Two warehouse automation stories landed within 24 hours of each other, and the contrast they draw is sharper than either alone. Walmart's multibillion-dollar campaign to roboticize roughly 200 distribution centers is hitting physical-world failures that its own supply-chain leadership calls "peak complexity"2,5. Meanwhile, Ultra Robotics closed $62 million in funding and expanded a partnership with Physical Intelligence, selling robots-as-a-service to the same class of facilities Walmart is struggling to automate1,6. ANALYSIS The juxtaposition frames a strategic fork in warehouse robotics: vertically integrated megaprojects versus modular, leased alternatives, with very different risk profiles for the buyer.
Why it matters
Warehouse automation is no longer a question of whether but how. Walmart accounts for roughly 85% of Symbotic's revenue and owns 12.6% of the robotics company's stock, a stake worth about $3.2 billion at recent prices. That deep entanglement means every technical setback ripples through both balance sheets. ◆ The scale of Walmart's commitment, and the persistence of its problems, suggests that brute capital alone does not resolve the physical unpredictability of warehouse environments.
The big picture
The Wall Street Journal's reporting cataloged a litany of real-world failures inside Walmart's automated warehouses: shipping boxes 1.5 inches too tall for the machines, robots idled by dust, and robotic arms that fumbled frozen turkeys in a California grocery facility before Thanksgiving. Walmart is not merely a customer of Symbotic; it sold Symbotic its in-store robotics unit, Advanced Systems and Robotics, after Walmart itself could not make the technology work, and is paying $520 million to finish that unit's development. ◆ The arrangement creates a circular dependency: Walmart funds the R&D, absorbs the equity risk, and still waits on the operational results.
Ultra Robotics occupies a different position in the stack. The Brooklyn-based startup raised a $50 million Series A led by Framework Ventures and a $12 million seed round led by Y Combinator and NextView Ventures. Its OP1 Operator robot, which resembles a humanoid torso mounted on a traditional robotic arm, uses forklift-prong grippers and has a vertical reach of up to 10 feet4. Ultra leases the devices under a monthly "robots as a service" model rather than selling large capital installations. The company says its robots have packed more than half a million orders for shipping.
Ultra CEO and co-founder Jon Miller Schwartz made a pointed observation in an interview with Fortune's Term Sheet: "Humanoid robots get the lion's share of attention, but it is other forms of robots that are having the most impact in the real world". The company also expanded its partnership with Physical Intelligence, the AI research firm founded by a team from Google DeepMind and valued at $5.6 billion, to supply AI software for Ultra's robots3.
ANALYSIS The two models diverge on where integration risk sits. Walmart's approach bundles hardware procurement, software development, facility redesign, and equity ownership into a single, tightly coupled program. When a box is 1.5 inches too tall, the entire system stalls. Ultra's RaaS model shifts integration risk toward the vendor: the customer leases a robot monthly and can, in principle, scale up or swap out without a capital write-down.
Ultra's decision to source its AI layer from Physical Intelligence rather than build it in-house further distributes technical risk. That contrasts with Walmart's position, where it ended up paying $520 million to develop technology it originally owned.
Schwartz's framing also carries competitive subtext. Ultra's OP1 Operator handles depalletization and kitting, separating mixed inbound shipments into single-SKU bins or assembling multi-item parcels. ANALYSIS These are precisely the tasks where Walmart's Symbotic installations have encountered the most friction, suggesting that the startup sees an opening in the gap between enterprise ambition and operational reality.
What's next
Ultra has raised prices, according to Fortune's reporting, a signal that demand currently exceeds supply. The Physical Intelligence partnership expansion will test whether foundation-model-driven AI software can handle the edge cases (dust, odd-sized boxes, frozen turkeys) that have tripped up Walmart's installations. Walmart, for its part, remains committed: its 12.6% stake in Symbotic and $520 million development commitment leave little room to reverse course. The next data point will be whether Symbotic can diversify its customer base beyond the single retailer that supplies 85% of its revenue.