Interview with Zhou Shunbo: why Ola Dimensions skips bipedal robots and targets the laundry room first
From a sinking feeling to a new start: a race more than a decade long
On August 12, Daily Economic News published an interview with Ola Dimensions founder Zhou Shunbo in its Future Business series. It took place in an old building at 505 Dongchangzhi Road in Shanghai, at the first closed-door offline demo day of the DGP gravity programme run by D-Robotics - Gravity DemoDay 2026 - with investors from Linear Capital, Didi Investment and Jinqiu Fund among the reviewers. Among the founders present, Zhou's background stood out: the first employee of Huawei's embodied intelligence lab and a Huawei Genius Youth, who chose the home consumer embodied AI track that the industry considers hardest and most patience-testing.
His connection to robotics goes back more than a decade. While earning a PhD in robotics at the Chinese University of Hong Kong, the full-stack training in physical modelling and hardware integration taught him early how high the barrier is for general-purpose robots reaching the mass market. The more deeply he studied, he said, the more his heart sank: robots then depended on professional engineers for deep development, and that wall of technical difficulty all but shut out consumer adoption - which planted his later goal of lowering the barrier and bringing robotics to ordinary people.
After finishing his PhD in 2020 he did not take a job; instead he spent a full year preparing a robotics startup, finishing an entire prototype system and believing only the final step of commercial deployment remained. The attempt stalled for reasons outside his control. The failure taught him something more valuable: on the robotics track, technical competence is merely the entry ticket, and softer skills such as engineering delivery and resource coordination decide how far a company can go.
After reviewing what went wrong, he made a pragmatic choice and joined Huawei in 2021 as a Genius Youth. Over nearly five years he served as the first employee of the company's embodied intelligence lab, building a research team from scratch that grew from a dozen people to more than a hundred at its peak, leading the launch of Huawei Cloud's embodied platform CloudRobo, and completing the from-zero-to-one technical validation and customer delivery of industrial robots. He describes that period as the most valuable business school he ever attended.
In March 2026 he formally left Huawei and registered Ola Dimensions in Qianhai, Shenzhen. Asked about giving up a large salary and the halo of a major platform, he says he felt no struggle beforehand: more than a decade of technical accumulation, a complete commercial apprenticeship at a big company and the window opened by large models reshaping robotics had all come together, and everything was ready. Co-founder Zhang Jing, according to public reports, was previously a product manager at Amazon AWS and a product director for Huawei's innovation cloud services.
Why no bipeds, and why laundry first
Most of the industry's attention sits elsewhere. China's embodied AI sector has split clearly: the vast majority of startups cluster in B2B markets such as factories, logistics and commercial services, where fixed-scope delivery brings revenue quickly and proves self-sufficiency, while only a handful of teams dare to commit fully to the consumer home. One benchmark in home services, the startup Zibianliang Robotics, builds an end-to-end unified world model with environment, user and task as adjustable variables and runs B2B industry and consumer housekeeping in parallel; according to media reports it has raised more than 5 billion yuan at a valuation of roughly 20 billion yuan.
Ola Dimensions' choice of the home comes directly from experience Zhou accumulated on industrial projects. Running industrial robotics work at Huawei, he saw with his own eyes how a promising B2B proof of concept is separated from volume production by a gap that is hard to cross: factories demand extreme reliability, and no matter how smoothly a prototype runs in a test environment, the moment it meets a real production line, non-standard customisation demands devour the margin. His conclusion: the custom nature of industrial deployments makes their capabilities and data difficult to transfer naturally into the home.
Put plainly, the first problems a home robot must solve are not about doing everything. Through multiple rounds of in-depth user interviews and research, the team settled on two high-frequency pain points: the repetitive labour of moving laundry from the washing machine to the dryer after work, and tidying flat surfaces such as entryways, coffee tables, desks and children's scattered toys. Difficult jobs such as folding clothes and sweeping an entire home are set aside for now.
Form factor matters as much as the wedges. Zhou Shunbo explicitly rules out a bipedal humanoid for now: the first-generation product uses a half-height wheeled base plus a robotic arm with the body kept under 1.4 metres, prioritising physical manipulation, object grasping and surface tidying and concentrating limited R&D resources on closing the technical loop. His explanation is blunt: bipedal humanoids entering ordinary homes still face safety, energy, cost and space-fit challenges. Keeping the body small also avoids the resistance many enthusiast households feel toward large equipment.
How does the product learn chores? The company's answer is a gradual, raise-it-yourself approach. The robot first gains general competence such as moving, grasping and recognising common objects; personalised tasks are then learned from user demonstrations - tidying a certain category of items, operating a particular appliance, adapting to a fixed home layout. The operating data from real homes flows back to the model for the next round of iteration.
A data flywheel, and the prototypes due in December
Commercialisation is staged to match. Zhou says the company plans to penetrate from a niche circle outward to the mass market: several dozen prototype units will be built in small-batch trial production in December 2026 and placed with screened veteran geeks and technology enthusiasts for an in-home beta, a small-scale real-user test. Product versions will keep lowering the barrier, with the aim of pushing operation and secondary-development privileges down from professional engineers to ordinary enthusiasts within the year.
The community is meant to be the moat itself. Ola Dimensions intends to reward co-creating users with compute credits, API access, product discounts and even project dividends or equity incentives, in exchange for large volumes of real household feedback data that feed an algorithm-iteration flywheel - a first-mover advantage that capital cannot easily replicate.
Why this logic matters deserves a line of background. Training data for language models can be harvested from public text almost for free; robot manipulation data cannot. It has to be generated by physical machines in real environments - slow to collect, expensive, and highly non-standard in the home, where layouts, furniture, objects and family habits all differ, so motions calibrated in a lab rarely move straight into the next house. Who gets into enough real homes first therefore stops being a marketing question and becomes a data-supply question. It is also why privacy and data compliance become an unavoidable threshold for any home robotics company.
No chasing hype: close the loop first
Zhou is notably cool-headed about the competitive field. In his view, digital-world intelligence will inevitably extend into the physical world over the long run, every technology giant will bet on general-purpose robotics, and the market is large enough for many players to develop together - not a zero-sum game. In consumer home robotics specifically, he argues startups have shorter decision chains and more concentrated resources than the giants and can break through on extreme product strength. On the rush to IPO in the first half of the year he stays cautious: after thirteen years in the field he is well aware that large-scale commercialisation of general embodied intelligence still carries multiple uncertainties, and that going public too early leaves a company hostage to short-term earnings, away from deep technical work. In his words, Ola Dimensions has exactly one core goal at this stage - close a complete commercial loop in the home, build a solid data iteration system and strengthen product delivery. Capitalisation, he says, is merely a natural consequence, and he will not let hype disrupt the company's established R&D and market rhythm.