Robot.com once filled college quads with small wheeled bots dropping off burritos and textbooks. Now the San Francisco company wants to put human-shaped machines to work beside line cooks, warehouse pickers and hospital staff. The bet rests on R-noid, a wheeled humanoid powered by custom AI models. Fewer than 40 units operate commercially across roughly a dozen sites. Yet the move signals something larger. Delivery robots have scaled. Manipulation comes next.
The company, formerly called Kiwibot, built its reputation on Level 4 autonomous delivery. It deploys about 500 robots total. Those machines have logged more than 2.5 million tasks for clients including Sodexo, Grubhub and Aramark. Business Insider first detailed the humanoid expansion on June 22, 2026. CEO Felipe Chavez sees clear continuity. “We already have a foot in the door with our delivery robots,” he said. “It made sense to also offer manipulation solutions.”
The Hardware and Its Immediate Targets
R-noid combines a mobile base with dual seven-degree-of-freedom arms. Each arm carries up to 11 pounds. A four-degree-of-freedom articulated torso adds reach. The machine includes autonomous docking, generative vision-language-action models and a dual-screen face for simple communication. One version rolls freely through kitchens, factory aisles and logistics zones. A stationary variant handles precision bench work.
Initial roles stay narrow. Restaurant assistant. Packer. Picker. Folder. Host. At Harbor Links Golf Course in New York an R-noid loads food into delivery robots and packs orders. In warehouses it moves semiconductor wafers, boxes and parts. GXO has tested it in live operations, according to the company’s site. Robot.com lists these applications alongside claims the robot works today, not in some distant lab.
But real deployment demands preparation. Customers wait eight to 12 weeks. Teams visit the site, map tasks, collect data — sometimes 50 hours for complex motions — then fine-tune the model. Initial autonomy hovers near 70 percent. Teleoperators and a remote service desk fill gaps. Chavez emphasizes the infrastructure his company already owns. “We have built operations, maintenance, remote operations, remote service desk, data infrastructure, and business development infrastructure,” he said. “We already know what it takes to deploy robots.”
The partnership with Physical Intelligence supplies the intelligence layer. The AI firm develops foundation models that work across robot shapes. Robot.com began collaborating last year. Custom versions now drive R-noid’s ability to interpret commands in messy, unpredictable settings. Physical Intelligence has raised substantial funding from investors including Jeff Bezos and OpenAI, though specific terms of the Robot.com deal remain private.
Competition surrounds the effort. Agility Robotics ships Digit for tote handling and recently signed a commercial deal with Mercado Libre. Apptronik’s Apollo has run trials at Mercedes-Benz. Figure AI demonstrated warehouse sorting that drew millions of views, yet experts note the systems still pause and require supervision. Wheeled humanoids from newer entrants like Sunday Robotics also chase similar hospital and logistics niches. Robot.com argues its delivery pedigree gives it an edge in maintenance, mapping and customer relationships.
Economics drive interest. Labor shortages plague food service, logistics and light manufacturing. Robots that work existing shifts without fatigue promise lower costs over time. Chavez frames the near-term goal differently: prepare workplaces and raise worker satisfaction by removing dull repetition. One motive, he repeats, remains “robots today, not someday.”
Technical limits persist. Payload stays modest. Full autonomy in unstructured spaces still needs remote backup. Data collection takes time and varies by customer layout. Scaling from dozens of units to hundreds will test the remote operations center and model refinement process. Yet the company has already turned delivery trust into infrastructure across 10 countries. That network now expands to manipulation.
Recent coverage reinforces the trend. A June 6, 2026 guide from Robozaps catalogs deployments at Amazon and Mercedes-Benz, noting humanoids’ advantage climbing stairs and using human tools. An April 2026 analysis in Spiceworks projects shipments growing from 13,000 units in 2025 toward millions by 2035. Morgan Stanley forecasts acceleration in the late 2030s once technology, regulation and public acceptance align.
Robot.com positions R-noid as practical rather than general-purpose. It won’t cook meals or fold linens in every home. Instead it targets specific, repeatable tasks inside facilities its delivery fleet already knows. The expressive screen, VLA models and existing service desk aim to make the machine approachable for workers who once only interacted with rolling delivery boxes.
Success will hinge on execution. Can the firm raise autonomy beyond 70 percent quickly? Will customers pay for eight-to-12-week onboarding? Early data from GXO and the golf course offer first signals. Broader adoption in food service and healthcare could follow if labor savings materialize and uptime matches delivery robots.
For now the company keeps the message direct. Its website declares robots for now, not someday. Chavez calls his team disciplined builders focused on scaling proven products. The R-noid rollout tests whether that discipline translates from sidewalks to factory floors. Industry watchers will track the next dozen deployments closely. The shift from delivery to manipulation marks a logical step. Whether it becomes standard workplace equipment depends on results gathered in real facilities, one task at a time.
WebProNews is an iEntry Publication