Proception is a Mountain View startup building dexterous robotic hands for humanoid robots. On June 29, 2026, the robotic hand startup said it had raised an $11 million seed round led by First Round Capital, with Y Combinator and BoxGroup also participating. The pitch is pretty clear: humanoid robots have made real progress in movement and perception, but fine hand control is still where a lot of them break down. Jay Li founded Proception in 2024 with Jack Xu after work on Tesla’s Optimus program convinced them that dexterous manipulation is still the hardest unsolved piece in the stack.
That’s why this round matters more than the dollar amount suggests. Proception isn’t trying to build a full humanoid first. It’s starting with the part nearly everybody admits is brutal: the hand. Li also had to fight off a trade-secret lawsuit from Tesla that was settled in June 2026, clearing a big cloud over the company.
What is Proception and how does the Proception funding-backed technology work?
Proception’s product is ProHand 1.0, a research-grade robotic hand meant for robotics companies and labs that need human-like manipulation rather than a simple two-finger gripper. The hand is built around 22 degrees of freedom and multiple joints per finger. It uses tendon-driven actuation and skin-like tactile sensors, so it can do contact-rich tasks that usually defeat basic end effectors.
Here’s the interesting part. Proception isn’t only selling hardware. It also built ProGlove, a wearable data-capture glove that uses the same sensor-skin idea as the hand itself. A person wears the glove and a headset, manipulates real objects directly, and captures touch plus motion data without needing a robot in the loop first. That means customers can collect training data from human hands before they even spin up a robot fleet.
That workflow fixes two headaches at once. It avoids the scaling limits of classic teleoperation, where data collection is bottlenecked by however many robots you own. It also avoids losing subtle human contact signals — grip shifts, pressure changes, tiny recovery moves — that often disappear when an operator controls a robot remotely instead of touching the object directly.
Proception says ProHand still supports standard teleoperation setups out of the box, so it’s not asking customers to throw away existing robotics workflows. But the company’s roadmap shows a bigger ambition: scale manufacturing and launch a data platform in 2026. Then build a full humanoid prototype in 2027 and reach market deployment in 2028 with cloud-based task planning and analytics layered on top. That’s a long roadmap for a young company. It’s also a much bigger swing than “we made a nice robot hand.”
Who founded Proception before the Proception funding round?
The founding story behind Proception funding
Proception was founded in 2024 by Jay Li and Jack Xu, both alumni of Tesla’s Optimus humanoid robot effort. Li had been a technical lead on Optimus, and Tesla later accused him of taking trade secrets to start Proception. Tesla filed that lawsuit in 2025 and dismissed it in June 2026 after a settlement. Li’s take on the whole episode was blunt: it was a “resilience test” and a pressure test for the company.
The startup is now active, part of Y Combinator’s Winter 2025 batch, and shipping its first batch of hands to researchers and robotics companies. That timing matters. Proception didn’t come out of stealth with a vague demo reel. It came out with hardware on the way to early users.
Founder market fit
Li looks like the kind of founder deep-tech investors usually want for a problem this specific. He’s a Stanford alum and worked on humanoid robotics at Tesla. He’s also been tied to earlier hardware roles at Apple, Aurora, and Aeva. That mix matters because dexterous manipulation isn’t just a robotics problem. It’s a brutal hardware reliability problem too.
Xu brings a similar operator profile. He’s a University of Waterloo alum who worked at Tesla on Optimus and vehicle systems. Earlier, he built medical exoskeletons at Trexo Robotics. His bio also points to autonomous racing work at Waterloo, which is a pretty good signal that he’s comfortable in messy robotic systems instead of just simulations.
Traction and fundraising
On the disclosed numbers, Proception is still small. Y Combinator lists the team at 10 people. But small is normal here, especially for a startup trying to build both electromechanical hardware and a data engine at the same time. The product is live enough that the first batch is shipping now, and the company has opened up broader orders for ProHand.
First Round Capital led the seed round, with Y Combinator and BoxGroup also in. Bill Trenchard from First Round said the firm backed Proception because it thinks the startup may have the best hand in market today, plus the data and models to support it. That’s basically the whole thesis in one sentence: not just a hand, but a learning system attached to a hand.
Competition and market positioning
There are already serious players here. Shadow Robot has spent decades selling dexterous hands and teleoperation systems, with 20 degrees of freedom, 24 movements, and 120 sensors in its current setup. Sanctuary AI is building tactile, high-degree-of-freedom hands inside its broader physical-AI platform. Persona AI is taking another route, commercializing NASA robotic hand IP for heavy industrial humanoids.
So where does Proception fit? Right now, it looks more like a picks-and-shovels supplier than a full-stack humanoid vendor. That’s smart. Lots of robotics companies want better hands and better manipulation data, but they don’t want to sink years into building both from scratch. Proception’s edge is that it’s selling the hand while also trying to own the data-collection layer that teaches the hand what to do. If that works, it becomes harder to swap out than a plain hardware vendor.
Why are investors backing this robotic hand startup now?
Because the legal risk got smaller and the technical wedge got clearer.
A settlement with Tesla doesn’t prove the technology works, but it does remove an obvious diligence problem for investors, partners, and early customers. That alone can change a startup’s trajectory. Buyers don’t love betting on a component supplier that might get tied up in court for another year.
Then there’s the product logic. Proception isn’t chasing general-purpose humanoids from day 1. It’s focusing on one subsystem that nearly every humanoid builder still struggles with. Elon Musk has said robot hands are among the hardest engineering problems still unsolved, and outside Tesla the consensus has been similar: useful human-like robotic hands are still years away. Proception is telling investors it can compress that timeline by pairing better hardware with scalable data capture. That’s ambitious. But it’s at least a focused ambition.
The round also gives Proception room to do the unglamorous stuff hardware startups always need: expand the team and scale production. It also needs to keep building the sensing and data infrastructure behind the hand. If the company only had a clever demo, this raise wouldn’t mean much. What makes it interesting is that the cash is tied to a very specific path: ship hands, gather data, improve manipulation, then climb toward a fuller humanoid platform.
How big is the humanoid robotics market for dexterous hands?
Goldman Sachs Research has projected the humanoid robot market could reach $38 billion by 2035, up sharply from its earlier $6 billion forecast. It also raised its shipment estimate to 1.4 million units by 2035 and said manufacturing costs for humanoids had already dropped about 40%, helped by cheaper components and improving supply chains.
That doesn’t mean every humanoid startup wins. Far from it. But it does explain why investors are willing to fund narrower component plays inside the category. Goldman’s own research says many hardware building blocks — cameras, motors, force sensors, batteries — are close to commercial readiness, while manipulation and interaction remain bottlenecks. If that’s true, then a company obsessed with hands and touch data isn’t early to the market. It may be exactly where the bottleneck is moving.
That’s also why Proception’s timing makes sense. Humanoid robotics is no longer just a moonshot story about flashy demos. It’s becoming a supply-chain story and a data story. It’s also a unit-economics story. A startup that can make dexterous manipulation easier for everyone else could matter even if it never becomes the humanoid brand consumers recognize.
What should you watch next from this robotic hand startup?
The next real test isn’t whether Proception can raise money. It just did that. The test is whether researchers and robotics companies actually adopt ProHand as a default building block instead of treating it like another cool lab device. That’s a much higher bar.
If the company can turn its glove-plus-hand loop into a reliable source of training data, it could become one of the more interesting infrastructure bets in humanoid robotics. Jay Li floated an even wilder signal: Tesla or another big humanoid builder showing up as a customer.
Read how Pocket raised $11M led by Accel to expand its AI-powered meeting recorder that captures offline conversations and turns them into searchable transcripts, summaries, action items, and enterprise-ready workflows.
FAQ
- What funding did Proception raise?
Proception raised an $11 million seed round announced on June 29, 2026. First Round Capital led the deal, and Y Combinator plus BoxGroup also participated, giving the company fresh capital to expand production and its manipulation-data stack. - How does Proception’s robotic hand work?
Proception’s system combines ProHand 1.0 with a wearable ProGlove. The hand uses 22 degrees of freedom and tendon-driven actuation. It also uses tactile sensing, while the glove captures human touch and motion data directly so customers can train manipulation models without relying only on robot teleoperation. - Who founded Proception?
Proception was founded in 2024 by Jay Li and Jack Xu, both of whom previously worked on Tesla robotics. Li came out of the Optimus effort and studied at Stanford, while Xu is a University of Waterloo alum who also worked at Trexo before Tesla. - Is dexterous robotic hands a big market?
Yes — but mostly because they sit inside the much larger humanoid robotics buildout. Goldman Sachs Research has estimated the humanoid robot market could reach $38 billion by 2035, and when manipulation is still one of the main unsolved technical problems, companies focused on hands can become critical suppliers rather than side bets.
