Amazon's new home robot starts with a $50,000 research kit

Amazon is building another home robot, five years after the first one quietly stopped being a product. CNBC reported this morning that the company has assembled a new team hiring for a "new-to-world robot product," with job posts promising a machine that navigates "dynamic home environments" and runs "autonomous operation alongside family members." There is no name, no price, and no launch date. There is, however, a $50,000 robot already on sale with Amazon's name on the company that makes it.

That robot is Sprout, and it is why this attempt looks different — not because the hardware is better, but because Amazon has stopped pretending it knows what a home robot is for.

What Amazon actually bought

Fauna Robotics spun out of New York in 2024, founded by Josh Merel, a former Meta and Google DeepMind researcher, and Rob Cochran, who ran product at CTRL-Labs before Meta bought it. Sprout shipped in January 2026: a 3.5-foot humanoid with 29 degrees of freedom, arms that grip, a rubber-like skin to soften accidental contact, and a list price of $50,000 for the Creator Edition aimed at universities, labs and corporate R&D groups, as Silicon Republic reported. Amazon bought the company in March, less than a week after it bought RIVR, a maker of quadruped delivery robots for doorsteps. Both now sit inside a new Personal Robotics Group, The Robot Report reported, with Fauna's roughly 50 staff moving over. The price was never disclosed.

Fauna's own framing matters here. Its announcement promised to keep selling Creator Edition units and supporting existing customers, and described Sprout as a platform, not an appliance: researchers build the applications. That is the opposite of Amazon's consumer hardware playbook — a finished product for tens of millions of people, sold near cost, with the margin made back on services. A $50,000 robot that does nothing until a graduate student programs it is a research instrument, and that is exactly why the deal is worth reading closely.

Astro failed at the one thing robots are for

The received version of the Astro story is that Amazon was early. Jeff Bezos told device executives in 2017 to build a household robot, and the result was a $1,450 tablet on wheels that recognised faces and patrolled the house. It shipped to "a few thousand" invite-only customers, and the business version launched in 2023 was discontinued within a year. Owners CNBC spoke to describe a party trick and a trip hazard; one said she would run hers "until it eventually dies, then send it to the trash heap."

The more useful diagnosis came from Ken Washington, who ran consumer robotics at Amazon until 2023. Executives had discussed adding an arm to pick things up. He wanted legs and was overruled. When Bezos sat in on reviews, Washington said, "I think he was disappointed with the lack of functionality, but I think he was impressed with its ability to map and navigate an undefined space under all lighting conditions."

Read that again. The hard part worked: mapping a messy home in bad light is a genuine achievement, and Astro did it in 2021. What failed was the value proposition. A robot that can see your kitchen but cannot touch anything in it competes with a $30 security camera and loses on price. Another owner told CNBC he would buy a second robot tomorrow "if it could bring a drink to me or dust furniture." Nobody wanted to look at their house. They wanted a machine that did something with its hands.

Sprout is the answer to that, in the most literal sense: arms, 29 degrees of freedom, and a body designed for spaces shared with people rather than fenced off from them. Amazon isn't buying a better screen. It's buying manipulation — the capability Astro lacked, and the most expensive one to build.

The price ceiling is the whole business

Here is the problem that no acquisition fixes. Sprout costs $50,000. A home robot needs to cost roughly what a laptop costs, and the consumer who spent $1,450 on Astro and felt burned is not the customer for a $5,000 humanoid either, let alone a $50,000 one.

That gap is why the research-first path is the sane one. Amazon can sell a few hundred units a year to labs, collect deployment data it cannot get from a warehouse, and let other people discover which tasks are worth automating before it commits a factory line. It is the warehouse business running in reverse: Amazon deploys more than a million robots because it controls the building, the lighting, the floor plan and the job. In a home it controls none of that. Jeff Burnstein of the Association for Advancing Automation put the constraint plainly to CNBC: "Children, pets, grandparents, stairways. Nobody's home is set up exactly the same."

The market everyone is circling is real but not imminent. Barclays values humanoid robotics at $2 billion to $3 billion today and projects roughly $40 billion by 2035 — a decade-long curve, not a product cycle, even though PitchBook counted $33.4 billion into robotics and physical AI in the first half of 2026 alone. Colin Angle, who built the Roomba and watched Amazon walk away from buying iRobot in January 2024, told CNBC the toolkit has changed out of all recognition: "The toolkit that the Astro team had when they did it was clubs with nails in them, just incredibly primitive."

What the sceptics have right

The honest read of this week's news is that it is a hiring signal, not a product. "New-to-world robot product" is a phrase from a job listing that has since been removed, and the surrounding language — robots "stepping out of the movies and into our homes" — is the kind of thing that gets written when the roadmap is a slide. Amazon declined to describe the device, its tasks, or its timeline. The team's lineage includes Lab126, which built the Kindle and the Echo and also the Fire Phone, a device that shipped with total confidence and died in months.

There is a second, less flattering reading of the Fauna deal: Amazon needed a vision-language-action team faster than it could hire one, and "home robot" is the framing that gets an acquisition signed.

The wider market has already done some of this repricing. Unitree's IPO was oversubscribed thousands of times over, then lost half its value as robot hype met its GPT-2-era reality. Agility's public listing showed what a real humanoid business looks like from the inside: a $465 million raise against $1.8 million of 2025 revenue. Amazon can absorb years of that math in a way a startup cannot, which is the one genuine advantage it brings.

What to watch

Three signals will tell you whether this is a product programme or an option being kept alive. First, whether Sprout stays a $50,000 developer kit: the moment it gets a consumer SKU name and a price under $3,000, the strategy has changed. Second, whether Alexa+ reaches the robot — Astro owners have complained for a year that Amazon's newest assistant never came to the device they paid for. Third, Apple. The Information has reported a home robot project including a tabletop device with an arm that resembles the Pixar lamp, and a fixed tabletop that reaches things is a cheaper answer to "bring me a drink" than a walking humanoid. If Apple gets there first at a home price, Amazon will have spent nine years and two acquisitions learning that the legs were never the point.

Would you pay $1,500 for a robot that does exactly one chore well, and which chore? Write to us via our contact page.