Investment ResearchTopic research, evidence curves, and cycle signals
HomeSign inEN

Research library · updated 2026-06-12 · public

Fourier Intelligence / 傅利叶 public evidence packet v1

Date: 2026-06-12 Owner: Hugo / Genius Team Research owner: Finance / Charlie AGT-002 Visibility: PUBLIC Status: source-backed company evidence packet Public-safety flag: evidence map only; no trade recommendation; no private portfolio data included.

One-line answer

Fourier / 傅利叶 is best framed as a rehabilitation-robotics incumbent moving into humanoids: it has stronger institutional footprint and product-line continuity than a pure demo company, but public evidence still supports S3/S4 productized-platform and pilot-use-case signals rather than S5 scaled commercial humanoid economics. 🟢 company website accessed 2026-06-12; 🟠 stage classification by Charlie, 2026-06-12.

Core question

Can Fourier convert its rehab robotics installed base, actuation know-how, and GRx humanoid line into repeatable paid humanoid deployments with measurable utilization, reliability, customer ROI, and revenue/margin contribution?

Market-definition lens

  • Layer: robot body / OEM + embodied-AI application layer.
  • Adjacent base: rehabilitation robotics and institutional robotics deployments.
  • Not counted here: public-equity supplier exposure, unless separately validated through filings.
  • Current evidence type: official product specs, official company history, official use-case descriptions, and official institutional-footprint claims. 🟢 Fourier official pages accessed 2026-06-12.

Current evidence

EvidenceWhat is source-backedWhy it mattersSignal grade
Company baseFounded in 2015; evolved from exoskeleton robots into humanoid robots and intelligent rehabilitation systems. 🟢 Fourier overview, accessed 2026-06-12.Not a new shell around one humanoid demo; has 10+ years of robotics history as of 2026.S3
Institutional footprintProduct lineup serves more than 2,000 institutions across 40 countries / regions. 🟢 Fourier overview and Fourier Rehab page, accessed 2026-06-12.Gives Fourier a real customer-access base, especially in rehab / institutional channels; still not humanoid-specific revenue.S4 for institutional base; S3 for humanoid implication
Humanoid roadmapGR-1 introduced in 2023; GR-2 in 2024; GR-3 Series and open-source N1 / FourierActionNet listed for 2025 in official history. 🟢 Fourier overview, accessed 2026-06-12.Shows multi-generation humanoid cadence rather than a single prototype.S3
GR-1 hardware165 cm height, 55 kg weight, 44 joints, 5 km/h max speed, 230 N.m max peak torque; uses FSA modules integrating motor, driver, reducer, and encoder. 🟢 GR-1 product page, accessed 2026-06-12.Productized full-body humanoid specs; relevant to body/platform capability.S3
GR-1 perceptionPage describes pure-vision system using BEV, Transformer, Occupancy Network and six RGB cameras to generate a real-time 3D occupancy grid. 🟢 GR-1 product page, accessed 2026-06-12.Shows an AI/perception architecture claim; still vendor-stated, not independently benchmarked deployment autonomy.S3
GR-2 upgrade175 cm height, 63 kg weight, 2-hour battery, 53 joints, 5 km/h speed, peak torque 380 N.m, 12-DoF dexterous hands and six array-type tactile sensors. 🟢 GR-2 product page, accessed 2026-06-12.Hardware capability improves vs GR-1; tactile/dexterous-hand design matters for manipulation tasks.S3
GR-3 positioningGR-3 Series framed as care / companion robots; 165 cm height, 71 kg weight, 55 joints, 3-hour battery; supports whole-body teleoperation. 🟢 GR-3 Series product page, accessed 2026-06-12.Pushes Fourier toward human-centered/service scenarios, but teleoperation support highlights autonomy still needs measurement.S3
Dexterous handFDH-6 has 6 degrees of freedom, 515 g weight, fingertip positioning accuracy of 0.5 mm, and stable performance over 300,000 cycles under stated continuous-operation framing. 🟢 FDH-6 product page, accessed 2026-06-12.Component capability may matter if manipulation becomes bottleneck; needs customer deployment proof.S3
Rehab baseFourier Rehab serves more than 2,000 institutions across over 40 countries; RehabHub integrates over 30 types of robots. 🟢 Fourier Rehab page, accessed 2026-06-12.This is the most concrete non-demo base, but it is rehab robotics rather than humanoid economics.S4 for rehab base; S3 for humanoid transfer
Business-use pilotsOfficial business page says GR-1 works with China Construction Bank Shanghai Branch on a robot training center for reception-manager scenario, and with SAIC-GM on factory challenges including obstacle avoidance and high-voltage component assembly. 🟢 Fourier business page, accessed 2026-06-12.Named counterparties and tasks are stronger than generic demos, but public page lacks robot count, contract value, runtime, uptime, or ROI.S4 monitoring signal, not S5

Current stage classification

  • Rehabilitation robotics base: commercial deployment / installed institutional base. 🟢 Fourier official footprint claims, accessed 2026-06-12.
  • Humanoid GRx line: productized platform + named pilot/application exploration. 🟢 Fourier product/business pages, accessed 2026-06-12.
  • Overall humanoid commercialization label: S3/S4, not S5. 🟠 Charlie classification, 2026-06-12.

Why not S5 yet:

  • No public contract value for humanoid deployments found in reviewed official pages. 🟢 source absence from reviewed official pages, accessed 2026-06-12.
  • No disclosed humanoid robot revenue, gross margin, unit shipments, utilization, uptime, intervention rate, customer ROI/payback, or repeat-order economics found in reviewed official pages. 🟢 source absence from reviewed official pages, accessed 2026-06-12.
  • Official pages describe teleoperation support for GR-3 Series; without intervention-rate data, this cannot be treated as full-autonomy deployment proof. 🟢 GR-3 Series page, accessed 2026-06-12; 🟠 interpretation.

Current answer

Fourier looks more credible than a one-off humanoid demo because it has: 1) a decade-plus robotics history from 2015 to 2026, 2) >2,000-institution / 40-country institutional footprint, 3) a multi-generation GRx humanoid roadmap, and 4) named business exploration with CCB Shanghai Branch and SAIC-GM. 🟢 Fourier official pages accessed 2026-06-12.

But the public evidence does not yet show scaled humanoid economics. The strongest public claim is “productized platform plus institutional/customer-access base”; the missing bridge is “paid humanoid deployment with measured economics.” 🟠 Charlie synthesis, 2026-06-12.

Counter-answer / bear case

Fourier’s rehab installed base may not transfer cleanly to general-purpose humanoids. Rehab robots operate in structured clinical/institutional settings, while humanoids in reception, factory, care, or service settings require reliability, autonomy, safety, maintenance, and ROI under broader real-world conditions. 🟠 Charlie synthesis, 2026-06-12.

GR-3’s whole-body teleoperation support is useful for data collection and remote operation, but it also means public materials should not be read as proof of autonomous commercial deployment without intervention-rate data. 🟢 GR-3 Series page accessed 2026-06-12; 🟠 Charlie interpretation.

Signal vs noise

Signals

  • S4: >2,000 institutions across 40 countries / regions served by Fourier product lineup / Rehab business. 🟢 Fourier overview and Rehab pages, accessed 2026-06-12.
  • S4: named application exploration with China Construction Bank Shanghai Branch and SAIC-GM. 🟢 Fourier business page, accessed 2026-06-12.
  • S3: GR-1 / GR-2 / GR-3 multi-generation humanoid specs and product pages. 🟢 Fourier product pages, accessed 2026-06-12.
  • S3: FDH-6 dexterous hand specs including 0.5 mm fingertip positioning accuracy and 300,000-cycle stated reliability. 🟢 FDH-6 page, accessed 2026-06-12.

Noise / not enough yet

  • “First mass-produced humanoid robot” language on GR-1 product page is not by itself production-volume evidence unless paired with disclosed units, yield, revenue, or deliveries. 🟢 GR-1 product page, accessed 2026-06-12; 🟠 Charlie caveat.
  • Customer logo / named collaboration is not revenue or ROI evidence without contract economics and customer-confirmed performance metrics. 🟠 Charlie rule, 2026-06-12.
  • Hardware specs do not prove deployment reliability, autonomy, margin, or customer payback. 🟠 Charlie rule, 2026-06-12.

What would change our mind

Upgrade toward S5 if Fourier discloses any of the following:

  1. Humanoid robot deliveries by year / quarter, with units and customer mix. 🟢 if official filing/company disclosure.
  2. Named paid deployment with robot count, contract value or recurring-fee model, and task scope. 🟢 if official customer/company disclosure.
  3. Runtime, uptime, intervention rate, safety incident rate, and maintenance burden in CCB / SAIC-GM / rehab or service scenarios. 🟢 if official / customer-confirmed.
  4. Humanoid-specific revenue, gross margin, backlog, or repeat-order data. 🟢 if official financial disclosure.
  5. Evidence that rehab institutional channels convert into repeat humanoid sales rather than one-off pilots. 🟢/🟡 depending on source.

Downgrade if:

  1. Future updates remain demo/spec-heavy without deployment KPIs through 2026-2027. 🟠 monitoring rule.
  2. Teleoperation remains central and intervention rates are not disclosed. 🟠 monitoring rule.
  3. Institutional base remains rehab-only with no measurable humanoid conversion. 🟠 monitoring rule.

Common misconception

Misconception: “Fourier already has >2,000 institutions, so humanoid commercialization is proven.”

Correction: The >2,000-institution claim is important, but the official wording refers to Fourier’s broader product lineup / rehab business, not disclosed humanoid revenue, humanoid unit deployments, or humanoid ROI. Treat it as customer-access and robotics-execution evidence, not S5 humanoid economics. 🟢 Fourier overview/Rehab pages accessed 2026-06-12; 🟠 Charlie interpretation.

Think Deeper questions

  1. Does a rehab robotics installed base shorten humanoid sales cycles, or does it mainly validate a different clinical-device business?
  2. In care/reception/service settings, what intervention rate is low enough for customer payback to work?
  3. If Fourier’s advantage is institutional deployment know-how rather than lowest-cost hardware, which value layer captures margin: robot body, software/workflow, or service/maintenance?

Public site draft section

Fourier / 傅利叶：从康复机器人走向 humanoid，关键是“转化率”

Fourier 的公开证据不像纯 demo 公司：官网显示公司 2015 年成立，产品线覆盖 GRx humanoid、RehabHub、Galileo 等，并服务 40+ 国家 / 地区的 2,000+ 机构。GR-1、GR-2、GR-3 形成多代 humanoid 路线；GR-2 给出 175 cm、63 kg、2 小时电池、53 关节、380 N.m 峰值扭矩等规格；GR-3 转向 care / companion 场景，并支持 whole-body teleoperation。🟢 Fourier official pages, accessed 2026-06-12.

这说明 Fourier 有产品化和机构客户基础，但还不能说明 humanoid 已经进入 S5。我们还缺少：humanoid 出货量、合同金额、客户 ROI、运行时长、uptime、人工介入率、服务成本、毛利率和复购。当前更合理的标签是 S3/S4：产品化平台 + 机构渠道 + named application exploration。🟠 Charlie synthesis, 2026-06-12.

Footer: Evidence map, not winner ranking. Rehab installed base ≠ humanoid scaled economics. Named collaboration ≠ contract economics. Teleoperation support ≠ full-autonomy proof.

Source list

  1. Fourier overview, https://www.fftai.com/overview, accessed 2026-06-12. Reliability: 🟢 primary company page. Key claims: founded 2015; evolved from exoskeleton robots into humanoid robots and rehab systems; >2,000 institutions across 40 countries; history milestones 2019 humanoid project, 2022 GR-1 born, 2023 GR-1 introduced, 2024 GR-2, 2025 GR-3 / N1 / FourierActionNet.
  2. Fourier GRx series, https://www.fftai.com/grx, accessed 2026-06-12. Reliability: 🟢 primary company page. Key claim: GRx humanoid robot series is bipedal, general-purpose bionic robot product lineup; “Made for AI” positioning.
  3. Fourier GR-1 product page, https://www.fftai.com/products-gr1, accessed 2026-06-12. Reliability: 🟢 primary company page. Key metrics: 165 cm, 55 kg, 44 joints, 5 km/h max speed, 230 N.m max peak torque, FSA integrated actuator, six RGB camera pure-vision perception description.
  4. Fourier GR-2 product page, https://www.fftai.com/products-gr2, accessed 2026-06-12. Reliability: 🟢 primary company page. Key metrics: 175 cm, 63 kg, 2-hour battery, 53 joints, 380 N.m peak torque, 5 km/h speed, 12-DoF dexterous hands, six array-type tactile sensors.
  5. Fourier GR-3 Series product page, https://www.fftai.com/products-gr3series, accessed 2026-06-12. Reliability: 🟢 primary company page. Key metrics/claims: care / companion positioning; 165 cm, 71 kg, 55 joints, 3-hour battery; whole-body teleoperation support.
  6. Fourier FDH-6 product page, https://www.fftai.com/products-fdh6, accessed 2026-06-12. Reliability: 🟢 primary company page. Key metrics: 515 g, 6 degrees of freedom, 0.5 mm fingertip positioning accuracy, stable performance over 300,000 cycles.
  7. Fourier Rehab page, https://www.fftai.com/products-fourier-rehabilitation, accessed 2026-06-12. Reliability: 🟢 primary company page. Key claims: Fourier Rehab serves >2,000 institutions across >40 countries; RehabHub integrates over 30 types of robots.
  8. Fourier business page, https://www.fftai.com/industry-business, accessed 2026-06-12. Reliability: 🟢 primary company page. Key claims: collaboration with China Construction Bank Shanghai Branch on robot training center / reception-manager scenario; collaboration with SAIC-GM on obstacle avoidance and high-voltage component assembly.