Humanoid robots are still dancing and serving coffee at trade shows, but dexterous hands have quietly hit 10,000-unit mass production with over 100 million yuan in orders.

36Kr exclusively reports that bionic robotics company Yuequan Bionic has completed a hundreds-of-millions-yuan Pre-A+ funding round, with investors including Changfa Fund, Huakong Fund, and ChinaAMC. What's more noteworthy than the funding amount is its commercial progress: its flagship product, the "bionic tension-compression dexterous hand," has already reached 10,000-unit mass production, with backlog orders exceeding 100 million yuan. At a stage when humanoid robots overall are still in the "impressive demos but nothing sells" phase, the dexterous hand component has already cracked the commercial closed loop.

Why Hands Are Breaking Through First

There's a strange timing to embodied AI commercialization: everyone's watching the "full body" humanoid robots, but the parts making real money first are often the hands, eyes, and legs.

The dexterous hand is the classic example. A full humanoid has dozens of degrees of freedom. Making it walk without falling and work without collapsing is extremely difficult technically, and the cost is sky-high — a single humanoid costs hundreds of thousands to over a million yuan, and what customers can do with it is still very limited. But dexterous hands are different: they're a component that can be deployed independently. Mount a dexterous hand on a factory robotic arm, and it can do delicate tasks previously only human hands could — tightening screws, inserting ribbon cables, sorting small parts, operating buttons. You don't need a whole humanoid. One hand is enough.

Yuequan Bionic's technical approach is also interesting. It doesn't follow the traditional "motor + reducer" route. Instead, it uses bionic tension-compression body technology — mimicking the human muscle-skeleton system, driving fingers with elastic bodies and tension-compression structures. The benefits? Lighter weight, cheaper cost, more robust (collision-resistant), and natural force feedback capability. A traditional dexterous hand might need three or four motors per finger — costs won't come down and they break easily. The bionic tension-compression structure is much simpler, and mass production costs can be pushed way down.

What does 10,000-unit production mean? Globally, the number of companies that can achieve 10,000-unit dexterous hand production can be counted on one hand. If Yuequan's numbers hold true, it might already be one of the highest-volume dexterous hand manufacturers in China and globally. More critically, 100 million yuan in orders — this means it's not just that production capacity went up with no buyers; there are real customers willing to pay.

The "Components First" Path for Embodied AI

Yuequan Bionic's story reflects an increasingly clear path in the embodied AI industry: sell components first, build whole robots later.

The old thinking was "build the full humanoid first, then gradually reduce costs and mass produce." But the reality is, the technical complexity of humanoid robots is too high — motion control, balance, perception, manipulation — each环节 still has tons of unsolved problems. Trying to deliver a complete product to customers in one go results in clients getting something that doesn't work well, with zero repurchase.

Now more and more companies are switching approaches: first get the hand right and sell it, polish the hand's technology and supply chain to maturity; then build the arm, then the torso, and eventually assemble a full humanoid. It's like how car manufacturing doesn't build whole cars in one step — you make the engine, chassis, and battery first, then integrate them.

This isn't just Yuequan Bionic's choice. At WAIC 2026, Insight Intelligence released its "embodied semantic agent system," and RoboSense launched its second-generation all-solid-state perception platform — both working on specific "organs" of embodied AI, not the whole person. The National Embodied AI Application Pilot Base also released the collaborative world model MoWorld 3D — solving the "brain" problem.

This "components first" approach has several advantages: first, commercial closure comes fast — components can be sold directly to industrial customers without waiting for the whole robot to mature; second, technology iterates faster — components used in real scenarios collect data that feeds back into technical improvements; third, supply chains mature earlier — the process of mass-producing components is the process of polishing the supply chain, so when you're ready to build whole robots, the supply chain is already ready.

Of course, the component route also has its ceiling. Making only dexterous hands limits market size; the whole-robot market is where the real imagination is. But for startups, surviving first matters more than anything — being able to support yourself by selling hands beats burning through funding while waiting for a whole-robot breakthrough.

Yuequan Bionic's hundreds of millions in funding and 10,000-unit production is a small milestone for embodied AI commercialization. It's not the flashiest one, but it might be the most solid one. After all, for robots to enter the real world, the first thing they need — is a pair of hands that can actually work.

See you tomorrow.