Walmart Spent Billions on Warehouse Robots. The Robots Keep Getting Beaten by Cardboard.

The warehouse of the future was supposed to run itself. At Walmart, the future still needs a hand lifting the dog food. A Wall Street Journal report published October 9 details the messy reality behind the retail giant's multibillion-dollar campaign to automate its warehouses, and it reads less like a victory lap than a field diary from the front lines of the automation revolution: cardboard boxes that defeat sorting machines, robotic arms that fumble frozen turkeys, robots that break down from dust, and electric bills that tripled.

None of this means the effort is failing. Walmart says its automation rollout is roughly on track, and it just opened another next-generation fulfillment center in Stockton, California. But the report is a rare, honest window into what happens when cutting-edge robotics meets the stubborn physicality of the real world, and why automating labor turns out to be a problem of boxes and dust rather than intelligence.

The turkeys and the cardboard

Start with the poultry. In a newly automated grocery warehouse in California, workers discovered that the machine's robotic arms could not handle large frozen turkeys. The problem surfaced ahead of the Thanksgiving rush, and the fix was decidedly low-tech: shifting turkey shipments to nearby grocery warehouses. The robots, for all their sophistication, lost a confrontation with a frozen bird.

Cardboard proved equally defiant. Sorting machines struggled with boxes that were too large for the equipment, so Walmart switched to box designs that work better with the machinery. This is the deeper pattern of the entire project: the world was not built for robots, so Walmart is rebuilding the world, one box at a time.

At the company's automation test site in Brooksville, Florida, the warehouse is now nearly fully automated. About 80 percent of inventory moves through the building with automation, according to Rob Montgomery, the head of supply chain operations at Walmart U.S., who oversees the automation of nearly 200 locations. The remaining 20 percent, heavy items like large-screen TVs and bulk bags of dog food, is where the robots still stall. That last fifth is the hardest fifth.

We are kind of in this peak complexity.

That is Montgomery's own description of where the program stands, speaking to the Journal. Peak complexity, in this case, means a system that mostly works and keeps failing in interesting ways.

The robots break down. The power bill doesn't.

The workhorse of the warehouse operation is a fleet of small, wheeled robots made by Symbotic, a Massachusetts robotics company in which Walmart holds a stake. The robots zip through warehouses on motion sensors, shuffling product into neatly built pallets. They are fast, precise, and, it turns out, finicky. Stoppages can cascade through a facility, and something as minor as dust buildup can knock machines offline.

Symbotic says the breakdown problem is improving: under 10 percent of its robots were being repaired at any given time last year, a figure that has dropped to about 5 percent in recent months. That is genuine progress, but it is progress measured in months of maintenance crews, spare parts, and engineering attention.

Then there is the electricity. Automated grocery facilities can run monthly power bills of $800,000 or more during peak months, up from about $250,000 when operations were mostly manual. The robots need juice, and so does the building that houses them. Walmart expects its automation spending to peak this year and next, but the honest accounting shows the robot future costs more before it costs less.

So do the people. The mechanical and electrical engineers needed to keep the automated systems running cost more than the unskilled warehouse workers they replace. Walmart employs more than 140,000 warehouse workers today, and the economics of swapping a forklift driver for a robotics engineer do not pencil out on day one. The bet is on the curve, not the starting point.

Building the plane while it ships

The deepest difficulty is structural. Walmart is not building robot-native warehouses from scratch, the way Amazon did. It is retrofitting automation into warehouses built in the 1980s and 1990s, and doing it while the buildings keep shipping goods. Tearing up a functioning warehouse to install robots, without stopping the trucks, is the logistics equivalent of open-heart surgery on a marathon runner.

It is also running a zoo of systems: different robotic technologies from multiple vendors, doing different tasks at different locations. Each integration is its own engineering project. The company's primary goal, assembling perfectly ordered pallets that unload cleanly at stores, remains a work in progress, though executives at both Walmart and Symbotic say they are getting closer.

The stores, meanwhile, are a separate saga. Walmart has spent nearly $1 billion since 2016 trying to automate online order fulfillment inside its retail locations, where space is tight and shoppers are underfoot. It bought Alert Innovation, the startup behind one system, for $400 million in 2023, then sold it to Symbotic at a 50 percent loss. Symbotic is testing a new version of the store system in Dallas, and does not expect a workable model until 2028 at the earliest.

What about the jobs?

The question that shadows every automation story: do the robots take the jobs? Walmart's answer is carefully calibrated. Executives told the Journal that automation will eventually mean fewer warehouse workers, but that the company's famously high turnover means attrition absorbs the change, with no layoffs. Sales growth could keep headcount roughly where it is today.

The timing is also telling. In the same week the report landed, Walmart opened its fifth next-generation fulfillment center, a 900,000-square-foot facility in Stockton, California, that will create more than 1,000 jobs. A sixth, at 1.5 million square feet, is planned for Carnesville, Georgia. These next-generation facilities, which use networks of robotic devices and conveyor belts to cut fulfillment from a 12-step process to five, are not jobless. They are differently jobbed.

The lesson for the automation era

Walmart's struggle is not a verdict on automation. It is a field manual. The lesson, one that supply chain veterans keep repeating, is that robots excel at predictable conditions and fail at messy ones: the oversized box, the frozen turkey, the dusty floor. Amazon understood this and built buildings designed for machines. Walmart is learning it by retrofitting buildings designed for people, while they operate.

There is a version of this story in which the Journal's findings are the best possible news for the long-term bull case. Every problem is tractable: redesign the box, tune the maintenance, scale the software. The pallet Tetris gets easier as more products flow through the system, because the software has more items to choose from. The complexity is peaking precisely because the system is nearly finished being complicated.

But the deeper truth is harder to schedule. Automating the physical world is not mostly an intelligence problem. It is a friction problem: dust, cardboard, gravity, Thanksgiving. The companies that win the automation race will not be the ones with the smartest robots. They will be the ones most willing to redesign everything around them, starting with the humble cardboard box.