Amazon closed every Amazon Go and Amazon Fresh location by February 2026 — the most visible concession yet that autonomous retail is far harder than Silicon Valley promised. But the technology didn’t die with the stores. More than 300 Just Walk Out deployments now operate across stadiums, airports, hospitals, and corporate campuses worldwide, with 150 added in the past year alone. AiFi cut deployment time from six months to under a week. And the autonomous checkout market, valued at $5.5 billion in 2024, is projected to reach $12.2 billion by 2032. The stores may be closing, but the technology is finding its real market — and it looks nothing like the cashierless grocery store that Amazon originally envisioned.
The gap between the autonomous retail vision and the autonomous retail reality has never been wider — or more instructive. For a decade, the pitch was simple: computer vision and AI would eliminate checkout entirely, creating seamless shopping experiences that would reshape brick-and-mortar retail the way e-commerce reshaped it in the 2000s. What actually happened is a case study in how promising technology meets the brutal economics of physical retail.
The Amazon Go experiment and why it failed
Amazon launched its first Go store in Seattle in January 2018, showcasing a shopping experience where customers could grab items and walk out without scanning anything or waiting in a checkout line. The underlying Just Walk Out technology — a combination of computer vision, object recognition, advanced sensors, and deep learning models — genuinely worked. Transactions were accurate, customer satisfaction was high, and the technology demonstrated that cashierless retail was technically feasible at convenience-store scale.
What the technology couldn’t solve was the business model. Operating small-format stores in expensive urban locations with sophisticated AI infrastructure produced unit economics that never reached profitability. Amazon expanded to roughly 30 Go locations and several dozen Fresh grocery stores before acknowledging the math didn’t work. In January 2026, Amazon announced the closure of all Go and Fresh locations, shifting its physical retail strategy entirely to Whole Foods expansion — planning over 100 new Whole Foods stores in the coming years.
The embarrassing 2024 revelation that roughly 1,000 workers in India were manually reviewing Just Walk Out transactions didn’t help. While Amazon maintained the system was overwhelmingly automated and human reviewers handled edge cases, the optics of an AI-powered store requiring a small army of human verifiers undermined the narrative of fully autonomous retail.
Where the technology actually works
The irony of Amazon Go’s closure is that Just Walk Out technology is thriving — just not in the places anyone originally expected. Amazon Web Services now licenses the technology to more than 145 third-party retailers across the US, UK, Australia, and Canada, with plans for significantly larger expansion in 2026 and beyond. The Seattle Seahawks’ Lumen Field alone has 15 Just Walk Out deployments, and the first shop reported an 85% increase in transactions and a 112% increase in sales per game compared to traditional concession stands.
The pattern is clear. Autonomous checkout succeeds in environments with three characteristics: high traffic density, limited product assortment, and time-pressured customers. Stadium concessions, airport convenience stores, hospital cafeterias, and corporate breakrooms all fit this profile. Traditional grocery stores — with tens of thousands of SKUs, produce that needs weighing, and price-sensitive customers who browse rather than grab — don’t.
This is consistent with what the robotics revolution in logistics has demonstrated across industries: automation succeeds fastest in constrained, predictable environments and struggles in open-ended ones. The warehouse robot that follows optimized paths performs brilliantly; the retail robot navigating unpredictable human behavior in a crowded grocery aisle does not.
The startup shakeout
The competitive landscape in autonomous retail has thinned dramatically. Grabango, which raised between $73 million and $93 million and secured a pilot with Aldi, shut down in October 2024 after failing to raise additional funding. Standard Cognition hit a $1 billion valuation on $150 million in funding but has narrowed its focus to specific niches rather than pursuing the original vision of revolutionizing all of retail.
The survivors are the companies that adapted. AiFi, backed by Aldi, Carrefour, and Zabka, has reduced store deployment time from six months to under a week using a camera-only approach that’s cheaper and less complex than Amazon’s hybrid camera-and-sensor system. Trigo, which uses the camera-and-shelf-sensor hybrid approach, has found traction with European grocery chains. Zippin has focused on small-format deployments in stadiums and corporate environments — the same niche where Amazon’s own technology is finding its best product-market fit.
The common thread among the survivors is restraint. The companies that tried to replace the entire grocery checkout experience failed or pivoted. The ones that identified specific high-value use cases where the technology’s advantages — speed, 24/7 operation, reduced theft, labor efficiency — clearly outweigh its costs are building sustainable businesses.
What this means for physical retail
The autonomous retail story is ultimately a story about where enterprise technology investment actually pays off versus where it’s venture capital subsidizing a compelling demo. The technology works. Computer vision can track shoppers and inventory with remarkable accuracy. AI models can process transactions in real time. The engineering problem is largely solved.
The business problem is harder. Full-store autonomous checkout requires significant upfront infrastructure investment — cameras, sensors, edge computing hardware, and ongoing AI model maintenance. For a convenience store doing $2 million in annual revenue, the payback period can work. For a grocery store doing the same revenue with thinner margins and more complex operations, it often doesn’t. The explosive growth in the broader robotics market hasn’t translated into the autonomous retail windfall that investors expected.
The most likely future isn’t the cashierless store that Amazon envisioned — it’s a hybrid model where autonomous checkout handles specific zones or product categories within conventional stores, self-checkout kiosks handle routine transactions, and staffed registers remain for customers who want human interaction. Retailers like Walmart and Kroger are investing in this layered approach rather than betting on fully autonomous stores.
Amazon Go’s real legacy won’t be the stores that closed. It will be the proof of concept that enabled hundreds of autonomous deployments in stadiums, airports, and corporate facilities — environments where the technology solves real problems at economics that actually work. Sometimes the most important thing a first mover discovers is that the market they targeted isn’t the market that wants their product.
