--- title: "Fourier's GR-3 care humanoid bets that being touchable matters more than being strong" date: 2026-10-21 category: Humanoids & Robotics site: NeuroAI canonical: https://neuroai.site/a/na-robot-fourier-gr3-carebot language: en --- # Fourier's GR-3 care humanoid bets that being touchable matters more than being strong > Shanghai's Fourier built the GR-3, a 165 cm care-focused humanoid wrapped in soft skin, betting eldercare and hospitals want a robot people don't flinch at. In a Shanghai rehab clinic, a 71-kilogram robot with a soft polymer shell leans toward a patient and offers a hand. Nobody backs away. That reaction — or the lack of it — is the entire product thesis. Most humanoid (人形机器人) makers are racing to lift heavier boxes and walk faster. Fourier Intelligence (傅利叶) is betting the opposite: that the first robot ordinary people will tolerate standing next to is the one built to be safe, gentle, and a little bit friendly. ## A rehab company that wandered into humanoids Fourier was founded in 2015 by Alex Gu (顾捷) and Zen Koh, and for its first decade it made rehabilitation exoskeletons and therapy robots — not bipeds. That background is the whole point. The company says it has shipped more than **2,000 rehabilitation robots** to over **40 countries** and serves more than 2,000 institutions. Selling into clinics taught it something most robot startups lack: how to clear procurement, training, and trust barriers in regulated care settings. The humanoid line is newer. The GR-1 arrived in 2023 as a billed "mass-produced" humanoid; the GR-2 followed in September 2024 with 53 degrees of freedom; the GR-3, unveiled in August 2025, is the deliberate turn toward care. ## The GR-3: designed to be touched The GR-3 is a full-size biped at **165 cm** and **71 kg**, with a standard **43 degrees of freedom** (rising to 55 with optional hands), about **3 kg of arm payload**, a walk speed near **6.1 km/h**, and roughly **3 hours** of runtime on hot-swappable battery packs. Its sensing stack is where the care intent shows: 31 distributed pressure sensors, a four-microphone array, and RGB-plus-structured-light face tracking. Fourier shaped the machine to be approached, not avoided. The shell is a soft polymer; the face is animated; the gait modes include a "bouncy walk" and even a "fatigue mode" that slows and shifts weight when charge runs low. Under the skin, the GR-3 pairs fast reflexive control with a large-model reasoning layer — quick reactions for balance and touch, slower language-model reasoning for conversation and intent. It made its U.S. debut at CES in January 2026, aimed at care facilities and research labs rather than factory floors. ## The money behind the bet Fourier is private, but its funding tells you how seriously investors take the care angle. The company closed a roughly **800 million yuan (≈ US$113 million / HK$870 million)** E round in January 2025, then a **300 million yuan (≈ US$42 million / HK$330 million)** E3 strategic round in June backed by listed materials firm Runyang Technology. Pre-money valuation reached about **8 billion yuan (≈ US$1.1 billion / HK$8.7 billion)**, putting Fourier among the unicorns of Chinese robotics. The proceeds were earmarked for humanoid mass-production lines, and the company has discussed delivering around 300 biped units in 2025. That is a fraction of what high-volume makers ship, and Fourier knows it. Its argument is not volume; it is fit. Where rivals chase factories and warehouses, Fourier is going after hospitals, eldercare, and public spaces, where certification and trust beat payload. ## Why care is a slower, harder market The honest version is that the care-bot is still a well-designed interaction platform looking for a proven job. No medical-device certification has been reported for the GR-3, so it cannot be billed as therapy. Its ~3 kg arm payload rules out real physical assistance — it cannot lift a person. And there is no published price; some reports cite institutional quotes above **200,000 yuan (≈ US$28,000 / HK$218,000)**, while distributor listings vary widely, so treat any number as indicative only. China's aging demographics are the tailwind everyone cites — state estimates put the "silver economy" near 7 trillion yuan in 2024 and project 30 trillion by 2035 — but a demographic trend is not a purchase order. Care buyers are cautious, regulated, and slow. ## Honest limitations This article relies on Fourier's own product disclosures and on Chinese trade-media reporting of its funding and delivery plans; the degree-of-freedom, payload, and sensor specs are the company's published figures, not independent measurements. The GR-3 is not in mass production by Fourier's own framing, and the "300 units in 2025" target is a company plan, not a shipment tally. The care use case remains largely demonstration and research today, with no reported medical certification. Price figures are inconsistent across sources and should be read as indicative. The deeper question — whether a soft, gentle humanoid actually earns a place in a clinic — is unanswered by any deployment data we have. ## What readers can do now - **Judge care humanoids by certification status, not demo videos** — until Fourier or peers report medical-device clearance, the GR-3 is a research and interaction platform, not a therapy device. - **Weigh payload against the actual job**: a ~3 kg arm rules out most physical assistance, so evaluate the GR-3 on interaction and monitoring, not lifting. - **Watch Fourier's rehab-install base as the real moat** — 2,000+ institutions in 40+ countries is the asset rivals from AI labs simply do not have when they enter care. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-robot-fourier-gr3-carebot Free to quote with attribution and a link to the original.