For years, the household robot has been the punchline of tech promises. Every CES demo showed a machine that could pour a glass of water on a stage but would fail spectacularly in an actual kitchen. That might finally be changing, and the shift is coming from a garage in Mountain View, not a corporate R&D lab.
Tony Zhao and Cheng Chi, two Stanford Ph.D. roboticists who helped build some of the most influential open-source robotics research of the past five years, dropped out to start Sunday Robotics. Their creation, a household humanoid called Memo, just pushed the company past a $1.15 billion valuation after raising $165 million in a Series B led by Coatue Management.
The Research That Made Memo Possible
Zhao and Chi aren’t first-time builders chasing a trend. At Stanford, they co-created ALOHA, a low-cost open-source hardware system developed in collaboration with Google DeepMind that used imitation learning to train robots on complex manipulation tasks. They also developed Diffusion Policy and UMI, two frameworks that changed how the robotics community thinks about teaching machines to move like humans.
The core insight behind their research was simple but powerful: robots don’t need millions of dollars in hardware to learn useful skills. They need better data and smarter ways to collect it. That philosophy drove everything Sunday built next.
A $200 Glove That Teaches Robots to Do Chores
Sunday’s secret weapon isn’t Memo itself. It’s the Skill Capture Glove, a $200 wearable device that lets anyone record training data for household tasks without needing a robot in the room. You put on the glove, fold a towel, load a dishwasher, or clear a table, and the device captures your hand movements, contact forces, and haptic dynamics with enough fidelity that the data maps directly onto Memo’s embodiment.
This approach solves the biggest bottleneck in robotics: data collection. As Zhao put it, “Data has always been the biggest bottleneck in robotics. We built the only pipeline that turns the complexity of real-world homes into autonomous intelligence.” Sunday has already generated tens of millions of movement episodes using this method, building a training dataset that no competitor can match.
What Memo Actually Does
Memo stands 1.7 meters tall and weighs 77 kilograms. It rolls on a wheeled base rather than walking on legs, a deliberate design choice that prioritizes reliability over spectacle. Two multi-degree-of-freedom arms with dual grippers give it the dexterity to handle real household objects. A central column raises and lowers the upper body so the arms can reach from floor level up to about 2.1 meters.
The tasks Memo handles aren’t glamorous, but they’re exactly what people actually want help with: folding laundry, clearing tables, loading dishwashers. These are the chores that eat 20 to 30 minutes of your day, every day, for decades. Sunday trained Memo on these specific tasks because they’re repetitive, high-frequency, and genuinely time-consuming.
A Race Worth Watching
Sunday isn’t the only company chasing the household humanoid market. Apptronik raised $520 million at a $5 billion valuation in February 2026, though its Apollo robot targets industrial applications like assembly line work rather than home chores. 1X opened pre-orders for its NEO robot with confirmed 2026 home deliveries. Figure launched its 03 model as a “general-purpose home humanoid” that TIME named one of the Best Inventions of 2025. And Tesla’s Optimus Gen 3 began mass production in January 2026, with Elon Musk projecting consumer availability by end of 2027.
The humanoid robot market was valued at $2.92 billion in 2025 and is projected to reach $15.26 billion by 2030. That growth rate explains why investors are pouring money into the sector despite the technical challenges.
What Makes Sunday Different
Most humanoid robot companies are building general-purpose machines and hoping use cases emerge. Sunday took the opposite approach. They picked the most common, most annoying household tasks and worked backward to build a robot that handles them reliably. Wheels instead of legs. Glove-based data collection instead of simulation. Chores instead of choreography.
The company tripled its engineering team and quadrupled its research staff since emerging from stealth in November 2025, growing to more than 70 engineers and researchers. Sunday plans to increase its real-world data collection fivefold by year-end, which would widen the gap between its training data and everyone else’s.
What Happens Next
Sunday is now accepting applications for its Founding Family beta program, where roughly 50 households will receive individually numbered Memo units in late 2026, with a target launch around Thanksgiving. The company has received thousands of applications and over 1,000 waitlist signups. Investors backing the beta push include Tiger Global, Benchmark, Bain Capital Ventures, and Fidelity Management.
For founders watching this space, Sunday represents a pattern worth studying. Zhao and Chi didn’t try to build the most impressive demo. They built the best data pipeline, picked tasks that real customers would pay for, and let the product follow the data. Whether Memo actually works in your kitchen by Thanksgiving remains to be seen. But for the first time in the household robot race, someone is asking the right question: not “what can the robot do?” but “what do people actually need done?”
