Proception, the robotic hand startup founded by former Tesla Optimus technical lead Jay Li, has settled the trade secrets lawsuit Tesla filed against it last year and raised an $11 million seed round led by First Round Capital, with Y Combinator and BoxGroup joining. Tesla dismissed the case earlier this month. On Monday the company also began shipping the first batch of its high-dexterity robotic hand to researchers and other robotics firms.
The hand has 22 degrees of freedom and multiple joints per finger, a specification Proception is pitching as the most capable on the market. The pitch to customers is straightforward: buy the hand from Proception rather than spending years building one in-house. Li wants the company to become the default supplier of dexterous manipulation hardware to humanoid programs that would rather focus on the rest of the robot.
Li was accused last year by Tesla of taking trade secrets from the Optimus program to launch Proception. The two sides traded legal blows for months before the settlement. Tesla did not respond to a request for comment.
Key facts
- 01Proception raised an $11M seed round led by First Round Capital, with Y Combinator and BoxGroup participating.
- 02Tesla dismissed its trade secrets lawsuit against founder Jay Li earlier this month after a settlement.
- 03The company's robotic hand has 22 degrees of freedom and multiple joints per finger.
- 04Proception is shipping its first batch of hands to researchers and robotics companies starting Monday.
- 05Northwestern's Kevin Lynch estimates functional human-equivalent robot hands are still a decade away.
Bill Trenchard, the First Round partner who led the deal, said Li was transparent about the suit throughout the fundraising process and kept the team focused while it played out. The round closed with the case still active, which is unusual — most seed investors avoid backing founders carrying open litigation from a company the size of Tesla.
The technical bet is that dexterous manipulation is the bottleneck holding back humanoid robots, and that solving it requires hardware and data collected together rather than separately. Most humanoid programs train their systems with teleoperators wearing VR headsets, controlling robots remotely so the robot can learn from the human's commands. Li argues that approach has two structural flaws: the human operator gets no tactile feedback from the objects being touched, and the data collection is capped by the number of robots a company has running at any given moment.
Proception's answer is a sensor-laden glove. Human testers wear the glove and a headset to capture hand-interaction data without a robot in the loop, dramatically expanding how much training data the company and its customers can gather. The same glove design serves as the sensor-packed skin on Proception's own hand, so the data captured by humans transfers directly to the hardware.
The scale of the engineering problem is the reason Tesla CEO Elon Musk has repeatedly singled out robot hands as one of the hardest remaining problems in humanoid robotics. Kevin Lynch, director of Northwestern University's Center for Robotics and Biosystems, told the Wall Street Journal last year that his team believes functional, human-equivalent robotic hands are still a decade away. Li thinks Proception can compress that timeline, citing the data-collection architecture as the unlock.
First Round's thesis on the deal is that whichever company supplies the best hand will sit at a chokepoint as humanoid programs scale.
There is a clear caveat. Proception is shipping its first units now to researchers and robotics companies, which means real-world performance data — durability under load, fine-motor accuracy on novel objects, how well the glove-collected data transfers to the hand in production — is still ahead of it. The 22-degree-of-freedom spec is a hardware claim; whether the underlying models deliver dexterous manipulation that customers will pay for at scale is the open question. Competitors with deeper balance sheets, including Tesla itself, are still working the same problem.
The broader read is that the supplier layer of the humanoid stack is starting to form. The robotics boom so far has been dominated by full-stack humanoid companies — Figure, 1X, Unitree, Tesla — each building hands, arms, torsos, and software in parallel. A specialist hand vendor with credible hardware and a scalable data pipeline is a different kind of bet: that not every humanoid program will, or should, build its own hand, and that the economics of the category will eventually look more like automotive components than vertically integrated platforms. Li expects Tesla itself may eventually be a customer. If Proception's hand performs in the field, that prediction will look less like bravado and more like a sober read of where the humanoid market is heading.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
