After Tesla lawsuit, Proception lands $11 million to chase the robot hand prize

Proception, the robotic-hand startup founded by a former Tesla engineer, has emerged from a bruising legal fight with its old employer with fresh funding, an early product launch and a clearer pitch: the hardest part of humanoid robotics may not be the legs or the torso, but the hand.

Jay Li, who previously served as a technical lead on Tesla’s Optimus humanoid robot program, told TechCrunch that being sued by Tesla was painful but ultimately instructive. The experience, he said, forced the company to harden quickly and sharpen its focus. That resilience test is now behind him after Tesla dropped its trade secret case earlier this month following a settlement.

With the litigation resolved, Proception announced a seed financing of $11 million led by First Round Capital, with participation from Y Combinator and BoxGroup. The company also said it has begun shipping the first batch of its high-dexterity robotic hand to researchers and robotics companies, while opening the door to broader orders.

The timing is notable. Robotics has become one of the hottest areas in artificial intelligence and hardware, with investors pouring money into humanoids, warehouse automation and physical AI. Yet even in a field full of ambitious claims, the robotic hand remains an especially stubborn challenge. Proception is betting that the companies able to solve it will hold an important advantage in the next wave of machines built to work in human environments.

Why the robotic hand matters so much

The attention around humanoid robots has often centered on whether a machine can walk, balance and navigate a factory floor. But the true bottleneck for many real-world tasks is manipulation: grasping, pinching, rotating, feeling and adjusting objects with the nuance of a human hand.

That problem is so difficult that even Tesla CEO Elon Musk has repeatedly described robot hands as one of the toughest engineering hurdles in robotics. Industry researchers have echoed that view. Kevin Lynch, who leads Northwestern University’s Center for Robotics and Biosystems, told The Wall Street Journal last year that his team expected it could take roughly a decade before such hands become truly functional and useful for human-like tasks.

Li believes that timeline can be compressed, not by relying solely on hardware, but by pairing hardware with better data generation methods.

The “last mile” of humanoid robotics

In the broad humanoid race, it is easy to overestimate how much progress can be made with a strong chassis, a capable model and a humanoid silhouette. But if the machine cannot handle delicate, repeatable object interactions, it will still fall short in most commercial settings.

That is why investors and robotics researchers often describe dexterous manipulation as the last mile problem of the humanoid story. Without it, robots may look impressive in demos but struggle with tasks that matter in factories, labs, warehouses and service environments.

Proception’s pitch is that the hand is not just a component; it is the product. Rather than treating manipulation as an afterthought, the startup is building its business around the idea that a best-in-class hand could become a core platform for other robotics companies.

From Tesla engineer to startup founder

Li’s path to founding Proception is tied closely to Tesla’s own humanoid ambitions. He worked as a technical lead on the Optimus program before leaving to start the new company. Tesla later accused him of taking trade secrets with him, setting off a legal fight that hung over the startup for months.

The dispute is now closed. Tesla dismissed the lawsuit after the parties reached a settlement, allowing Li to move forward without the cloud of litigation. Tesla did not respond to a request for comment.

For a young startup, a lawsuit from a powerful former employer could have been destabilizing. Instead, Proception appears to have used the period to keep building, tightening its fundraising narrative around technical differentiation and speed of execution.

“I think it’s kind of like a resilience test, or pressure test,” Li said, describing the lawsuit as a challenge that may have strengthened the company. “People say that what doesn’t kill you makes you stronger, right?”

Li also signaled that he expects the market for sophisticated hands to grow as humanoid efforts expand. In his view, the companies that can supply a practical solution will become indispensable to others that do not want to spend years and significant capital developing the capability in-house.

The new funding round and who backed it

Proception’s $11 million seed round gives the startup new room to scale manufacturing, refine data collection and serve early customers. First Round Capital led the investment, while Y Combinator and BoxGroup also participated.

For First Round, the bet appears to rest on both the technical thesis and the founding team’s ability to navigate adversity. Bill Trenchard, a partner at First Round who led the investment, said Proception’s approach to combining hardware and scalable data stood out.

Trenchard said he believes the company is building “the best hand in the market” or possibly the most advanced version available today, backed by the data and model stack needed to improve it over time.

He framed dexterous manipulation as a foundational element of humanoid success, arguing that robot hands are central to whether these systems become truly useful beyond demos and controlled lab environments.

He also said Li and his team handled the legal dispute with discipline, noting that they stayed focused and were transparent with investors once the lawsuit became public.

That combination of technical ambition and operational steadiness seems to have helped Proception make its case in a crowded robotics market where hype can often outrun execution.

How Proception says its system works

At the center of Proception’s strategy is a sensor-rich glove designed to collect high-quality human hand interaction data. The company says the glove can be worn by human testers, alongside a headset, to capture interactions without needing a robot physically performing the task during every data collection session.

That matters because much of the industry currently relies on teleoperation to train humanoid systems. In a typical teleoperation setup, a human wearing a VR headset sees through the robot’s perspective and manipulates objects on the robot’s behalf. The robot then learns from those actions.

Li argues that method has major limitations. The operator does not directly experience tactile feedback from the objects the robot touches, and the amount of data generated is constrained by how many robots are available for training.

Why teleoperation is not enough, according to Proception

Teleoperation has become a useful bridge for training robotic systems, but it is still a bottleneck. It often requires expensive equipment, human supervision and limited throughput. It also may not capture the full richness of human touch.

Proception’s answer is to collect hand interaction data directly from people, then use that information to improve both the design and the learning process behind its robotic hand. In theory, that should make the company less dependent on scarce robot time and more capable of scaling data collection across tasks.

Li says the approach also allows more task-specific learning, which can improve realism and precision. Rather than relying on generic motion capture alone, the startup aims to gather nuanced examples of how the human hand handles different shapes, textures and forces.

Li said Proception is building “highly dexterous hardware plus highly scalable data,” arguing that both elements are required if dexterous manipulation is going to be solved at scale.

Inside the hand: degrees of freedom and sensor-packed skin

Proception says its robotic hand has 22 degrees of freedom and multiple joints per finger, enabling a broad set of dexterous motions. In robotics, more degrees of freedom generally mean more flexibility, more complexity and more control over movement.

The same glove used for data collection also serves another purpose on the actual robotic hand, acting as a sensor-laden skin layer. That allows the company to capture information about how the hand is interacting with objects in a way that can feed back into training and control.

The result, if Proception’s approach works as intended, would be a hand that can do more than grasp large objects or execute scripted motions. It would be designed to handle the subtle, everyday interactions that make human hands so versatile: adjusting pressure, reacting to slip, repositioning a grip and adapting to object shape on the fly.

What 22 degrees of freedom can enable

• More natural finger articulation
• Improved grasping of irregular objects
• Greater adaptability for fine manipulation tasks
• Potentially better performance in human-oriented environments

In practice, that level of dexterity could matter for manufacturing, lab work, logistics and service applications where a robot has to handle a mix of rigid, fragile and oddly shaped items. It could also become relevant for future humanoid platforms seeking to move from controlled demonstrations to real labor.

A crowded robotics market, but a narrow wedge

Robotics is attracting huge interest from investors and big tech, particularly in areas tied to AI, autonomy and labor replacement. But much of that attention has gone to robots that can move around, navigate spaces or perform broad tasks. Fewer startups are trying to own the highly specialized layer of human-like manipulation.

That creates a potential opening for Proception. Rather than compete head-on with every humanoid company, it can position itself as a component supplier and software-data platform focused on a single, difficult problem.

That strategy has precedent in hardware markets. Specialized parts suppliers can become valuable if they solve a key bottleneck better than the companies trying to build full-stack systems. For robotics, a reliable hand could end up being as strategically important as the chassis it attaches to.

Still, the challenge is steep. Robotic hands are notoriously hard to engineer because they require durability, precision, lightweight design, sensing, power efficiency and control all at once. Add in the need for scalable data and learning systems, and the complexity rises quickly.

What Tesla’s role means for Proception

Proception’s backstory gives it a unique place in the robotics conversation. Tesla is one of the best-known companies in humanoid robotics, and Optimus has become one of the industry’s most closely watched programs. Li’s previous role there gives him credibility in a space where practical experience matters.

At the same time, the lawsuit underscored the risks of founding a startup after leaving a high-profile employer. Trade secret disputes can be especially disruptive in robotics, where competitive advantage often depends on technical know-how, process details and engineering judgment as much as on formal patents.

For Proception, the legal episode may also have sharpened its messaging. The company can now frame itself as an independent effort that is no longer defined by its origin story alone, but by the distinct data strategy and hardware stack it is developing.

Li suggested he would not be surprised if Tesla eventually turned to Proception for help as the company grows, saying that after the legal fight he would not rule out future collaboration or interest.

That comment reflects both confidence and a wider truth about the robotics market: even rivals often depend on one another for components, talent and technical breakthroughs.

How the company’s strategy compares with the industry

Many robotics startups focus on building complete humanoids or on software that controls robots in general terms. Proception is carving out a narrower but potentially more defensible niche: the hand as a specialized product category.

That distinction matters because the market for robot hands could evolve differently from the market for whole humanoids. A startup that develops a superior manipulation platform may sell into multiple ecosystems, including industrial automation, research labs and future humanoid fleets.

There are also practical reasons the company may find customers. Researchers often need hardware that can produce repeatable results. Robotics firms may want to avoid spending years building a hand from scratch. If Proception can offer both robust hardware and useful data tooling, it could become attractive even before humanoids scale broadly.

Potential customer segments

1. Academic robotics labs seeking advanced manipulation platforms
2. Humanoid startups building prototype and production systems
3. Industrial robotics teams testing more flexible end-effectors
4. Companies collecting hand interaction data for model training

Timeline of Proception’s latest chapter

Why investors are paying attention now

The robotics investment cycle often moves in waves. First comes the excitement over a new capability, then the realization that real-world deployment is harder than the demos suggest, and finally the hunt for specific bottlenecks that unlock broader adoption. Proception is trying to occupy that third phase.

By focusing on dexterous manipulation, the company is addressing a gap that remains large even as AI improves. The software layer may get smarter, but without a hand capable of translating intelligence into physical action, the robot remains limited.

That is likely what drew backers to the deal: not just the vision of humanoids, but the unglamorous engineering work that makes them useful.

For investors, the proposition is simple enough. If robots are going to leave the lab and work in the physical world, they will need hands that can handle the world’s messiness. Proception wants to be the company that builds them.

The bigger question: how fast can the field move?

One of the biggest unknowns in humanoid robotics is the pace of commercial progress. Musk has suggested factory use could arrive within a few years. More cautious researchers think the timeline is much longer, especially for high-skill manipulation.

That divergence is important because it shapes capital allocation, product strategy and customer expectations. If the timeline is short, companies may race to ship rough systems quickly. If it is long, the opportunity shifts toward infrastructure, components and data pipelines that can compound over time.

Proception is behaving as if the long game is still winnable on a startup timeline. That is an ambitious bet, but one that investors in frontier hardware often reward when the technical thesis is precise enough.

What success would look like

For Proception, success probably will not be judged by a flashy demo alone. The company will need to show that its hand is durable, adaptable and useful to customers who are trying to solve real tasks. It will also need to prove that its data method produces meaningful improvements over existing approaches.

If it can do that, the startup could become more than a robotics point solution. It could become part of the enabling layer for humanoid machines, industrial manipulators and future physical AI systems.

Key facts at a glance

What comes next

With the lawsuit behind it and new funding in hand, Proception now faces the more familiar but equally difficult challenge of proving its technology in the market. The company must convince buyers that its hand is not only advanced, but dependable and practical enough to matter in real deployments.

For now, the startup has at least three signs of momentum: a resolved legal fight, a fresh seed round and a product that is moving into the hands of early users. In robotics, that combination can be enough to turn a promising thesis into a serious company.

Whether Proception ultimately becomes a key supplier in humanoid robotics will depend on execution. But its central argument is already clear: if the future of robots depends on hands that can do human-like work, the companies that solve dexterity first may hold the advantage.

superintelligencenews@gmail.com