A founder who once built machines inside a major internet company's research lab is now racing to prove that a cable-and-pulley body can out-compete the rigid arms most rivals still ship.
His Shenzhen startup has become the center of gravity in China's embodied intelligence (具身智能) push after a fresh injection of capital arrived faster than most teams manage in two years.
The debate among investors is no longer whether the machine can move, but whether it can earn its keep on a real factory floor before the cash runs out.
What StarDust AI actually raised
StarDust AI (星尘智能) announced in mid-October 2026 that it had closed a new financing round worth about 1 billion yuan (US$141 million; HK$1.1 billion). The round was described by the company and local press as a B+ stage raise, the fourth the firm has completed since March 2026.
The lead investor was Toukong Donghai, the investment arm of Shenzhen Investment Holdings (深投控), a municipal state-owned platform. It was joined by Hongtai Fund, the Ninghai Xingtaihe vehicle managed by Guanghua Yangguang, Shanheng Capital, and an existing shareholder that followed on. Prior backers include Yunqi Capital, Matrix Partners China, ByteDance's Jinqiu fund, Ant Group, and Guoke Investment.
The new money lifted the company's post-money valuation to 15 billion yuan (US$2.1 billion; HK$16.5 billion). Three months earlier, after its June 2026 B-series raise of more than 1 billion yuan (US$141 million; HK$11 billion), the value had first crossed 10 billion yuan (US$1.4 billion; HK$11 billion). In other words, the paper value climbed by roughly 5 billion yuan (US$704 million; HK$5.5 billion) in a single quarter.
Why a rope, not a gearbox
The company's bet is unconventional. Instead of rigid joints and harmonic reducers, StarDust builds rope-driven (绳驱) humanoids, using flexible cables to pull joints the way tendons move a hand. The claim is better force control and safer human interaction, at the cost of harder engineering around friction, tension, and real-time compensation.
Founder and chief executive Lai Jie spent more than seventeen years in robotics and was an early architect at Tencent's Robotics X lab. The core team largely came from that group. The product line spans the S-series flagship wheeled dual-arm robot and the newer T1, a lighter humanoid priced from 89,900 yuan (US$12,700; HK$98,900) that the company says targets homes, shops, labs, and light industry.
StarDust says it reached thousand-unit delivery of its S-series by late 2025 and that combined S and T capacity now sits at the ten-thousand-unit level. Those are company-stated figures; independent verification of field reliability is still thin.
Where the money goes
The proceeds are earmarked for model iteration, hiring, commercial deployment, and capacity build-out, plus continued work on what the company calls its "AI model – embodied OS – rope-driven body" full-stack architecture. It is also pushing a "robot mart" retail format that has been deployed in six Chinese cities since late 2025.
Customers and partners named publicly include ThunderSoft, XianGong Intelligence, and BoDeng Intelligence, with thousand-unit order frameworks announced over the past year. The company frames its path as dual-track: commercial services first, then a developer ecosystem and eventually home users.
What the round signals
The round is a useful signal for the broader embodied-machine sector. State-backed municipal funds are no longer content to watch from the sidelines; they are now lead investors in late-stage robotics names, supplying the patient capital that purely commercial funds hesitate to provide. For StarDust, that backing also hints at where early deployments may land — Shenzhen's own factories and public-service sites rather than distant export markets.
The valuation question nobody can dodge
Here the numbers get uncomfortable. Registry-linked data cited by financial media puts StarDust's 2025 revenue at roughly 20.9 million yuan (US$2.9 million; HK$23 million) against a 15 billion yuan paper value — a multiple that one quoted analyst called pricing a decade of future success into today. The company has not published audited financials, and it has not yet filed a prospectus.
That matters because a Hong Kong listing under the 18C chapter for pre-commercial technology firms sets a high bar: with revenue well below the 150 million Hong Kong dollar threshold, the firm would need research spending to exceed half of total operating costs in at least two of the past three years. StarDust re-registered in August 2026 as a joint-stock company, a step widely read as preparation for a Hong Kong float.
Trend reading
According to Li An, chief scientist at BrainNet (脑机网), the rope-driven path will live or die on total lifecycle cost rather than on a headline sticker price, and investors are right to scrutinize maintenance intervals before celebrating a round size.
Honest limitations
This article relies on company announcements and Chinese financial press; StarDust has not filed a prospectus, so audited figures are unavailable. The 2025 revenue figure comes from registry-linked media estimates, not the company, and should be treated as indicative. Deployment and capacity claims are company-stated and lack independent field verification. Valuation math depends on venture-round pricing that a public market may not confirm. Currency figures use fixed rates of 7.1 yuan per US dollar and 1.1 Hong Kong dollars per yuan.
What readers can do now
- Founders should track StarDust's maintenance-cost disclosures, not just its round size, when benchmarking their own rope-driven or rigid-body roadmaps.
- Investors should watch for the eventual 18C filing and its research-to-operating-cost ratio before treating the 15 billion yuan paper value as a floor.
- Operators evaluating humanoids should request field logs and failure rates on thousand-unit deployments rather than relying on showcase videos.