Jeff Bezos made a bold bet in 2017. Convinced that consumers would soon have a robot at home, he told a group of Amazon's top device executives to go build one. Within a few years they produced the $1,450 Astro — essentially an Alexa-equipped tablet on wheels that navigates a house on its own, recognises faces and works as a roaming security camera. It was unveiled in autumn 2021, shipped to a few thousand customers on an invite-only basis, and never reached open sale. A business version followed in 2023 and was sunset less than a year later.

Amazon is not giving up on Bezos's dream. According to CNBC's analysis of job postings, the company has spent recent months assembling a new team working on a "new-to-world robot product". One listing, since removed, promised: "Robots are about to step out of the movies and into our homes." It described a product that navigates "dynamic home environments" and operates "autonomously alongside family members".

What is being built

In March, Amazon acquired New York-based Fauna Robotics, which built a 3.5-foot humanoid called Sprout — equipped with expressive yellow eyebrows and so far used mainly by developers, commercial and academic researchers. Fauna was to become part of a new division called the Personal Robotics Group.

At the same time Amazon redeployed much of the Astro team and brought in an MIT roboticist, according to LinkedIn profiles. Some group staff came from Lab126, the R&D unit behind the Kindle, Echo and Fire TV — and behind the Fire Phone, one of the company's biggest flops.

The listings hint at the direction: a field support engineer for customer support and on-site repair; a UX designer to "define what it feels like to live with a Fauna robot". Another posting states the goal plainly: "make robots people actually want to live and interact with in everyday human spaces."

An Amazon spokesperson told CNBC the company is "committed to tackling the challenges that can unlock the personal robotics category", while acknowledging a substantial gap between what works in a controlled lab and what works in a living room or a lobby.

Why Astro failed

Ken Washington, Amazon's vice president of consumer robotics until 2023, told CNBC he pushed for a bipedal robot and leadership disagreed. Watching Bezos at show-and-tells, he 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."

A former employee put it more bluntly: "Our poor little Astro was around $2,000 when it came out, and that's a lot. That's a lot for a consumer, and they want a robot that does a lot." An owner in North Carolina who has lived with Astro since 2023 still uses it to check on pets and appliances when she is out, but says it has become "a party trick and a trip hazard". Her plan is to let it run until it dies — "then send it to the trash heap".

Why it might go differently this time

Colin Angle, co-founder and former CEO of iRobot, is blunt about the earlier effort's tools: "The toolkit that the Astro team had when they did it was clubs with nails in them, just incredibly primitive." It is, he says, difficult to understate how much the robot industry has changed in two years. Advances in AI models, sensors, batteries and actuators have driven a new wave.

The money agrees. Barclays estimated in January that the global humanoid market, currently worth $2–3 billion, could reach roughly $40 billion by 2035, or as much as $200 billion under the most optimistic scenarios. Venture firms put $33.4 billion into robotics and physical-AI companies in the first half of 2026, against $28.7 billion for all of 2025, per PitchBook. Agility Robotics — maker of the bipedal Digit, with Amazon as an investor — went public via a SPAC on Thursday.

The home remains the hardest market of all. "Children, pets, grandparents, stairways. Nobody's home is set up exactly the same," said Jeff Burnstein of the Association for Advancing Automation. Amazon has deep robotics form: it bought Kiva Systems for $775 million in 2012 and now runs more than a million robots across its warehouses. Those were built for standardised halls, not living rooms.

What to watch

- Whether Amazon positions price and capability differently this time — Sprout is a developer product, not a consumer device. - Whether the group shows an arm, manipulation and specific household tasks, rather than navigation and a security camera. - Whether Astro owners keep getting support; in their Facebook groups, complaints about diminishing service and the absence of the Alexa+ assistant are mounting.