Amazon Go

Amazon Go was Amazon's chain of checkout-free convenience stores. It opened to the public in Seattle in January 2018, peaked at about 30 stores, and Amazon closed it in 2026.

What it was

Each Amazon Go store used cameras and sensors to work out what shoppers took, and charged them when they walked out.

Shoppers scanned the Amazon Go app at a turnstile to get in, AP reported at the opening. Anything they took off a shelf went into a virtual cart, and anything they put back came out of it.

Amazon called it Just Walk Out shopping. Jeff Bezos wrote in his 2017 letter to shareholders that it ran on "a custom-built combination of computer vision, sensor fusion, and deep learning".

Bloomberg later described the first store as one near Amazon's headquarters, selling salads, sandwiches and snacks to office workers who wanted a meal on the run.

What happened

Amazon Go took more than a year to open to the public after it was announced.

AP reported that Amazon announced the store in December 2016 and said it would open by early 2017, then delayed the debut while it worked on the technology and its employees tested it. Go opened to the public on January 22, 2018.

In September 2018, Bloomberg reported, citing people familiar with the plans, that Amazon was considering as many as 3,000 Go stores by 2021. Amazon declined to comment.

The chain peaked at about 30 stores in early 2023, Retail Dive reported. Bloomberg reported in March 2023 that Amazon was closing eight of them, and three New York stores followed in 2024. By January 2025 Amazon had 16 Go stores, a spokesperson told Retail Dive.

In April 2024, as Amazon was rolling back Just Walk Out in its US grocery stores, Fortune asked Dilip Kumar, an Amazon vice president and one of its inventors, about a report that humans were involved in 700 of every 1,000 Just Walk Out sessions in 2022. Kumar said "that sounds high", that humans review "a small sample of videos", and "We don't share those numbers."

On January 27, 2026, Amazon said it would close all its Go stores and its Amazon Fresh grocery stores. AP counted 15 Go stores still open, Bloomberg 14.

Amazon said the last day would be February 1, AP reported, except in California, where stores would stay open longer to comply with state requirements. The Orange County Register later reported that the remaining California Fresh and Go stores would close on March 13, 2026.

Why it failed

Amazon said it hadn't yet built a distinctive enough store with an economic model that could scale, while reporters pointed to the cost of opening each location, ordinary merchandise and an office-lunch trade that remote work cut.

What Amazon said

In its closure post, Amazon wrote that "we haven't yet created a truly distinctive customer experience with the right economic model needed for large-scale expansion." It was closing the stores "to prioritize investment in growth areas", and planning to open more than 100 new Whole Foods Market stores over the next few years.

Jason Buechel, Amazon's vice president of worldwide grocery and the CEO of Whole Foods, gave staff the same reason in a memo CNBC printed.

The post doesn't blame the Just Walk Out technology. It calls the Go stores "innovation hubs" where Amazon developed it, and says the technology runs in more than 360 third-party locations across five countries.

On the call for the fourth quarter of 2022, CFO Brian Olsavsky said Amazon had decided "to exit certain stores with low growth potential." Grocery Dive reported that he said Fresh and Go accounted for the bulk of a $720 million impairment charge that quarter.

A spokesperson told Bloomberg in 2023 that Amazon remained "committed to the Amazon Go format". Three New York stores "were performing well", Amazon told C-Store Dive in 2024, but "we couldn't make the economics work with the lease cost". In January 2025 a spokesperson told Retail Dive that Amazon would keep investing in its Go stores.

What reporting showed

Bloomberg reported in 2018 that the first store "required more than $1 million in hardware alone", according to a person familiar with the matter, and called the high cost of opening each location the challenge to Amazon's plan. In the same story, NACS vice president Jeff Lenard said that if a store is a quarter mile from where people walk and bike, "the novelty of the technology won't matter."

Bloomberg wrote in 2023 that the format "was a victim of the pandemic", because so many people worked from home and went to the office only a few days a week. Amazon's statements on the 2023 and 2026 closures don't mention the pandemic.

At the closure, Fortune's Phil Wahba wrote that "the dazzling tech was not enough to camouflage how blah the merchandise was." Fortune quoted Neil Saunders, managing director at GlobalData, saying neither Fresh nor Go offered an appealing in-store experience. In emailed comments quoted by Bloomberg, he said neither was "delivering the sales needed to make them fully economic".

What later analysis found

C-Store Dive argued, the day after the announcement, that Amazon built a convenience chain without fuel, which it called "the primary driver of visits to c-stores in the U.S." It said fuel wasn't the only reason, and that it's possible Amazon "never gave customers a reason to visit its convenience stores in the first place."

In an opinion column for Computerworld, Evan Schuman wrote that shoppers "visited Go mostly for the novelty" and didn't find it meaningfully better than their usual stores. He wrote that Amazon didn't drop the technology; it "realized that the value of the technology lay elsewhere."

The stage it reached

Amazon Go never left the introduction stage. I found no published figure for Go's sales, so this call rests on store counts, costs and Amazon's own words.

In the product life cycle, growth is when adoption accelerates and unit costs fall as volume builds. Go never got near the scale Bloomberg's sources described, and no source I read reports its costs falling. I read Amazon's 2026 statement, that it hadn't yet found the right economic model needed for large-scale expansion, as an introduction-stage verdict.

It wasn't a decline either. Decline is when sales fall because the market has moved on to a substitute. Amazon's closure post says US shoppers are "turning to online delivery", but that's a claim about shoppers in general, not evidence about Go's customers.

The lesson

I think Go shows that a better way to pay isn't a reason to visit.

Bezos said in 2018 that Amazon was "very interested" in physical stores, but only with something new to offer. "If we offer a me-too product, it's not going to work," he said, Bloomberg reported. Go's checkout was new. By Fortune's account, its shelves weren't.

Go looks to me like a store built outward from its technology instead of from the shopper, the first pattern in why AI projects fail.

If you're building a product around a clever model, count repeat visits apart from first visits, and work out what one location costs and earns before you plan the next ten.