- Robot demos impress, but unit economics remain a major challenge
- Beyond robot hype, humanoid companies face a tough profit test
When Unitree launched its A-share IPO subscription on August 10, it was set to become the first humanoid robot company on China’s stock market, with a valuation of about 61 billion yuan ($9 billion) and a price-to-earnings ratio approaching 220 times.
The listing triggered a new round of valuation debate around embodied intelligence. More importantly, it established a benchmark the capital market had long been waiting for: humanoid robots are no longer just a technology story — they can become a commercial business.
But as Unitree moves toward capital-market validation, its closest rival AgiBot (智元机器人) is pursuing a very different path.
On July 24, AgiBot confirmed that it had begun preparations for a Hong Kong IPO, with cornerstone investors reportedly targeting a valuation of about 36.3 billion yuan to 43.3 billion yuan, broadly in line with Unitree.
Yet the significance of AgiBot’s listing goes beyond fundraising. If Unitree represents the IPO of a successful product company, AgiBot is attempting to answer a much bigger question: Will embodied intelligence ultimately be dominated by a single winner, or by an ecosystem?
Betting big on a “1+5+N” ecosystem
AgiBot has built one of the most distinctive ecosystem strategies in China’s embodied intelligence industry.
The structure is built around a “1+5+N” model.
The “1” refers to the listed parent company, which focuses on the core capabilities of embodied AI models and humanoid robot platforms.
The “5” represents five controlled subsidiaries, each targeting a specific segment: dexterous hand maker Agilink (临界点), quadrupedal robot producer Agiquad (智元拓酷), data service provider Maniformer AI (觅蜂科技), robot rental platform Sharebot (擎天租), and cleaning robot startup Brawn Robotics (智鼎机器人).
Each subsidiary targets a different vertical and has completed its own initial financing round.
Among them, Agilink reached a valuation above $1 billion less than five months after inception, becoming another fast-growing unicorn in the embodied intelligence sector.
The “N” refers to more than 20 early-stage companies invested in by AgiBot, covering key segments across the industry chain.
The logic behind this structure is straightforward: the parent company protects the core intelligence layer; subsidiaries specialize in specific markets while isolating risks; investments secure future possibilities.
At a stage when technology paths remain uncertain, AgiBot is using diversification to hedge against both technological and market risks.
This approach differs sharply from Unitree’s strategy of focusing on in-house robot platforms, engineering execution and commercialization.
When the biggest bet is also the weakest point
However, AgiBot’s ecosystem rests on one critical assumption: its embodied intelligence model must become sufficiently powerful.
Founder Deng Taihua (邓泰华) said research and development of the company’s “brain and cerebellum” AI systems account for three-quarters of both its R&D workforce and spending.
That fund allocation reveals the company’s core challenge.
Its subsidiaries may operate independently and raise their own capital, but their long-term valuation logic remains tied to the capabilities of the parent company’s AI model.
In other words, the ecosystem can only go as far as the brain allows it to go.
The financial pressure is also significant. From 2023 to 2025, AgiBot accumulated about 280 million yuan in attributable net losses and expects to achieve overall profitability only by 2028.
The losses represent a stark contrast to industry peers such as Unitree and DEEP Robotics, which posted net profits of 278 million yuan and 28.68 million yuan, respectively, in 2025.
Contrary to what many showpiece videos may lead viewers to believe, the economics of robot hardware remain challenging.
An industry consensus is that a humanoid robot may sell for 200,000 yuan to 500,000 yuan, while the annual cost of a factory worker is roughly 100,000 yuan to 150,000 yuan. The return-on-investment ceiling for hardware sales is therefore visible.
Robot-as-a-Service (RaaS) models may improve utilization, but their impact on cash flow remains limited. Financing raised by subsidiaries ultimately cannot replace true operational profitability.
AgiBot’s Deng has argued that the company is focused on long-term value creation rather than short-term profits.
But as embodied intelligence moves from the era of storytelling into the era of commercial validation, investor patience will eventually depend on whether financial performance begins to improve.
The model’s inherent contradiction?
The deeper challenge lies in AgiBot’s decision to open its internal capabilities to the broader industry.
By commercializing its dexterous hands, offering data services, and opening rental platforms, AgiBot is effectively helping the entire industry catch up.
Take the robot leasing platform Sharebot, for example. It has even integrated products from competitors including Unitree.
Such openness may help establish industry standards during the early stages of development. But it also means AgiBot is giving up some potential competitive barriers.
Unitree is using a business model with gross margins of around 60% to prove that humanoid robots can generate profits.
AgiBot is attempting to prove something far more ambitious: that a company can simultaneously be a manufacturer, a platform and an ecosystem builder — and that this model can be commercially viable.
That is a much harder proposition.
As embodied intelligence enters a new phase focused on profitability rather than fundraising narratives, AgiBot’s IPO will become a public test of whether a platform-based humanoid robotics company can truly work.
The outcome will depend on two things: how far its AI “brain” can advance, and whether investors are willing to pay a higher premium for an ecosystem than for a product.
