Amazon plots 250k-sku mega-marts to gut walmart’s grocery lead
amazon is quietly drafting a store that swallows Walmart whole. Internal blueprints reviewed by TechBloom reveal a 200k-square-foot hybrid—part supermarket, part robot hive—where a quarter-million items orbit above shoppers’ heads in an AutoStore lattice and drop into bags before the customer hits the exit. The code name is Project Kobe, and the first site, hidden in plain sight outside Chicago, opens late 2027.
The back room is the front room
Flip the floor plan and you get the point: 45 % of the footprint is windowless warehouse. Conveyor fingers reach into refrigerated vaults, plucking strawberries and phone chargers with the same mechanical shrug. Walmart keeps fulfillment “out back”; amazon wants it dead center, visible only through polycarbonate panels that fog with condensation every time a 500-foot grocery run is compressed into 38 seconds of robot transit. The company’s own math shows each picked unit costs 12 % more than a same-day Prime van drop, a surcharge amazon will try to bury inside higher basket totals and membership inertia.
Three more sites—Cherry Hill, Edison, Oak Brook—are already zoned and leaking drywall dust. If the pilot proves the algorithm can sling 250k SKUs without choking on its own cable trays, dozens follow, then hundreds. The long-term roster dwarfs the 500-ish Whole Foods flagships that never cracked 3 % grocery share and now sit like expensive bonsai inside the larger jungle of amazon’s retail experiment.
Orbital freezers and the last human mile
AutoStore grids are dense but dumb: they can’t tell a rib-eye from a Popsicle. Enter Orbital, amazon’s still-vaporware chilled cube designed to shove ambient, refrigerated and frozen inventory through the same bot highway. Engineers say it’s two years out, which in Amazon time means “maybe before the heat death of the universe.” Until then, flesh-and-blood pickers will jog the produce aisle, scanning QR codes on $4-a-pound peaches while robots overhead ferry Bluetooth speakers. The split labor model is deliberate: 70 % of chilled orders still start on human feet because the cold chain is a jerk to automation.
Even with $33 million in capex per site, the budget line that makes finance VP’s palms sweat is perishables shrink. A single mis-temperatured pallet of avocados can erase the margin on 10,000 units of detergent. Amazon’s workaround is an AI layer—nicknamed Frida—that re-plans the entire store nightly, telling clerks which mango stack to cull and which robot tote to refrigerate. Frida is still crib-learning; planners admit they’re nowhere near the “North Star” where the machine orders its own HVAC tweaks.
Walmart’s moat is asphalt
While Amazon drills concrete, Walmart already touches 93 % of U.S. households with same-day groceries using parking lots it bought in 1994. The math stings: Walmart’s grocery share is 21 %, Amazon’s 3 %. Kobe’s 250k SKU headline doubles a Supercenter’s assortment, but Walmart can retaliate tomorrow by dropping another 20k items into existing shelf caps overnight—no construction permits, no union drywall crews, no Waiting for Orbital.
Jason Buechel, CEO of Whole Foods and newly crowned czar of all Amazon groceries, owns the scoreboard. He’s told insiders the goal isn’t to beat Walmart at Walmart’s game; it’s to make the game irrelevant—turn the supermarket into a node on the same cloud network that already ships cat litter by drone. The rhetoric is classic Amazon: redefine the playing field until competitors look like they’re standing on a chessboard with checkers pieces.
Come 2028, either Kobe’s robots will have chopped grocery margins thinner than deli parchment or Walmart’s 4,700 Supercenters will still soak up sunrise traffic while Amazon’s glass-fronted fulfillment cathedrals hum half-empty, waiting for Frida to learn what grandmothers in Ohio want for dinner. Place your bets; the first 500-foot strawberry run happens in 1,200 days.