Genesis AI says wheeled robots will beat legged ones

While the robotics industry continues pouring billions into machines that walk like humans, one startup is making a deliberate detour. On Tuesday, Genesis AI unveiled Eno, a wheeled robot designed with a foldable tower, dexterous hands, and a foundation model its creators claim delivers human-level manipulation capabilities. This is a pointed rejection of the prevailing assumption that useful robots must look and move like people.

The bet against Genesis is enormous. Figure AI holds a $39 billion private valuation and has begun deploying its Figure 03 humanoid in a Catalyst Brands warehouse, handling logistics for the parent company of JCPenney and Brooks Brothers. Hyundai and Boston Dynamics plan to produce 30,000 Atlas humanoids annually by 2028 for deployment at car factories in Georgia. Norway’s 1X Technologies sold out its first production run of 10,000 home robots within five days of opening preorders. The boom extends to China, where more than 150 companies are chasing a market that delivered roughly 14,000 units in 2025. Yet only 23% of buyers reported satisfaction, a gap between supply-side ambition and demand-side reality that Genesis believes vindicates its approach.

“The hardest problem in robotics is not locomotion,” Zhou Xian, Genesis AI’s co-founder and CEO, told Business Insider. It is manipulation, the ability to handle objects with the dexterity and adaptability of a human hand. The race to solve that problem has attracted vast investment on its own. Chinese robot-hand maker Linkerbot is targeting a $6 billion valuation on the strength of its dexterous grippers alone, shipping more than 1,000 units a month. Zhou, who holds a PhD from Carnegie Mellon, founded Genesis with Théophile Gervet, a former Mistral researcher. The company has offices in Paris and San Francisco and employs roughly 60 people.

Eno’s wheeled base is the direct consequence of that diagnosis. Wheels are cheaper to build, simpler to stabilise, and safer to deploy around humans than bipedal legs, which remain an unsolved engineering challenge at commercial scale. The trade-off is that Eno cannot climb stairs. Genesis argues that for the use cases it is targeting,logistics, manufacturing, hospitals, and hotels,that limitation rarely matters.

Most robotics companies teach their machines through teleoperation, in which a human operator remotely controls the robot while its sensors record every movement. That process reportedly costs up to $6,000 per hour. Genesis developed a sensor glove instead, one that maps directly between a human hand, the glove, and the robot’s hand. The company says the gloves cost roughly $300 a pair, approximately 100 times less than conventional teleoperation setups, and collect up to five times more usable training data per session. That data feeds GENE-26.5, the foundation model Genesis demonstrated in May. In a public demo, a single model running on the same hardware performed cooking tasks, solved a Rubik’s cube, played piano, and assembled wire harnesses.

Genesis is not the only company questioning the humanoid orthodoxy. Sunday Robotics raised a $165 million Series B at a $1.15 billion valuation for Memo, a wheeled domestic robot trained on data from sensor gloves worn in more than 500 real homes. German startup Sereact has taken a different tack, building a “robot brain” that plugs into existing industrial hardware. Its Cortex model powers more than 200 deployed systems for BMW, Daimler Truck, and PepsiCo, sidestepping the form-factor debate entirely. Meanwhile, Agility Robotics’ bipedal Digit has quietly become the only humanoid generating revenue from paying commercial customers, having moved more than 100,000 totes in a GXO Logistics warehouse. The gap between that single milestone and the industry’s multi-billion-dollar valuations tells its own story. The logistics sector Genesis is targeting is not short of incumbents either. Amazon has invested more than €10 billion in its European fulfilment network and recently deployed Proteus, a warehouse robot that takes plain-language instructions.

Genesis raised $105 million in seed funding co-led by Eclipse Ventures and Khosla Ventures, with participation from Eric Schmidt, Xavier Niel, MIT robotics professor Daniela Rus, and former Intel Labs head Vladlen Koltun. “Even in the most automated industries, the robot-to-human ratio rarely exceeds 1:30,” Eclipse partner Charly Mwangi said. The company plans to begin production and customer deployments by the end of 2026, starting with logistics and manufacturing before expanding to hospitality and healthcare. Whether the humanoid industry’s IPO-bound giants or Genesis’s wheeled contrarianism proves correct depends on a question the market has not yet answered: what shape does a useful robot actually need to be? For now, Genesis is betting the answer has wheels.