How Kroger bet $3 billion on robots, then went back to its stores

Kroger spent seven years and close to $3 billion building robotic warehouses to fill online grocery orders.

Then it shut the whole thing down. It wrote off $2.5 billion, paid Ocado $350 million to walk away, and went back to filling orders from the 2,700 supermarkets it already owned.

And the quarter it made that call, its online business turned a profit for the first time.

This is interesting because Kroger is one of the few companies that has tested both answers to the same question at full scale: should an online order come from a robot warehouse, or from the store down the street? 

Kroger tried both. Only one worked out.

In this breakdown, I dive into:

Let's get into it.

In case you aren't familiar with Kroger: it's the largest traditional supermarket operator in the US. About $147 billion of groceries a year through roughly 2,700 stores, mostly under names you wouldn't connect to the company: Ralphs, Fred Meyer, Harris Teeter, King Soopers, Smith's. 

It has been selling groceries since 1883, on the thin margins that come with it. It also makes a big share of its own store brands, runs its own trucks, and owns a data business built on the purchases of 62 million households.

Before any of this, Kroger ran a conventional grocery supply chain, and ran it well.

Goods flowed from suppliers and from Kroger's own plants into distribution centers, then out to stores on a mix of contracted carriers and Kroger's own trucks. Groceries are mostly perishable, so the system was built for speed. In fiscal 2018, Kroger turned its inventory about 14.7 times a year, holding stock for roughly 25 days. Walmart that year turned about 8.5 times; Costco about 11.8. A grocer either moves food fast or throws it away.

Three parts of that setup matter later.

What Kroger didn't have was an answer to online grocery.

Four forces reshaped Kroger's supply chain, two from outside the company and two of its own making.

Amazon bought Whole Foods in 2017, delivery apps scaled, and customers began expecting groceries at the door. Demand was never the issue; the cost of filling it was. Sending a worker down the aisles with a cart and then driving a few bags to a single house is expensive, and grocery margins leave no room to absorb it. 

Every grocer faced the same math. Kroger's answer became the biggest supply-chain bet in its history.

The pandemic sent online grocery vertical in 2020 and 2021, and made the fulfillment buildout look urgent. What came next was harder: years of food inflation, then a customer with less to spend.

When pandemic-era SNAP emergency allotments ended in 2023, recipient households lost meaningful monthly benefits overnight, and grocers absorbed the demand shock. Shoppers traded down to smaller baskets, cheaper items, and store brands. The pressure hasn't let up. CEO Greg Foran told investors in 2026 that "high gas prices and reduced SNAP benefits are squeezing budgets. Customers are managing spend carefully and shopping with real intent."

Some of what looks like weak demand is accounting, not lost volume. New Medicare drug-price rules shave about 130 basis points off reported sales while staying profit-neutral, and egg deflation took off another 64 basis points in the most recent quarter. Roughly two points of the headline softness is deflation and pharmacy mechanics. The customer is stretched, but demand is stronger than the headline number suggests.

For the supply chain, the squeeze came down to one instruction: fund lower shelf prices without giving up margin the company doesn't have. That pushed Kroger toward private label, direct sourcing, and cost cuts at every node.

In October 2022, Kroger agreed to buy its largest traditional rival, Albertsons, for $24.6 billion. The deal would have combined Kroger's roughly 2,700 stores with Albertsons’ roughly 2,200 under banners like Safeway, Vons and Jewel-Osco, the largest supermarket acquisition on record.

The case was scale. A combined company would buy more from every supplier, which lowers the cost of goods. It would spread the fixed cost of distribution centers, technology and manufacturing over a much larger base. And it would carry more weight against Walmart, Amazon and Costco, which had been taking food share for a decade. In Kroger’s telling, the savings would fund lower prices.

To clear regulators, the companies agreed to sell 579 stores to C&S Wholesale Grocers for about $2.9 billion. Regulators weren't convinced. The Federal Trade Commission and several states sued, arguing the merger would raise grocery prices, weaken union bargaining power, and hand the divested stores to a buyer too weak to keep them competitive. In December 2024, courts in Oregon and Washington blocked the deal. Albertsons terminated the agreement the next day and sued Kroger for the $600 million termination fee plus billions in damages. Kroger countersued. C&S filed its own suit over a $125 million fee. All of it is pending in court.

For the supply chain, the collapse meant one thing: the scale option was gone. Every cost advantage Kroger wanted now had to come from inside its own network. The merger also ate two years of management attention, much of it overlapping with the stretch when the warehouse bet was coming apart. For most of 2023 and 2024, leadership was fighting for a merger while its flagship fulfillment program stalled.

Rodney McMullen, CEO since 2014 and the champion of the warehouse bet, had framed the Ocado partnership as a “relentless drive to innovate way ahead of the market.” He resigned in March 2025 after a board investigation into personal conduct unrelated to the business. Lead director Ron Sargent stepped in as interim CEO and ordered a site-by-site review of the Ocado network. That review produced the reversal. 

In February 2026, the board named Greg Foran CEO. Foran ran Walmart US for six years and most recently led Air New Zealand. His career was built on store operations. So was the direction Kroger was already heading.

The warehouse bet, and the unwinding of it, played out in five decisions.

Decision one, 2018: Own the technology. Kroger signed Ocado as its exclusive US partner and took a minority stake. The plan was up to 20 automated customer fulfillment centers, the giant warehouses Ocado calls sheds. Inside each one, thousands of robots roll across a grid of stacked crates, pulling totes of groceries and carrying them to human packers at the edge. One site could pick many stores' worth of online orders with a fraction of the labor.

The network was hub-and-spoke:

Concentrate the picking in a few high-throughput hubs, and the cost per order should drop below anything a store picker could match. Own the best fulfillment technology in grocery, and competitors can't copy the edge. The whole plan rested on one belief: centralized scale beats proximity.

Decision two, 2020 to 2021: Scale into the surge. The first two centers opened in Monroe, Ohio and Groveland, Florida in early 2021, into a pandemic that had put online demand on a vertical. More followed across the Mid-Atlantic, the Upper Midwest, and the South. Volume was never the problem. The surge proved the sheds could process orders. It said nothing about whether they made money per order, and that question got buried for two years.

Kroger's own financials told a different story than the press releases. Sales crossed $132 billion in fiscal 2021, but inventory stayed pinned near $7 billion, and capital intensity fell to 1.9% of revenue by fiscal 2022, the lowest point in the whole window. A transformational infrastructure program doesn't usually coincide with the lightest capital spending in a decade. The capital line suggests the build was losing internal priority before anyone said so out loud.

Decision three, 2023: Stop building. As demand normalized, the math surfaced. It came down to density. Two things worked against the sheds:

Picking cost per item fell inside the building. Total cost per delivered order didn't, because the savings went right back out the door on trucks and vans. By late 2023, Kroger had slowed new openings, with eight centers built against the plan's 20.

Decision four, 2024 to 2025: Unwind. The first three spoke facilities closed in March 2025 because they "did not meet the benchmarks we set for success." Sargent's review followed, and in November 2025 the company restructured the network:

The reversal cost roughly $2.5 billion in pre-tax charges, $1.9 billion after tax. Alongside the closures, Kroger set a target of $400 million in improved e-commerce profitability for fiscal 2026.

Decision five, 2026: Run the store as the warehouse. Kroger now fills most online orders from its 2,700 supermarkets. A worker walks the aisles with a cart and a handheld, picks the order, and stages it for pickup or a delivery driver. The store sits close to the customer, which is exactly what the remote sheds lacked.

Two numbers explain why it works:

Kroger also stopped trying to own delivery. 

DoorDash and Uber Eats now fulfill from the store network, and Instacart became the primary delivery provider inside Kroger's own app. Renting three delivery networks was faster and cheaper than building one. Convenience orders delivered in under an hour drove about half of digital growth last quarter, a speed the remote sheds could never have hit.

The result came early. Kroger's e-commerce business, including retail media, turned an operating profit in the first quarter of fiscal 2026, ahead of the schedule set only months earlier. CFO David Kennerley said the company kept "nearly all" of the households the closed centers had served, and that cost to serve is at record lows.

The seven-year numbers. From fiscal 2018 to fiscal 2026, revenue grew from $122.7 billion to $147.6 billion. Inventory stayed inside a $6.5 to $7.6 billion band the whole time. Turns drifted up from 14.7 to 16.4, and days of inventory tightened from about 25 to about 22. Capital intensity fell to its 1.9% trough in 2022, then climbed back to 2.73% by 2025. Based on our analysis of company filings, Kroger turns inventory faster than any large peer: Costco about 13.1 times, Albertsons 11.9, Walmart 9.3, and Target 6.0.

The automation era is nearly invisible in those numbers. The core supermarket business kept doing what it always did while the warehouse project was built and unwound beside it. What moved the e-commerce P&L was the retreat to the stores.

Automation isn't gone. Kroger is adding smaller, lighter automation inside high-volume stores, where the order density already exists. One large centralized bet became many small ones, placed where the volume is.

The squeezed customer pushed Kroger deeper into the part of the supply chain it controls outright.

Its private-label business, Our Brands, runs more than $39 billion a year across over 13,000 items, a bit more than a quarter of sales, led by Private Selection, Simple Truth, and the Kroger brand. As shoppers traded down, store brands outgrew national ones; last quarter Our Brands beat national brands by 175 basis points in share gain even with dairy deflating. The whole industry is moving the same way. US private-label sales hit a record $282.8 billion in 2025, and by industry estimates Kroger's own-brand share, around 27%, sits near Walmart's (about 30) and Albertsons' (about 26, targeting 30). Aldi runs near 80.

The 33 plants make about 20% of those units in-house, which gives Kroger control over cost, supply, and launch speed that a pure reseller doesn't have. A private label, at that scale, is a physical supply-chain asset, not a branding exercise.

Kroger is also buying more directly, cutting out intermediaries it historically used in fresh imports, national brands, and its own brands. Private label and direct sourcing do the same job from different ends: they fund lower shelf prices without surrendering margin the company can't spare.

The 84.51° data arm has moved from merchandising into the supply chain itself.

Kroger now runs AI tools built on its own sales and shipment data to manage inventory, especially in fresh, flagging what's selling and what's close to expiring. The company credits this with cutting shrink, the industry's term for product lost to spoilage, damage, and theft, one of the largest controllable costs in grocery. The same signal that reduces waste also tightens orders and frees working capital.

Foran has been blunt about the cost problem the tools are meant to solve: "operating costs have been growing faster than our sales. That's not sustainable, and frankly, it's not acceptable." His program pairs fewer organizational layers with a new Capability Center and a push to apply AI across operations.

The same shopper data runs Kroger's most profitable growth business, and that business helps pay for everything else.

Kroger Precision Marketing sells advertising to brands against Kroger's purchase data. Retail media networks run operating margins of 50-70%, a level groceries never touch, and Kroger's alternative-profit operating income has climbed from about $1.3 billion to $1.5 billion over three years. Every large retailer is building the same engine: Walmart Connect is around $4.4 billion in ad revenue, Target's Roundel around $649 million. Across the industry, media profit is the cushion that funds price investment. In 2025, Kroger unified 84.51°, the media business, and loyalty marketing into a single organization.

One caveat. The "profitable e-commerce" milestone is e-commerce including media, and media is the high-margin half. An ad business at more than 50% margins can carry a fulfillment operation that, on its own, sits much closer to break-even. The store pivot made picking and delivery cheaper. The ad engine makes the reported number look better than the picking economics alone would.

Kroger runs its own fleet alongside contracted carriers and has worked for years on fuel efficiency and alternative-fuel trucks, because diesel swings its logistics costs in ways it can't control. That exposure is live: last quarter, higher oil prices put about 15 basis points of pressure on gross margin through transportation costs. Through 2025, the company repeatedly cited lower supply-chain costs as a margin tailwind, even with fuel moving against it.

Kroger's supply chain is settling into a clear shape: stores as the fulfillment base, partners handling delivery, light automation where volume justifies it, private label and direct sourcing holding down cost, and one data organization underneath all of it.

Kroger isn't alone. 

Ahold Delhaize is closing six of its own centralized e-commerce facilities on the same store-based logic. Walmart sold its robotics unit to Symbotic and committed $520 million to co-develop automation for hundreds of store-attached pickup and delivery centers. Amazon is piloting a micro-fulfillment center inside a Whole Foods. Albertsons is the holdout still building standalone warehouse automation. The contest is no longer store versus shed. It's what kind of automation goes inside the store, and Target sets the pace, filling most online orders from stores and getting nearly 80% of them out within a day.

Four things will decide whether Kroger's version holds:

Kroger's stores were a fulfillment network the whole time. Nobody called them that, but that's what they were. The building is paid for, the shelves are stocked, the staff is already clocked in. Putting an online order on top of that is cheap.

The robot warehouses never got that head start. Each one had to cover its whole cost from online orders alone, in places that just didn't have enough of them.

So the first move isn't to build. It's to go through what you've already got. The cheapest capacity in your business is usually the capacity you're already paying for.

The Ocado machines worked. That was never the issue. It was where Kroger put them.

They went up on cheap land outside the cities, miles from the customers. So the picking got cheaper, the driving got a lot more expensive, and the driving won.

Automation like that only pays off when it's running flat out, close to steady demand. If you're sizing up a big automated site, you've already made the real call the moment you pick the location. Map the actual order density around each one, at normal demand and not some pandemic peak, and be honest with yourself about whether it clears the bar.

This is the one I'd actually steal. Kroger's e-commerce didn't turn a profit just because store picking got cheaper. A big chunk of it came from something with nothing to do with boxes or trucks: ads.

Kroger sells advertising against its shopper data, at margins nothing like what you make selling groceries. That business helps pay for the one that moves the boxes, which lets the whole thing run leaner and keep prices down.

If you've got proprietary data and traffic, there might be a profit layer sitting right next to your supply chain, one that can fund the parts that will never really make money on their own. 

The best assets earn twice. Once by making the operation better. Once by helping foot the bill.

Kroger spent seven years and billions of dollars looking for its e-commerce answer in new warehouses and a bigger company. It ended up building that answer out of the stores, plants, trucks, and data it already had.