The “picks and shovels” model
On 17 September D-Robotics (Digua Robotics) closed a USD 400 million Series C. Mirae Asset led, with Meituan strategic, Hefei Guotou, Nanshan Industrial, Jingquan Capital and Cathay Capital following, while Hillhouse, 5Y Capital, Temasek’s Vertex Growth and Aramco’s Prosperity7 were among more than twenty existing backers adding funds. A single USD 400 million round is one of the largest in the robotics sector in four years. The stranger detail: the company that took the money builds no robot bodies.
D-Robotics positions itself as the “shovel seller”: it makes only chips, operating systems and development toolchains, and lets others build the robots. In its founders’ words the goal is to become the “Wintel of the robotics era”, even the “Nvidia of embodied intelligence”. While humanoid makers cool and valuations come under pressure, a body-free shovel seller took USD 400 million. The question is why robotics money is flowing upstream.
The stack has three layers. The chip layer: the Sunrise series spans 5 to 560 TOPS, from floor cleaners to humanoids. The software layer: the RDK developer kit supports many algorithm deployments and open-source repositories so developers need not build from scratch. The ecosystem layer: the “Gravity” acceleration programme has served over 500 small teams, 100,000 developers and 500-plus universities.
The flagship Sunrise S600, released in November 2025, carries 560 TOPS and the key selling point is “brain and body in one”: an 18-core CPU for perception and decisions plus a 6-core MCU for real-time motor control, all on one chip. Older designs used one AI chip plus an external MCU, adding cost and latency across the chip boundary. S600 does it on one chip, saving a chip’s bill of materials and cutting cross-chip delay.
Commercially the bet is paying: Sunrise chips have shipped over 8 million units, with first-half 2026 revenue up several times year on year. Within half a year of S600’s launch, over twenty top customers adopted it, including Itchi, Qianxun, Zizhi, UBTECH and PaXini, with embodied-intelligence client coverage above 50 per cent and most projects at production scale. But note: most of those 8 million units come from mature categories like cleaners, mowers and drones, not embodied AI. Mature categories are the base. Embodied AI is the growth option.
Why the funding is speeding up
D-Robotics’ cadence is an outlier: Series A USD 100 million in May 2025, B1 USD 120 million in March 2026, B2 USD 150 million in April, C USD 400 million in September. Four rounds in two years, about USD 770 million total, USD 670 million across three rounds in one year, with only twenty days between B1 and B2. As the whole-body humanoid track cooled, a body-free company raised ever larger rounds, a sign that capital’s logic is shifting from “who can build a robot” to “who gives robot builders something they cannot avoid”.
The investor list confirms it. Mirae Asset is Korea’s largest independent financial group and in 2026 backed MiniMax and AgiBot, building an AI and semiconductor portfolio. Meituan invested two rounds to secure bottom-layer chip supply across delivery, quadruped, humanoid and industrial robots, a strategic position. Three state platforms, Hefei Guotou, Nanshan and Beijing, each have industrial motives: Hefei builds a humanoid cluster, Nanshan is D-Robotics’ headquarters, and Beijing counts D-Robotics as the chip supplier to its humanoid innovation centre. When industrial capital, financial investors and local state all crowd in, some surely fear missing out, but the signal is clear: USD 400 million into a company that builds nothing means the market is now pricing infrastructure.
D-Robotics spun out of Horizon Robotics in early 2024. Its core team came from Horizon’s AIoT unit. CEO Wang Cong led edge AI there from 2018. Horizon, born from automotive chips, already proved the “chips only, standardised shipping, big business” path. D-Robotics aims to move the same method into robotics. Horizon’s 2025 report showed its stake in D-Robotics diluted from near 70 per cent to about 40 per cent as D-Robotics kept raising, and from 31 March 2026 D-Robotics left the consolidated statements and became an associate. Horizon is no longer the controlling shareholder but remains the largest single shareholder and technical partner, and Yu Kai stepped back from board appointment. D-Robotics is now weaned but still holds two lines: largest shareholder and technical cooperation.
Fighting Nvidia on its own terms
Yu Kai said plainly at the 2026 Yabuli Forum that the rival in this field is still Nvidia. Nvidia’s Jetson has been the de facto robot-compute standard for a decade. Yole puts Nvidia at about 69 per cent of the global robot SoC market. Agility’s Digit and Boston Dynamics’ Atlas run on Jetson Thor. D-Robotics does not win on specs: S600’s 560 TOPS sits between Nvidia’s older Orin at 275 TOPS and the newest Thor, and the two use different precision units so they are not directly comparable.
D-Robotics’ three real cards are cost, power and supply security. Jetson keeps rising in price, pushing chip cost uncomfortably high in thousand-yuan consumer robots. D-Robotics’ single-chip design drops the external MCU. Sunrise 5 draws about 3 watts against Nvidia’s Thor at 40 to 130 watts. Home robots often have only a 25 to 40 watt budget, so high-power modules simply do not fit. And export-control uncertainty turns “whose chip” from a technical choice into a risk choice, making domestic silicon a safety choice for many Chinese robot makers.
The three cards are not enough. D-Robotics’ real challenge is ecosystem. Nvidia’s JetPack, Isaac, Omniverse and GR00T form a toolchain from simulation to deployment, with millions of developers bound to CUDA, an inertia no parameter or price can break. Wang Cong concedes that simulation and training may stay on Nvidia first, then adapt to the deployment chip. So in the short term D-Robotics does not replace Nvidia but occupies the “train on Nvidia, deploy on D-Robotics” slot.
The bet is that robotics, like cars, moves from vertical integration to horizontal division of labour, with professional chip and software platforms. The answer is not in today’s USD 400 million but in the production data its customers ship over the next twelve to eighteen months.
Editor’s note: This is an adapted translation of the original OFweek Robotics report. It has been trimmed and restructured for readability for an international business audience. The full original (in Chinese) is at https://robot.ofweek.com/2026-09/ART-898890-12003-30704773.html.