Eka's 1,500 Picks an Hour Is a Lab Number, and Its Own Investors Say So

Eka Robotics announced a $150 million Series A on October 20, led by Andreessen Horowitz (a16z) and F-Prime, for a robot manipulation model that reads force as well as camera images. TechCrunch reported that the round values the company at $1.45 billion post-money.

The figure driving the coverage is 1,500 picks an hour. The company's lead investor describes it as a laboratory result.

What Eka says it has built

Eka calls its model a vision-force-action model, in contrast to the vision-language-action models used by most robot-learning groups. In a16z's announcement, authors Raghu Raghuram and Yoko Li say Eka reports up to "1,500 picks per hour" in laboratory induction and bin-picking tasks, "two and a half times faster than humans."

TechCrunch put the human comparison at about 600 picks an hour for objects up to 1 kg, and reported that Eka says it picks reliably more than 95% of the time. Co-founder Pulkit Agrawal told TechCrunch that vision gives rough location and pose while force handles the grasp: "Vision tells you roughly where the object is, what the pose is, but the actual grabbing, the actual searching is happening through forces."

The company was founded in 2024, according to The Robot Report. The a16z post describes Agrawal as leading the Improbable AI Lab at MIT CSAIL and Tuomas Haarnoja as a co-creator of Soft Actor-Critic who later worked at Google DeepMind.

Lab numbers and what is missing

The a16z authors write that "Sustaining that performance through full production shifts is the next test." They also note that contact dynamics are "notoriously difficult to model" in simulation, and they are investors in Eka.

Claim	Source	Setting
Up to 1,500 picks per hour	a16z, TechCrunch	Laboratory induction and bin picking
About 600 picks per hour for a human	TechCrunch	Objects up to 1 kg
More than 95% pick reliability	TechCrunch, citing Eka	Not specified
Gripper under $1,000	TechCrunch	Rated for 5 million duty cycles

TechCrunch reported that the gripper is rated for 5 million duty cycles against a goal of 20 million, the standard for industrial arms. It also reported pilots with several industrial customers in pick-and-place and wire manipulation. None is named in the article, and The Robot Report's account names no customers or deployments either.

The outside view

Kris Hauser, a professor at the University of Illinois Urbana-Champaign, told TechCrunch the demo had a "wow factor" and that the robot adapted to many objects without retraining. He also said the hard test is real assembly lines running for a full year with integrators, and that it is unclear how far the technique generalises beyond pick-and-place.

TechCrunch also cited Rodney Brooks as a critic of generative robot models that ignore force and touch, and wrote that Eka's results suggest he has a point.

The training method is described in general terms only. Both the company and a16z say policies are trained with reinforcement learning in physics simulation, and TechCrunch wrote that the founders are secretive about specifics. No paper, weights or public benchmark accompanied the announcement in the sources read.

What the round does and does not show

The Robot Report lists MFV Partners, Bison Ventures, Glasswing Ventures and Samsung Next among other participants, with seed investors Obvious Ventures and E14 Fund continuing. Raghuram is quoted there calling it "the first approach I've seen that truly solves manipulation with the generality and throughput needed for deployments." He is also the partner who led the investment.

The next checkable event is a named customer running the system through a full production shift. Eka has not announced one. The sources describe use of funds only as scaling the technology and customer deployments. Related coverage on The Terminal: Warp Sciences' robot-assisted chemistry lab and Generative Bionics' humanoid data factory.