In late January 2026, by the trade press's dating of the announcements, Amazon published two statements about the same technology, days apart, pointing in opposite directions.
The first said Amazon is closing every Amazon Go and Amazon Fresh location. The company's stated reason, in its own words: "we haven't yet created a truly distinctive customer experience with the right economic model needed for large-scale expansion." Go was the store Just Walk Out was invented for, the one where the 2018 launch video was filmed. All of them are closing.
The second statement, published the same week under the title "An update on Amazon's plans for Just Walk Out and checkout-free technology," said this: "We have strong conviction that Just Walk Out technology will be the future in stores that have a curated selection where customers can pop in, grab the small number of items they need, and simply walk out."
Read that description again. Curated selection. Pop in. Grab a few items. Walk out. That is a description of Amazon Go. Amazon asserted the future of the format in the same week it exited the format entirely.
The tempting frame is hypocrisy, and it is wrong. The two statements reconcile cleanly, and the reconciliation is more useful than the gotcha: the technology works well enough to sell to other people, and the store it was invented to prove did not work as a business. Those are different claims. The 2018 launch made only the second one, and only the first one survived.
Just Walk Out did not die in January. It moved. It is now a product Amazon licenses to other retailers, and by Amazon's own product site, "more than 175 third-party retailers are leveraging Just Walk Out technology at locations across the U.S., UK, Australia, Canada and France." Note the noun: retailers, not locations. One retailer can run many locations, and Amazon's own phrasing is the count worth quoting.
Meanwhile the stores are gone. By the tallies the trade press assembled at the announcement, 72 US locations, 57 Amazon Fresh and 15 Amazon Go, most closing by early February, plus 19 UK locations, some slated to become Whole Foods. And Amazon is not leaving grocery. The same closure post promises more than 100 new Whole Foods Market stores over the next few years. Amazon is leaving the grocery format it invented a technology to power, while expanding the grocery chain it bought, which runs checkouts.
The Just Walk Out update was not written by anyone in Amazon's retail organization. It is authored by a Vice President of AWS Applications. That is the whole story in a job title: checkout-free shopping is no longer a bet about how Amazon runs stores. It is a cloud product line, run by the division that sells infrastructure to other companies, and the post reads exactly like what it is, a vendor's roadmap for a technology with reference customers.
Look at where the vendor says the product is going. Amazon's own list of Just Walk Out venue types: convenience stores, sports and entertainment venues, airports, healthcare, higher education, hospitality, business and industry. Stadium beer stands and hospital gift shops and campus grab-and-go. It is a good list, and one venue type is conspicuously not on it: a supermarket. For the big weekly trip, Amazon's own update says the quiet part plainly: "customers so far prefer Amazon Dash Cart," the smart cart that shows a running total, which is to say, a shopping cart with a screen, a much less magical machine that customers apparently trust more with a hundred dollars of groceries.
The 2018 demo had two pieces of magic. The store knew what you took, and the store knew who you were. The second piece, at its most theatrical, was Amazon One, the palm scan: hold your hand over a reader and the building recognizes you. It made the demo feel like the future in a way cameras never could, because it gave the future a gesture.
The palm is being retired. By the reporting in the trade press, Amazon is discontinuing Amazon One for limited customer adoption, removing the readers, and deleting enrolled palm data, with the service reportedly continuing only for patient check-in at existing healthcare sites. The same reporting put its usage in the prior year at 36.7 million items across 17.7 million shopping sessions, numbers that sound large until you remember they describe the flagship identity gesture of a company with hundreds of millions of customers. Those details come from the outlets, not from an Amazon statement I can point to, and they deserve that attribution. But the direction is not in dispute: walk-in-and-walk-out survives as a licensed capability, and the gesture that sold it does not.
There is one more honest change worth naming without a gotcha. Amazon's current product description says Just Walk Out "combines AI, computer vision, sensor fusion, and RFID." RFID was not part of the 2018 story; tagged merchandise is the technology the launch was implicitly positioned against. Its arrival is normal engineering, the kind of pragmatic widening every shipping product undergoes. It also means the thing being licensed to more than 175 retailers is not quite the thing that was demonstrated eight years ago. The demo was a sensor-and-vision moonshot. The product is a toolkit that uses whatever works.
We have written before about hyped systems whose afterlife diverged from the launch, and this one completes a pattern worth stating generally. When we covered Devin, the "first AI software engineer", the afterlife story was underperformance: the thing did not do what the demo implied. AutoGPT's hundred thousand stars were a third shape again, enthusiasm that never converted into a buyer at all. Just Walk Out is the opposite pole from Devin and the more common one: the thing works, and it relocated. Nothing in the January announcements says the computer vision failed. The store around it failed, as a business, at exactly the layer no launch video ever demonstrates: rent, shrink, staffing, basket sizes, the economics of selling sandwiches.
So the question an operator should carry out of this story is not "did the technology work." It is: which layer had a buyer. Amazon's checkout-free bet had three layers. A sensing technology, an identity gesture, and a retail format. The sensing technology found a buyer, more than 175 of them by Amazon's count, and it survives with a vendor's roadmap and a VP whose title says cloud. The identity gesture found a widely circulated session figure whose provenance our companion piece could not establish, and whatever it counted, it was not enough, and the gesture is being unplugged. The retail format found no economic model by its own builder's admission, and it closes in February.
Those same three layers sit inside every ambitious internal technology, and they do not survive together. The one that lasts is usually the one somebody outside the building would pay for, which is usually not the layer in the launch video, because launch videos are made of interface and story.
Amazon told us both halves in one week, in its own words. The future of stores, and the closing of every store it built that future for. Both true. You just have to notice which organization signed each statement.