East Asian Technology Intelligence
Japan & China technology, translated and contextualized for Western readers
Industrial Automation & Robotics
傅利叶智能 / 上海傅利叶智能科技股份有限公司
A private robotics developer that transitioned from rehabilitation equipment into humanoid robots, building embodied-AI platforms, actuators, and manipulation data.
Fourier has a credible technical bridge from rehabilitation robotics to human-facing humanoids, differentiating it for care and rehabilitation use cases. However, its commercial visibility is constrained by a lack of public financial disclosures, unproven unit economics, and high exposure to export controls on advanced compute.
Fourier is a Shanghai-based private robotics company that transitioned from rehabilitation equipment into humanoid robots. Founded in 2015, the company supplies physical embodied-AI systems, including the GR-1, GR-2, and GR-3 humanoid robot families, as well as the open-source N1 humanoid platform. It also maintains its legacy business in rehabilitation robots and exoskeletons. Its products address rehabilitation, eldercare, research, education, industrial training, and selected factory workflows.
Fourier's structural position is that of an edge-device and enterprise-application provider. It builds the physical platforms, actuators, dexterous hands, and robot operating tools that allow AI models to interact with the physical world. Rather than developing general-purpose foundation models or cloud infrastructure, Fourier focuses on the hardware and manipulation data required for embodied AI. This positions the company as a critical enabler for researchers and enterprises looking to deploy physical AI systems in clinical, institutional, and industrial environments.
Fourier's AI relevance is hardware-led, focusing on embodied-AI platforms rather than foundation models or cloud infrastructure. Its primary AI product is the GR-2 humanoid platform, which features up to 53 degrees of freedom, proprietary FSA 2.0 actuators, and self-developed dexterous hands with tactile sensing. The company states that the GR-2 has experimental applications in industrial manufacturing, medical rehabilitation, academic research, and consultation.
Fourier's other direct AI asset is the ActionNet dataset. ActionNet contains more than 30,000 teleoperated bimanual-manipulation trajectories, representing about 140 hours of interaction data. Fourier releases this dataset under a creative commons license while open-sourcing the training pipeline and evaluation tools.
AI demand reaches Fourier through the growing need for embodied-AI research platforms, humanoid pilots, robot data collection, and healthcare automation. This link is direct for its GR-series and N1 sales, though the market remains early-stage and revenue exposure to AI is not publicly disclosed.
Fourier's principal technical distinction is its rehabilitation-robotics base. Medical rehabilitation requires compliant actuation, body-contact safety, controlled force delivery, repeatable motion trajectories, and clinical-use reliability. These capabilities are highly relevant to humanoid robots intended to interact physically with people, particularly in rehabilitation, eldercare, and assistance settings.
The GR-2 platform combines whole-body mobility with 53 degrees of freedom, FSA 2.0 actuators, dexterous hands, and tactile sensing. While this does not make Fourier unique across the humanoid sector, it provides a more clinically relevant physical-interaction starting point than companies that begin in warehouse, logistics, or consumer robot markets.
The hard-to-copy part of Fourier's position is its accumulated know-how around human physical interaction and rehabilitation workflows. The easier-to-copy part is its general humanoid hardware architecture, where Chinese competitors can source overlapping motors, reducers, sensors, controllers, and batteries from the same regional supply base. The N1 open-source release broadens its developer ecosystem but may reduce hardware differentiation.
Fourier depends on a supply chain of electric motors, reducers, encoders, batteries, sensors, compute modules, high-performance materials, and manufacturing partners. Specific upstream suppliers are not publicly disclosed, though its developer documentation references Alibaba Cloud's Wulanchabu object-storage endpoint for distributing software assets.
On the customer side, named commercial deployments are limited. Public reporting places Fourier robots in SAIC and OnStar-linked automotive factory training settings, as well as China Construction Bank's Shanghai Pudong branch. Fourier's rehabilitation systems are reported to operate internationally, but customer concentration and the share of sales from each customer are not disclosed.
The company is not a sole-source provider to cloud, chip, memory, server, or data-center supply chains, meaning its supply-chain centrality is limited to the emerging embodied-AI and rehabilitation robotics niches.
Fourier faces high exposure to U.S. and allied export controls. Humanoid robots depend on advanced processors, GPUs for training, high-performance sensors, machine-vision components, and development software. Some of these critical inputs may be subject to U.S. export restrictions or licensing requirements for China-based users. Fourier's ActionNet, simulation, teleoperation, and embodied-AI training work require substantial compute capacity, and domestic alternatives may not be fully interchangeable for all model-development stacks.
As a Chinese robotics developer, Fourier also faces regional trade and cross-strait risks. Taiwan-related disruptions could affect its access to semiconductors, sensors, and electronics, even though Fourier itself does not fabricate chips.
Furthermore, its focus on healthcare and eldercare applications exposes the company to data-security regulations. Teleoperation data, hospital data, and multimodal robot-interaction data create compliance exposure under China's data-security, personal-information, and cross-border-data rules.
| Risk | Severity | Why it matters |
|---|---|---|
| U.S. and allied export controls | High | Humanoid robots depend on advanced processors and GPUs that may be subject to export restrictions. |
| Dependence on imported advanced compute | High | ActionNet and embodied-AI training require compute capacity where domestic alternatives may not be interchangeable. |
| Component supply-chain concentration | High | Actuators, reducers, and sensors can become bottlenecks without publicly disclosed supplier diversification. |
| Early commercial demand | High | Humanoid deployments remain dominated by pilots and lack audited revenue demonstrating repeatable unit economics. |
| Strong domestic competition | High | Competes with Unitree, UBTECH, AgiBot, and others for capital, talent, and supply-chain capacity. |
| Private-company disclosure risk | High | Investors cannot independently assess revenue quality, cash burn, or customer concentration. |
| Healthcare and eldercare regulation | Medium | Rehabilitation applications expose Fourier to demanding device safety and clinical-validation requirements. |
| Data-security regulation | Medium | Teleoperation and hospital data create compliance exposure under China's data-security rules. |
Fourier is an independent private company, not a chaebol, keiretsu, or listed enterprise. Gu Jie serves as the recorded legal representative and is publicly identified as founder, chairman, and chief executive officer.
The company is not state-owned but is state-influenced, based on the participation of Shanghai government-background investment funds in its January 2025 E-series financing. These include Guoxin Investment, Pudong Venture Capital, and Zhangjiang Hi-Tech Investment. Prosperity7 Ventures, associated with Saudi Aramco, also participated in the round.
Because Fourier is private, its cap table, voting rights, board composition, shareholder agreements, and beneficial ownership percentages are not disclosed. This creates significant disclosure risk, as investors cannot independently assess revenue quality, cash burn, or customer concentration.
We haven't written about Fourier yet. Subscribe below to get it when we do.
Compiled with AI-assisted research from company filings, market data, and published reporting as of September 14, 2026, then reviewed by AsiaAI.FYI. Figures marked Estimate are not company-reported. Check primary filings before relying on any number.
Confidence: D. Corporate product documentation and press reports exist, but audited company financial disclosures, unit economics, and ownership percentages are not public.
Main sources: Corporate product documentation; Corporate-registration data; Transaction disclosures; Chinese financial press; Robotics trade reporting.
East Asian Technology Intelligence
Japan & China tech news — translated, contextualized, and delivered for Western readers.
Free. Unsubscribe anytime.