- Published on
The Robot Economy Is Here: Who's Actually Making Money in 2026
- Authors
- Name
- Mikhail Liublin
- https://x.com/mlcka3i
A robot dog named Royal Barkla is sniffing whisky barrels in Scotland, hunting for leaks that cost Bacardi millions. A fleet of drones is delivering baby wipes and eggs to homes across Dallas in under 19 minutes. A $50,000 humanoid the size of a nine-year-old is checking what's in the fridge at a startup office in San Francisco.
None of this is a prototype. All of it is generating revenue in 2026. And together, these stories tell you more about where the robot economy is headed than any CES keynote.
The Humanoid Business Model Problem
Everyone is building humanoid robots. The harder question is: who's making money?
At CES in January, Google DeepMind and Boston Dynamics announced a partnership to deploy Gemini-powered Atlas robots at Hyundai auto plants. The pitch: give humanoids the contextual awareness to navigate unfamiliar environments and manipulate objects with their hands. Not choreographed demos—actual factory work.
"The real value going forward is for our robots to be contextually aware of their environment and able to use their hands to manipulate any object," Boston Dynamics CEO Robert Playter told WIRED. "Manufacturing environments, like in Hyundai factories, are a perfect place to deploy that today."
Google DeepMind CEO Demis Hassabis sees Gemini as Android for robots—a universal AI brain that can run on anyone's hardware. The business logic is familiar: own the platform layer, let others build the bodies. If it works, DeepMind becomes the operating system for physical AI, collecting a margin on every robot that ships.
The problem is that most humanoids still can't do useful work reliably. At the World AI Conference in Shanghai last July, humanoids danced, boxed, and backflipped—but look closer and you'd spot humans holding game controllers, telling the robots which way to walk. Many humanoids don't even have functional hands. Their arms end in stumps.
Still, the investment is enormous. Over 200 Chinese companies are now developing humanoid robots, according to CMRA, a Chinese industry association—so many that the Chinese government has warned of overcapacity. The US has about 16 prominent firms. Morgan Stanley forecasts a billion humanoids in use by 2050, with nearly a third in China.
The company leading on volume is Unitree, based in Hangzhou. Unitree sold nearly 24,000 quadruped robots in 2023—ten times the volume of Boston Dynamics. Its humanoid, the G1, costs tens of thousands of dollars. An American equivalent often costs hundreds of thousands. Unitree is reportedly targeting a $7 billion IPO in Shanghai.
The playbook is pure Chinese hardware economics: build cheaper, iterate faster, sell more. Unitree's CEO, Wang Xingxing, built his first robot on a budget of $28. The company's headquarters is reportedly battered from robots practicing on the stairs. "Everywhere there are broken robots, broken parts, and broken concrete," one visitor told WIRED.
Where the Early Revenue Is
The companies finding actual customers right now are not selling general-purpose humanoids. They're selling narrow, specific machines.
Fauna launched Sprout, a $50,000 humanoid the size of a nine-year-old, designed for hospitality, research labs, and entertainment—not warehouse work. Its first customers include Disney and, interestingly, Boston Dynamics itself. Sprout has mechanical eyebrows to express surprise and confusion. It can walk to a fridge, peer inside, and report back what's there. The bet: the first real market for humanoids is service and entertainment, not manufacturing.
Amazon shelved its Blue Jay warehouse robot system in February 2026, just months after launch, pivoting toward micro-fulfillment centers in the back of Whole Foods stores. Even the biggest player in logistics is still searching for the right robot form factor.
Qualcomm acquired Arduino and announced the Ventuno Q, a single-board computer with a 40 TOPS NPU designed specifically for building autonomous robots. This is an infrastructure bet: sell the silicon to everyone building robots, regardless of who wins the humanoid race.
Honor, the Chinese smartphone company, announced a pivot to humanoid service robots at MWC in February 2026—a signal that consumer electronics companies see robotics as the next hardware frontier after smartphones plateau.
The pattern: the money is in components, platforms, and narrow applications. General-purpose humanoids remain largely a capital expenditure, not a revenue generator.
Drones: The Delivery Economics Are Finally Getting Tested
Drone delivery has been "almost here" for a decade. In 2026, the numbers are finally being forced into the open.
Walmart and Wing Go Big
Alphabet's Wing has been running drone deliveries for Walmart in Dallas for nearly two years. The operation: 18 stores, 18 drones each, roughly 1,000 orders per day. Average delivery time: 19 minutes. Top items: baby wipes, eggs, a forgotten pint of milk.
In June 2025, they announced expansion to 100 additional stores in Atlanta, Charlotte, Houston, Orlando, and Tampa. The FAA authorized Wing to make up to 30,000 deliveries per day in Dallas alone. Walmart+ members (20 per shipment.
Wing CEO Adam Woodworth declined to comment on the business model, but said Wing envisions taking a delivery fee cut on each transaction. The question nobody has answered publicly: is this profitable at scale, or is it still subsidized by Alphabet's balance sheet?
MIT researcher Matthias Winkenbach is skeptical. "It's unlikely that it will become commercially viable in the foreseeable future," he told WIRED, citing regulatory costs, the expense of employing drone pilots, and the challenge of working with unpredictable customers and environments. "It's hard to beat the efficiency and price of a good old UPS truck."
The drones also don't fly in bad weather. The expansion is exclusively to warm, temperate cities. Cold batteries need warming. Wind and rain ground flights. This is a fair-weather business, literally.
DoorDash Bets on Robots on Wheels
DoorDash launched Dot, a 350-pound, 5-foot-tall delivery robot, in metro Phoenix in late 2025. It travels up to 20 mph on roads, bike lanes, and sidewalks, carries 30 pounds of cargo, and runs six hours per charge.
DoorDash is pursuing a "hybrid" model: human Dashers, drones (via its Wing partnership), and autonomous robots all dispatched through a single platform. The pitch to investors is obvious—replace expensive, hard-to-manage human contractors with machines.
But delivery robots have a brutal track record. Amazon and FedEx both abandoned their programs. Starship Technologies, the biggest survivor, is mostly confined to university campuses. Serve Robotics, a Postmates spinoff, has built only 400 robots after years of operation.
The core problem, as Urban Robotics Foundation director Bern Grush put it: "You're trying to solve a much harder problem with far, far, far less capital and far, far, far less compute." A delivery robot has to navigate sidewalks, bike lanes, parking lots, driveways, and apartment complexes—understanding cars, strollers, children, dogs, squirrels, and aggressive mopeds. Robotaxis at least stay on roads.
The Weird Niches That Actually Work
The most commercially successful drone and robot deployments aren't glamorous. They're industrial.
Bacardi's whisky-sniffing robot dog (a modified Boston Dynamics Spot) patrols Dewar's warehouses in Scotland, using an ethanol sensor to detect leaky barrels. The robot costs under $100,000. In its first deployment, it flagged 10% of casks as needing remediation. Given that Dewar's has 100+ warehouses with 25,000 casks each, aging for up to 12 years, the ROI is immediate and obvious.
Sewer inspection drones from Swiss company Flyability and rival Flybotix are replacing human inspectors in underground pipes. Macomb County, Michigan, spent 1 million in manual inspections every three years—a 40% cost reduction with better defect detection. Flyability reports "huge sales growth." The value proposition is simple: robots go where humans shouldn't, for less money, with better data.
Wing delivers blood between hospitals in London for the UK National Health Service. Medical logistics—time-sensitive, high-value, low-weight—may be the drone delivery use case that makes economic sense before consumer groceries ever do.
These deployments share a common trait: they solve a specific, expensive problem in a controlled environment. No toddlers. No squirrels. No bad weather (or at least, predictable bad weather).
The Platform War
The real business story of 2026 robotics is not which robot wins. It's who owns the brain.
Google DeepMind is positioning Gemini as the Android of robotics—a universal AI model that runs on any manufacturer's hardware. Anthropic demonstrated Claude controlling a robot dog. Every major AI lab is eyeing robotics as the next platform.
The logic: if AI models need to understand the physical world to achieve general intelligence, then robotics data is strategically essential. The lab that collects the most real-world interaction data—through robots in factories, warehouses, hospitals, and homes—will have a lasting advantage.
This explains why DeepMind hired the former CTO of Boston Dynamics. It explains why OpenAI is reportedly developing humanoids. And it explains why Chinese companies like Agibot are training robots using teams of human workers who demonstrate tasks the robots then learn to replicate.
The platform play also explains why Qualcomm bought Arduino. If you can't own the AI model, own the chip. If you can't own the chip, own the development board. Every layer of the stack is a potential toll booth.
What the Numbers Say
Some concrete figures worth holding onto:
- $50,000: Price of Fauna's Sprout humanoid (hospitality/research)
- 20,000: Price of a fully equipped Unitree humanoid (China)
- $75,000+: Price of a Boston Dynamics Spot (industrial)
- $100,000+: Typical US humanoid price point
- $7 billion: Unitree's reported IPO target
- 200+: Chinese companies building humanoids
- ~16: US companies building humanoids
- 24,000: Unitree quadrupeds sold in 2023 (10× Boston Dynamics)
- 1,000/day: Wing drone deliveries in Dallas
- 30,000/day: FAA-authorized Wing capacity in Dallas
- 19 minutes: Average Wing delivery time
- $100,000: Bacardi's whisky robot dog cost
- 40%: Cost reduction from drone sewer inspections vs. manual
- 1 billion: Morgan Stanley's forecast for humanoids in use by 2050
What This Means for Business
Three takeaways:
1. Narrow beats general. The robots making money today solve one problem well: sniffing barrels, inspecting sewers, delivering eggs in warm weather. General-purpose humanoids remain a capital sink. If you're evaluating robotics investments, look for specificity and a clear cost advantage over the human alternative.
2. China will win on volume. The US will try to win on intelligence. Unitree sells ten times the robots Boston Dynamics does, at a fraction of the price. The US response is to compete on AI—Gemini, Claude, proprietary models that make robots smarter per unit. Whether intelligence can offset a 10× cost disadvantage is the central question of this industry.
3. The platform layer is the biggest prize. The company that becomes "Android for robots"—providing the AI brain that runs on any hardware—will capture more value than any single robot manufacturer. Google, OpenAI, and Anthropic are all positioning for this. The robotics hardware companies know it, which is why Boston Dynamics is partnering with DeepMind rather than building its own foundation model.
The robot economy in 2026 is real, but it's smaller and weirder than the hype suggests. The money is in whisky barrels and sewer pipes, not dancing humanoids. The race that matters is not who builds the best robot body—it's who builds the brain that runs on all of them.