Nearly six years ago Walmart sent 500 shelf-scanning robots home after concluding that workers picking online orders could see the
Conversations On Retail
July 20, 2026
At Amazon’s Delivering the Future event near London on June 4, the company showed a version of its Proteus warehouse robot that workers direct in conversational language instead of code. Scott Dresser, vice president of Amazon Robotics, said the machine works out “the priority, the route, the timing” once a worker tells it what needs doing. The original Proteus, in service since 2022 and running in 25 U.S. fulfillment centers, was confined to dock areas and moved carts weighing close to 400 kilograms. The new model is built to operate anywhere items need to move across a site, and Amazon plans to deploy it in Europe in the first half of 2027.
The robot arrived as one line in a much larger commitment. Amazon tied the announcement to more than 10 billion euros, around 11.6 billion dollars, to expand and modernize its European fulfillment network, alongside two other systems it is scaling, STARK, a tote-handling robot piloted in Barcelona and slated for 15 European sites by 2027, and Vulcan, its first touch-sensitive robot, which it is expanding across Europe. The company said it would add 25,000 European roles over the coming years and committed a billion dollars to its Career Choice training program by 2030. More than a million robots already work across its network.
The detail that matters for anyone selling into these networks is not the robot’s vocabulary. It is what a plain-language, deploy-anywhere machine signals about the shape fulfillment automation is taking, and how sharply that shape now differs across three of the largest retailers in the United States. Amazon is widening where automation operates and lowering the skill needed to direct it. Walmart is pushing automation down to the store. Kroger has just paid to pull back sharply from the centralized model it spent seven years building.
Kroger said in November it would close three automated customer fulfillment centers, in Pleasant Prairie, Wisconsin; Frederick, Maryland; and Groveland, Florida, and would take an impairment and related charges of about 2.6 billion dollars in its fiscal third quarter, according to Grocery Dive and Chain Store Age. In December it canceled a planned center near Charlotte and moved to close a spoke facility in Nashville, with roughly 132 jobs affected, and agreed to pay Ocado 350 million dollars, per filings reported by Grocery Dive and Supply Chain Dive. The network it built with Ocado from 2018, an initial commitment to capacity equivalent to as many as 20 customer fulfillment centers, only ever reached eight open sites and had not met the grocer’s financial expectations. Ron Sargent, then Kroger’s chairman and interim chief executive, framed the move around faster delivery and profitable growth, citing five consecutive quarters of double-digit e-commerce sales growth. The company is redirecting toward in-store fulfillment and deeper third-party delivery partnerships, targeting roughly 400 million dollars of e-commerce profitability improvement in 2026.
Walmart’s newest bet pushes automation past the distribution center and down to the store itself. In early 2025 it sold its Advanced Systems and Robotics business to Symbotic, its automation partner since 2017, for 200 million dollars plus up to 350 million in contingent payments, and committed 520 million dollars to fund development and to buy and deploy 400 Accelerated Pickup and Delivery systems, according to Supply Chain Dive and Digital Commerce 360. Those APDs are store-level micro-fulfillment units, which suits Walmart’s structural position, since roughly 90 percent of the U.S. population lives within ten miles of one of its more than 4,600 stores, per DC Velocity. Symbotic continues to run automation across Walmart’s 42 U.S. regional distribution centers.
Read together, the three disclosures describe a category sorting itself out, not a single arms race. The model under pressure is the one Kroger just wrote down, large and purpose-built and capital-heavy, the kind that has to fill its own throughput before it pays off. The money in these three cases is moving toward systems that are more distributed and more flexible, whether that means Walmart’s store-level APDs or an Amazon robot that can be told what to do without a programming interface and put to work wherever a site needs it. For commercial and supply-chain leaders, the implication is less about which retailer owns the most robots and more about how quickly fulfillment cost and capacity can now change. A system that deploys without a bespoke building and without specialized programming compresses the timeline on which a retail partner can reconfigure how, and from where, a brand’s orders get picked.
That timeline lands differently across the audience. For a large CPG selling into all three networks, the near-term work is in the planning assumptions, since store-based fulfillment changes where inventory needs to sit and how fill rates behave, and Kroger’s reversal is a caution against treating any one fulfillment design as permanent. For a brand scaling into these channels, a lower barrier to deploying automation cuts in its favor, because capacity can expand without waiting on multiyear warehouse construction. The labor dimension stays unresolved and contested. Amazon says it is creating jobs alongside the technology and has hired hundreds of thousands of workers since introducing robotics, while internal documents reported by The New York Times in October described an expectation that automation could let the company avoid hiring more than 160,000 U.S. workers by 2027, a characterization an Amazon spokesperson said gave an incomplete picture. For all the attention the new Proteus drew, it remains a lab pilot with no commercial deployment scheduled before the first half of 2027. The changes brands will feel sooner are the store-level systems Walmart is deploying through Symbotic and the centralized capacity Kroger is taking offline in January.