Eka's numbers land, 1,500 picks an hour per robot
A robotics roundup attaches performance figures and a valuation to Eka's raise, and nobody outside the company has checked them.
Eka Robotics' $150M Series A now has numbers attached to it. In a weekly robotics roundup, Lukas Ziegler says the round closed at a $1.45B valuation led by a16z and F-Prime, and lists performance figures for the company's picking system. Roughly 1,500 objects picked per hour, against roughly 600 for a human, with 95%+ reliable pickup.
force and touch close the last-millimeter problem vision alone cannot solve
~1,500 objects/hour picked vs ~600 for a human, with 95%+ reliable pickup
The hardware number is the other eye-catcher. Eka's custom gripper is listed at under $1,000 and rated for 5 million duty cycles. For a company selling picking throughput, a cheap end effector that survives millions of grasps matters as much as the model driving it.
The funding itself was already known. What is new here is the valuation, the pick rate and the gripper spec.
Force instead of language
Eka was founded by MIT professor Pulkit Agrawal and Google DeepMind's Tuomas Haarnoja. Their pitch is a vision-force-action model rather than the vision-language-action models (systems that turn camera images and text instructions into robot actions) that most of the field is building. The argument is that cameras run out of information at the moment of contact, and touch is what finishes the job.
Ziegler sums the thesis up as force and touch closing "the last-millimeter problem vision alone cannot solve."
Shown versus claimed
None of these figures came from Eka directly in this source. They appear in a third-party roundup post, with no video, no test conditions, and no description of what the 1,500 objects actually are. Pick rates swing wildly depending on object size, bin clutter and how close together the bins sit. Nobody outside the company has run the system and reported back.
The force angle is not unique to Eka either. The same roundup notes Robotiq launching the 2FC, an adaptive gripper with force sensing and onboard compute at the wrist for slip detection and grasp checks, rated for 2 million cycles and shipping Q1 2027. Zurich-based Enact Intelligence came out of stealth with a $30M pre-seed to build robots that sense force through their motors instead of separate sensors, founded by @manumerous, formerly software lead on 1X NEO and at Physical Intelligence.
Touch is having a moment. Eka just put a price and a throughput number on it.
Sound like you were at the demo
One email a day. Under three minutes. The demos, the deployments and the part the press release left out.