Feather Robotics pulled in $7.6 million in pre-seed funding announced on September 24, 2026, led by Gradient Ventures and joined by Builder Capital, Geometry, SEED Innovations and Virgo VC. The San Francisco company has a straightforward pitch: sell a relatively affordable bimanual humanoid platform to AI developers who want to experiment without committing to a single software stack, and do it while already booking revenue.
The $29,990 robot — a bargain by humanoid standards — has crossed seven figures in revenue, the company said, though it declined to break out unit sales or specify when it reached that milestone.
Hardware without the handcuffs
Feather's approach diverges from the vertically integrated model favored by higher-profile competitors like Tesla's Optimus and Figure. Those companies build the chassis, write the software and train the models in-house. Feather, by contrast, sells the hardware and SDK separately from what it calls the AI "brain," allowing customers to plug in models from whichever robotics AI provider they prefer.
"You can't buy a Tesla robot today and develop on top of it," CEO Hoa Mai told TechCrunch in September. The comment, perhaps more pointed than diplomatic, captures Feather's bet: that a modular ecosystem will appeal to startups and integrators who'd rather not get locked into proprietary systems.
The robot itself sports two seven-degree-of-freedom arms on a torso capable of 600 millimeters of vertical travel. The holonomic base reaches speeds of 1.3 meters per second. Dual hot-swappable batteries — 48-volt, 50-amp-hour units — are rated for 10 hours of operation, according to specifications the company published. Configurable compute ranges from Jetson Nano up to Jetson Thor. The software development kit supports Python 3.9 and higher, ROS2, teleoperation out of the box, inverse kinematics and collision detection.
Those specs matter less for what they reveal about technical prowess than for what they signal about market positioning. Feather is undercutting Chinese manufacturers on price while offering flexibility. It cited Unitree's H2 EDU at around $68,000, Agibot's G2 near $135,000, and Dexmate's Vega at roughly $83,000. US distributors have listed the Unitree H2 EDU between $66,800 and $68,900 recently, according to dealer websites.
The team and the timeline
Mai brings unusual credentials. He previously founded Kind Humanoid, which 1X acquired earlier this year, TechCrunch reported. Chief Product Officer Parsa Bakhtiari worked through Tesla's Model 3 production troubles in 2018, reporting directly to Elon Musk during what the company describes as the "production crisis." Whether that experience translates to scaling a robotics startup remains an open question, but the pedigree doesn't hurt.
The company is incorporated in Delaware, with a business address at 2261 Market St. Ste 22925, San Francisco, CA 94114, according to an SEC filing. The company said it shipped its first prototype within two months and landed its first order in nine months, a timeline that suggests either exceptional product-market fit or a small number of early adopters willing to take risks on unproven hardware.
Customers and traction
The startup said it already serves a manufacturer with $4 billion in annual revenue, though it wouldn't name the client. TechCrunch noted robots deployed in restaurants and science labs. The vagueness is typical for early-stage companies reluctant to lose customers to competitors or expose commercial relationships before contracts solidify.
Still, any revenue at this stage separates Feather from the many robotics ventures that burn through capital on R&D without landing paying customers. The company crossed $1 million in revenue, according to its announcement, though again without specifying the timeframe.
The funding mechanics
The round was structured as SAFEs. SEED Innovations previously disclosed a $1 million SAFE with a $60 million valuation cap, according to a regulatory filing on the London Stock Exchange. An SEC Form D showed $6.015 million sold to 15 investors, with a first sale date in late March. The gap between that figure and the announced $7.6 million total suggests additional SAFE commitments closed afterward.
Feather had between 11 and 50 employees as of late September, according to its LinkedIn profile — a range that underscores how early-stage the operation remains.
The company frames its mission in sweeping terms. Mai talks about building "infrastructure to scale physical AI" for teams "no matter their size." That language positions Feather less as a hardware vendor and more as an enabling platform for what it calls "Physical AI" companies tackling manufacturing, logistics and service work.
Whether the market coalesces around open platforms or vertically integrated stacks — the perennial question in emerging technology categories — will determine if Feather's bet pays off. For now, the startup has funding, early revenue and a product shipping to customers. That puts it ahead of most robotics ventures, even if the path from here to sustainable scale remains uncertain.
