valuation after a $150 million round to scale physical AI, bringing total funding to $500 million for its purpose-built Physical AI platform. Intelligent CIO Physical Intelligence previously raised $600 million at a $5.6 billion valuation. Bloomberg Skild AI raised $1.4 billion to develop a brain usable in robots of all shapes and sizes. Bloomberg Video Arm reorganized to create a Physical AI unit to expand its presence in the robotics market. Reuters

For real warehouses, the questions go beyond the demo. Picks per hour need background details: object mix, how often items are hidden, how often human help is needed, and how lighting or packaging changes affect results. At 1,500 picks per hour, more than 95% success still means dozens of misses per hour, so retrying well and signaling problems matter as much as the first try.

In my view, the retrying is the most important detail. Tests usually measure one-try success. Real value comes from recovery on its own. A system that feels a failed or weak grip and tries again without a human reset cuts the most expensive part of warehouse automation, fixing exceptions.

One point worth flagging is language. Language instructions in robotics are often seen as staged for demos. Paired with touch, they have a clearer use: telling the robot to be gentle, what is fragile, and where to hold when cameras alone cannot tell. The test is whether that works on new product types without retraining, which is what Hauser pointed to.

Looking ahead, if Eka's numbers hold up, the near-term change is not replacing people with human-like robots. It is getting more from existing sorting, kitting, and returns lines where speed and handling varied shapes set output. That is less dramatic than the usual robotics story, and more useful.