Spirit AI is building around China's industrial data loop

Spirit AI and D-Robotics announced a strategic partnership on May 20, 2026, pairing Spirit v1.5 with the D-Robotics Sunrise S600 edge-AI compute platform. The announcement added another layer to a fast-forming stack.

Spirit AI, also known in Chinese as Qianxun Intelligence, was founded in 2024 and has moved quickly from model-and-robot development into industrial partnerships. Its public robot is Moz, a force-controlled humanoid platform tied to Spirit VLA work. The company has harder-to-replicate assets in the environments forming around that robot: places where machines can fail, collect data, encounter edge cases and be pushed through real operating constraints.

Founder Mix: Industrial Robotics And Embodied AI

Spirit AI rapid rise is easier to understand through its founders. Han Fengtao previously co-founded Rokae Robotics and served as CTO, bringing more than 12 years of industrial robotics experience. ION Analytics reported that he led delivery of more than 20,000 robot units across over 20 industries. Rokae is an industrial robot company, not an academic lab, so Han background points directly toward manufacturing customers, robot reliability and deployment execution.

Co-founder Gao Yang brings the other side of the company: embodied AI, computer vision, machine learning and robot manipulation research. Yicai described Gao as having more than 10 years of research experience in those areas, including visual-language work associated with ViLa and robotic manipulation research using foundation vision-language models. The pairing mirrors Spirit strategy: industrial robotics deployment on one side, foundation-model-style embodied intelligence on the other.

The founder mix helps explain why Spirit accumulated Bosch, JD, CATL, D-Robotics and large industrial/state-linked capital so quickly for a company founded in 2024. The company could speak to factory execution, not only model capability, while also presenting a VLA roadmap that fit the current embodied-AI investment cycle.

Rapid capital

Spirit disclosed nearly RMB 2 billion in two February rounds, roughly $275 million at early-2026 exchange rates. The investor list mixed financial capital with strategic and state-linked backing: Yunfeng Capital, HongShan, Chaos Investment, Synstellation Capital, TCL Capital, 360 Capital, HouXue Capital, funds from Chongqing and Hangzhou, and returning investors including Shunwei Capital, Prosperity7 and Fortune Capital.

On April 7, the company announced another RMB 1 billion round, about $140 million, co-led by Shunwei Capital and YF Capital with Fortune Capital, Galaxy Yuanhui Investment, Xin Ding Capital and Glacier Capital participating. Spirit AI framed the combined sprint as $450 million secured within 30 days.

Source: Spirit AI

The financing followed an earlier JD-led round. Spirit said in its Pre-A+ funding announcement that JD.com led a nearly RMB 600 million investment, with China Internet Investment Fund, Zhejiang Science and Technology Innovation Mother Fund, Huatai PCPD Investment and Fosun RZ Capital participating. The cap table gives Spirit more than money: it links the company to consumer retail, logistics, batteries, electronics, industrial groups and public-sector robotics priorities unusually early in its life.

JD Gives Spirit Retail And Logistics Contact

Spirit and JD Group signed a strategic partnership covering 2026 through 2029. The agreement spans consumer product customization, technical integration, deployment and joint marketing, with Moz robots placed in JD MALL physical stores for high-precision coffee brewing and service demonstrations. JD had already become a shareholder through the Pre-A+ round, making the relationship both financial and operational.

Retail stores and logistics channels bring human traffic, ordering patterns, service interruptions, maintenance routines and object variability. Coffee service is a narrow task, but a narrow task in a live store can still produce interaction data that a staged demo cannot. JD also has a broader embodied-intelligence data strategy across retail, warehousing, healthcare, home service and urban operations, giving Spirit a possible route into repeatable environments where robot behavior can feed model iteration.

CATL Puts Moz Into Battery-Pack Work

CATL is the clearest production reference. Spirit AI account of the deployment says Moz robots entered the CATL Zhongzhou Base on a new-energy battery PACK production line, handling high-voltage connector insertion in end-of-line battery-pack testing. The company reported connection success above 99 percent and cycle time at the level of skilled human workers. The same account describes Moz as developed by Spirit AI Robotics, an ecosystem partner of CATL, and powered by CATL self-developed batteries.

Battery-pack testing is a useful task because it is physically constrained but unforgiving. Connector work involves position variation, flexible harnesses, force control and safety-critical electrical context. Public material still does not show fleet size, uptime, maintenance intervals or economics.

Bosch Adds Factory Scenarios And Component Depth

Spirit May 2026 agreement with Bosch China moved the company deeper into industrial infrastructure. The Bosch announcement describes a partnership to industrialize a general-purpose robot "Universal Brain," combining Spirit VLA models with Bosch China-based factories, logistics centers, sensors, actuators and manufacturing support. The release names a "Real-World Data - Embodied AI - Real-World Scenarios" pipeline inside Bosch environments.

Bosch brings surroundings a robot model company cannot fabricate alone: production cells, logistics flows, component constraints, validation expectations and industrial customers. A model that works only in a clean lab gives limited proof. A model tested against Bosch factory and logistics scenarios has a better chance of exposing edge cases that matter for deployment, especially when the partner also contributes sensors, drives and manufacturing feedback.

D-Robotics Moves The Stack Toward Edge Inference

The D-Robotics agreement adds compute to the sequence. According to the May 20 announcement, Spirit v1.5 has been adapted to the D-Robotics Sunrise S600 embodied-compute platform, combining Spirit model and deployment work with D-Robotics edge-AI computing hardware. Embodied models have to run with latency, power and reliability constraints that cloud-only architectures cannot always meet inside robots.

The partnership also shows how quickly the Chinese embodied-AI stack is splitting into specialist layers. Spirit brings model, robot and deployment references, D-Robotics brings robot compute, Bosch brings factories and component systems, JD brings retail and supply-chain surfaces, and CATL brings a battery-production task with energy-system connection. This increasingly resembles industrial platform coordination: models, deployment surfaces, compute, manufacturing infrastructure and strategic capital forming around the same robot program.

What Spirit Has Proved, And What It Has Not

Spirit has proved that it can attract major capital, secure strategic partners and place Moz in more concrete environments than most young humanoid companies. The differentiated part is not Moz alone; it is the operational network around Moz, where partners can supply data, infrastructure, validation and deployment pressure.

The public record does not yet prove deployment scale, revenue, unit economics, robot uptime, autonomy under shift conditions, fleet size or production cost. CATL reported connector success rate is meaningful inside that task, but it should not be stretched into a general claim about humanoid readiness.

Have a robotics update Korthos should review? Send news, deployments, product releases, funding rounds, research, or media to tips@korthos.xyz or reach out on X at @agkorthos.