Avatar Robotics Raises $6.5M for VR-Piloted Warehouse Robots

A seed round built on remote control

Avatar Robotics, a San Francisco startup, has closed a $6.5 million seed round. Its plan for humanoid robots runs against the usual pitch: it does not wait for machines to work on their own. Remote human operators run its robots in warehouses today. AlleyCorp led the round, defy.vc led an earlier pre-seed, and Headline, Henry Ford III, Refashiond, Paul Vogel, Jack Huffard and Samuel Udotong also invested.1516 The company describes its goal as an "unlimited workforce" for industrial labor. In practice, that means robots paired with a global pool of people who control them, plus AI meant to take over more of the work over time.15

The system works like this. Operators wear Meta Quest headsets and hold handheld controllers, and they guide wheeled robots with two arms through packing and sorting jobs.11 Business Insider's profile describes a pilot in Dallas shortly before Christmas 2025. One robot packed beauty products into pouches for a full eight-hour shift while its operator worked from an Airbnb a few miles away.11 Since launching in December 2025, Avatar says its robots have packed, sorted or helped ship more than 900,000 products. Customers include what it calls a leading global beauty retailer, and the company says a multi-billion-dollar warehouse operator has expanded its use beyond a pilot.1519

The business model: cheap remote labor, and the data it produces

Avatar does not sell robots. Customers pay a recurring fee, which founder and CEO Colin Webb says costs less than hiring warehouse workers directly.11 The savings come from where the operators live. Many are contractors in Mexico, Colombia and the Philippines who earn between $5 and $20 an hour depending on location.11 Customers know the robots are remote workers in robotic bodies, according to Webb.11

The second source of value matters more to investors. Each teleoperation session records what the robot sees and how the operator responds in real time. Avatar uses that data to adapt open-source robotics models so the machines can gradually handle more tasks without help.12 AlleyCorp principal Brannon Jones said the human-in-the-loop system performs about as well as skilled human operators. He added that it is building large datasets of robot-control data, something he says the industry lacks.15 Defy.vc partner Amy Yin said full autonomy will need both better hardware and huge amounts of real-world data. She thinks the approach could produce one of the world's largest humanoid fleets.1519

This is the most interesting part of the deal. Robotics companies are competing for the real-world data needed to train the "brains" of physical AI. Startups such as microagi, Turing and micro1 pay people to film themselves cooking or cleaning.12 Webb's argument is that paying people to do useful work produces better value than paying them only to wear sensors. His logic is that the work earns money now and the data comes as a byproduct.11 Mecka AI takes the opposite approach. It puts sensors and instrumented gloves on humans instead of having them control robots, and its CEO says it has passed $100 million in annualized revenue.17 Both approaches have backing, so neither has clearly won. Avatar's version does have one structural advantage: customers pay for the labor that produces the training data.

Where the coverage diverges

The reporting differs in a few places. The company's announcement gives the output as "more than 900,000 products."1519 Business Insider rounds that to nearly 1 million items. It also says the robots work across several warehouse facilities, yet reports that Avatar would not say how many robots it has deployed, how many customers pay, or how many facilities it serves.1112 The location of the operators also varies. Business Insider's main profile names Mexico, Colombia and the Philippines.11 Its later list of promising startups mentions only Mexico and the Philippines.17

The framing differs too. The press release talks about safer, more comfortable work and a remedy for the US labor shortage.15 Business Insider spends more time on the wage gap between US warehouse jobs and offshore operators, and on the physical strain of the job. Teleoperations supervisor Angel Avila said eight-hour shifts can cause arm and shoulder cramps, and the office keeps a massage gun.1112 Webb calls the job "like playing a video game for eight hours." The Business Insider reporter who tried it managed to place only two packets in her first few minutes.11 The press release presents a clean story about infrastructure. The hands-on reporting shows a job that is real, tiring and paid at offshore rates.

The supply chain problem

Avatar buys wheeled humanoids from China instead of building its own. That sets it apart from Agility Robotics, Figure AI and Tesla.11 New Federal Communications Commission restrictions on foreign-made "advanced robotic devices" could cut off that supply and push Avatar toward assembling robots in the US.11 Webb says the business would be unaffected because Avatar's software runs on many types of robots. He calls the rules a chance to move toward his longer-term goal of building Avatar's own fleet.11

That answer is plausible but not proven. Being able to run on different hardware is a real advantage for a teleoperation company. Still, US-built hardware would almost certainly cost more than the Chinese robots Avatar now uses, and lower costs than human labor are the core of its pitch. That is my assessment, since the reporting does not give hardware costs. Other startups already treat supply-chain origin as something to sell. Minerva Humanoids, which also uses VR teleoperation, says it will build robots in Germany and the US without relying on single-source Chinese components.9

Teleoperation becomes a funding category

Avatar is one of several companies betting on remote control. Two months after its round, Minerva Humanoids came out of stealth with about $10 million in pre-seed funding led by General Catalyst. Its robot, Roger, is semi-autonomous and is controlled by a specialist through a VR headset.710 Minerva aims at oil rigs, bomb disposal, hazardous-materials response and airport security, so its pitch is about danger, not labor shortages.310 Robotics and Automation News reports that Roger also uses force-feedback haptic gloves.9 One trade outlet noted that Minerva has published no specifications, latency figures or named customers.3 In the UK, Extend Robotics raised £2.6 million to expand a model in which customers pay for completed robotic work instead of buying robots. It is also building a network of remote operators.6

Taken together, these deals point to a clear pattern. Investors increasingly see remote human control as a bridge to autonomy, not an admission of failure. Avatar's model is the most commercial version: it sells warehouse labor right away and treats autonomy as a gain that comes later.1519

How small the round is

The size of the round is notable. A tracker of pure-play humanoid funding from August 2025 to July 2026 counted $6.2 billion across 25 disclosed deals, with a median round of $145 million.1 Seed rounds made up only 0.74% of that capital, and no disclosed rounds fell below $5 million.1 Figure's $1 billion Series C and NEURA Robotics' $1.4 billion Series C account for much of the total.1

Against those numbers, $6.5 million is very little. That may be the point of Avatar's approach. Companies building humanoids from scratch have to raise huge sums before their machines can earn money. Avatar earns revenue from shift one because people supply the intelligence the robots lack.1115 Investors seem to agree. Business Insider included Avatar in its inaugural list of 25 promising robotics startups, nominated by defy.vc's Amy Yin, who is also one of its investors.17

My read

The "side hustle" angle in the coverage overstates things for now. Most Avatar operators work from dedicated facilities. Working from home is something Webb hopes for, not how the company runs today.11 The real story is about economics and data. Avatar sells warehouse labor priced on offshore wages through robots made in China, and it collects training data as it goes.1112 Its growth targets are ambitious. Webb wants each customer to reach dozens of robots this year and hundreds next year.11

The obvious objection is whether operators are training themselves out of a job. Webb says they are not. In his view, humans will supervise larger fleets as the robots gain autonomy, and new industries will create jobs nobody can picture yet.11 That is a hope, not a plan. The data flywheel only pays off if each operator handles fewer of the tasks over time. If Avatar succeeds, its contractors will become supervisors or they will become unnecessary. The funding announcement does not say which. What is already clear is that teleoperated humanoids have moved from lab demos to paying warehouse shifts. The company says nearly a million packed items show the model works, while it has not disclosed how many robots or customers produced that number.