On September 17, D-Robotics, a supplier of robotics hardware and software infrastructure, announced it had closed a $400 million Series C round. Mirae Asset led the round, with Meituan investing strategically. Government-backed platforms including Hefei State-Owned Assets Holding, Nanshan Strategic Emerging Industries Fund and Jingquan Capital took part, alongside Cathay Capital, Huamei International, GF Xinde and More Than Moore.

Existing shareholders including Hillhouse Ventures, 5Y Capital, Linear Capital, Huangpu River Capital, Temasek's Vertex Growth, Prosperity7, Yunfeng Capital and Meituan's Longzhu fund put in additional money.

D-Robotics said the proceeds will go mainly toward filling out the Sunrise chip family across all computing power tiers, and toward building a full-stack software platform spanning data collection, model training, simulation testing and inference deployment, giving different classes of robots a common hardware and software foundation.

No Robots of Its Own, Positioned as the "Brain Layer" for Embodied AI

Spun out of Horizon Robotics, D-Robotics positions itself as a supplier of robot computing platforms and development infrastructure. Unlike companies that build humanoid or industrial robots outright, it sells compute chips, development tools and integrated hardware-software platforms to robot makers.

Robots must continuously process data from cameras, microphones and other sensors, then handle environmental perception, task interpretation, path planning and motion control within tight power and latency budgets. Requirements for compute performance, interfaces, power draw and model deployment vary sharply between robot form factors.

D-Robotics wants the Sunrise line to span that range of computing power needs, with accompanying software tools that lower the effort robot makers spend on model porting and product development.

With the round closed, the company plans to keep expanding its chip lineup while building out a development path that runs from data generation to on-device inference, letting customers train models, run simulations and deploy to robot hardware on a single platform.

Sunrise Chip Shipments Pass 8 Million

D-Robotics also released updated business figures. By its own accounting, revenue in the first half of 2026 grew several times year over year, and cumulative Sunrise shipments have passed 8 million units.

Sunrise chips have so far gone mainly into robot vacuums, smart imaging devices and service robots. As interest in embodied AI builds, the company is pushing further into humanoid robots, industrial mobile robots, robotic arms and on-device embodied models.

Its current focus is the Sunrise S600, released in November 2025 for perception, decision-making and motion control workloads, paired with a development toolchain that spans cloud and device.

The company says more than 20 robotics firms and research institutions adopted the S600 within six months of launch, including TARS, Spirit AI, X Square Robot, Zhizai Wujie, UBTECH, Paxini, the Beijing Humanoid Robot Innovation Center, National and Local Co-built Humanoid Robotics Innovation Center in Shanghai, and Xiaoyu Zhizao.

Those partnerships span humanoid robots, industrial wheeled robots, flexible assembly for consumer electronics, embodied foundation models and multimodal sensing. D-Robotics says several have reached production-grade shipments, though it has not disclosed order sizes, revenue contribution or production timelines for individual projects.

From Chips Toward a Full-Stack Software Platform

Whether a robot chip reaches scale depends on more than peak computing power. Model compatibility, development tools, power efficiency and manufacturing consistency all matter.

Embodied AI companies typically have to run through data collection, labeling, model training, simulation testing, hardware bring-up and on-device deployment. Without common interfaces between chips, toolchains and robot systems, they burn engineering resources on integration work and product timelines stretch out.

One of the main uses for this round is building a software platform that covers those steps. As planned, it will link cloud training with on-robot inference, letting customers deploy vision-language models, motion control models and other embodied models onto robot hardware.

That marks a shift from selling chips alone toward a bundle of chips, development tools and a model deployment platform. Compared with one-off hardware sales, a complete software stack raises switching costs and platform stickiness, but it also demands more of the company on model support, tool reliability and technical service.

Developer Base Spans More Than 20 Countries

Beyond chips and software, D-Robotics is expanding a robotics ecosystem aimed at developers, universities and startups.

Its DGP program offers robotics startups help with product development, supply chain sourcing, marketing and investor pitching. It currently covers ten areas, including open-source humanoid robots, AI fitness coaching, embodied models, AI imaging and home companion robots.

D-Robotics says its platform now connects more than 500 small and mid-sized robotics developers, over 500 academic institutions, and more than 100,000 developers across more than 20 countries, who have built several hundred kinds of robot products on it.

Those figures for developer numbers, customer reach and product counts come from the company itself, and there is no consistent third-party industry data to verify them.

Mass Production Is the Next Test

Embodied AI is moving out of lab prototypes and demos into product validation and small-batch delivery. What robot makers ask of a compute platform is shifting from whether it can run a model to power draw, cost, real-time performance, reliability and the ability to supply at volume.

D-Robotics' advantage is that Sunrise chips already ship in meaningful numbers in consumer robots and smart hardware, giving it experience in edge computing and volume manufacturing. As the S600 lands in more humanoid and industrial robot projects, the company has a chance to carry that chip and toolchain capability into the embodied AI market.

Still, embodied AI remains early, and most humanoid robot projects have yet to generate steady large-scale orders. Design wins do not necessarily translate into volume shipments, and the adoption figures the company discloses say nothing directly about revenue or market share.

The things to watch next, then, are actual S600 shipments, revenue from the embodied AI business, how far customer projects get toward mass production, and whether the full-stack software platform actually cuts development and deployment costs for robot companies.