（本文作者为 Chelsea_Sun，钛媒体经授权发布）
NextFin News — Fourier Intelligence has spent more than a decade in robotics without becoming a household name outside the industry. Founded in Shanghai in 2015, it started with rehabilitation machines, moved into bipedal platforms, and by early 2025 had delivered more than 10,000 robots across categories to institutions in dozens of countries.
What it had not done was dominate the humanoid shipment tables. A 2025 CCID market study put its humanoid deliveries at roughly 300 units that year—useful for research, education, and data work, small next to volume leaders such as Unitree.
That quieter profile broke when Meta needed a body for lab experiments tied to Muse, its personal AI assistant, and to a broader push into machines that can act in the physical world. At Fourier’s factory, Meta’s robotics product leadership spent extended time on an assembly line.
In about three months the two sides produced Max: a humanoid about 1.5 meters tall and 48 kilograms, able to pick up voice and gesture at short range, with a face display for basic states. The stack runs Meta’s agent and interaction logic on a chassis derived from Fourier’s GR series, heavily reworked in proportions, head, cameras, hands, and footwear. A Max 2 pass has been described as redesigning torso and manipulators and adding stair-related sensing.
Meta says the project is internal research, not a product for sale. Fourier has not offered a detailed public narrative. No purchase volumes or contract values have been disclosed. For Fourier, the visible prize is validation—that a team steeped in medical-grade interaction and three generations of GR hardware can absorb a hyperscaler’s brief and return a walking prototype on a startup calendar.
Fourier’s economic base is easier to describe than to measure. Rehabilitation robotics taught it safety, compliance, and sustained contact with hospitals and care settings. Overseas revenue has been reported around 20% of the total. Humanoid revenue, according to trade coverage, still leans on research, education, and data-collection customers rather than fleets of workers or household machines. Complete 2025 revenue and profit figures are not public, so the cash contribution of the rehab line versus the burn of humanoid development remains opaque to outsiders.
The GR platform is the bridge Meta used. It is also the asset Fourier must eventually turn into either branded end products or a manufacturing role others pay for at scale. Peer comparisons are unkind on pure humanoid volume, kinder on breadth: few Chinese robotics firms combine a long rehab franchise with a credible bipedal line and a factory that can customize deep. That combination explains why a company that is not the shipment leader can still win a prototype slot after Meta had, according to earlier Bloomberg reporting, also talked to other humanoid players.
Investors have already marked a price. SoftBank’s Vision Fund participated in a 2022 round; further funding in 2025 included a capital injection disclosed by Runyang Technology at a pre-money valuation of 8 billion yuan, with payment reported completed in August that year. Fourier completed a corporate restructuring toward a listing-ready form in 2025 and has pointed at Shanghai’s STAR Market. As of early November 2026 it had not filed a prospectus or appeared in formal tutoring records. Meanwhile competitors have listed or entered exchange review, and public valuations in the sector have moved quickly with narrative, scarcity, and revenue mix.
An 8-billion-yuan mark grounded in rehab plus a few hundred humanoids a year invites a simple question: what converts into durable profit? Prototype work for a U.S. tech giant supports the technology story and may help recruiting and future fundraising. It does not substitute for the recurring revenue, customer concentration, and loss-narrowing trajectory that public-market investors—and, if informal regulatory focus reported in the industry is any guide, sponsors preparing robotics IPOs—will eventually interrogate.
Fourier has shown it can do the first job in the chain: take a specification and deliver a research body fast. The next jobs are harder. Volume manufacturing rewards cost, yield, and supply assurance. Platform leadership rewards models, data, and standards—territory Meta is explicitly building in-house. Brokerage snapshots of robotics value chains have long shown thinner margins in whole-machine assembly than in some components. Being early can still leave a hardware partner in a low-price slot once the design stabilizes and procurement sharpens.
External constraints apply if any Fourier-built machine is ever marketed into the United States. FCC rules updated in 2026 tightened equipment authorization for certain foreign-produced advanced robotics, a compliance gate distinct from export bans but material for commercial entry. Meta’s own job ads in China around supplier and production readiness signal preparation for scale somewhere in the program—not a published commitment that Fourier will own that scale.
Fourier’s bind is therefore straightforward. It must keep the rehab business healthy, lift humanoid shipments and revenue quality beyond demo and education accounts, and decide whether Meta-like work is a showcase, a repeating ODM line, or a bridge to Fourier-branded products in care and service settings. Catching Meta’s hand moves the company into the global frame.
Escaping a long valuation plateau still depends on Fourier’s own orders, margins, and listing readiness—not on whose logo appears beside a prototype in a Menlo Park hallway.
更多精彩内容，关注钛媒体微信号（ID：taimeiti），或者下载钛媒体App