The true cost of a robot that doesn't work: anatomy of a $2.5 billion failure

In November 2025 Kroger told the SEC it would take about $2.6 billion in impairment and related charges on its automated fulfillment network [1]. Two weeks later it filed again to add that the figure included a cash payment to its vendor, Ocado, of about $350 million [1]. Nothing had broken. No vendor went bankrupt. The robots worked.

This is the most expensive documented automation failure in the sector. Every detail sits in public filings. It is worth reading closely, because the failure modes are ordinary ones.

The sentence that matters most

The headline was three closures. The accounting note says something larger. Kroger tested each site as its own asset group. It then reported that the test showed 'the carrying value of each asset group exceeded the sum of undiscounted cash flows' the assets were expected to produce [2].

Read that again. Every site failed, not only the three being closed. And it is a low bar. Undiscounted cash flow asks whether an asset will ever produce more cash than its book value. It ignores the time value of money completely. The whole network failed that test.

How big the gap was between plan and outcome

When the partnership launched in 2018, the two companies identified 20 potential sites. They opened eight [8]. Three closed in January 2026, one planned site in Charlotte was cancelled [3], and five remain. The chief financial officer said the retained sites would be monitored 'with a focus on improving operating efficiency'.

So a programme scoped at 20 sites delivered eight, kept five, and wrote off $2.5 billion on the way [4].

Four causes, each with evidence

Kroger never named a root cause. Its filings and its critics between them make four visible.

Not enough orders. Kroger says it will keep using automation 'in geographies where Kroger sees higher density of demand' [3]. A former Kroger executive was blunter. The centers were 'just not processing enough orders to pay for all that technology investment you had to make' [5].

Sites in the wrong places. The same executive called locating the centers outside cities 'a key flaw' [5]. Kroger's annual report gives this away in an aside. It reports eCommerce growth 'excluding the effect of fulfillment center exits in markets where Kroger does not operate stores' [4]. Two of the three closed sites sat in markets with no Kroger stores to anchor them.

The wrong bet about customers. The model assumed shoppers would trade delivery speed for better prices, which is how the technology works in the UK. US shoppers chose speed, and rivals moved to same-hour delivery [6].

Throughput that never covered the capital. This is the accounting consequence of the first three, and it is what the recoverability test measured [2].

The warnings were visible two years early

Kroger paused development of new fulfillment centers in September 2023. In March 2024 it closed three smaller spoke facilities, with a spokesperson saying they 'did not meet the benchmarks we set for success' [6]. The full site-by-site review came in September 2025, and the write-off followed in November.

So the operator had benchmarks, had sites failing them, and kept the network running for two more years. The lesson is not that nobody measured. It is that measuring is only useful if a failing number triggers a decision.

What was never disclosed, and why that matters to you

Neither company has ever published a throughput figure for any of these sites. No utilisation number. No orders per site. Ocado called the closed sites 'underutilised' without quantifying it [7]. The public record of the sector's largest automation failure contains no operating metric at all.

That is the gap this field lives in. Buyers commit hundreds of millions against performance claims. When the projects fail, the numbers that would teach the next buyer never appear. You will not get them from a case study. You have to measure them yourself, in your own building, before you scale.

What it cost beyond the write-off

• The $350 million cash payment to exit the vendor relationship [1], on top of the impairment.
• Ocado received closure fees of £261 million in January 2026 [7], so one side's loss was the other side's revenue.
• Spoke facilities and jobs went with the hubs, including a Nashville site closing with 132 layoffs [6].
• Two years of capital that the chief executive later said would have been better spent on stores [5].

The uncomfortable general case

It would be easy to read this as one company's siting error. The survey data says otherwise. Across 767 US operations leaders, 89% say technology investments have not fully delivered [9]. Among leaders who have deployed warehouse robotics, only 34% are fully satisfied [10].

Kroger is not an outlier. It is the case where the number was large enough, and the company public enough, that we get to read the filing.

What to do differently

• Test the demand assumption before the automation assumption. Kroger's technology worked. The order volume did not arrive.
• Set the benchmark before go-live, and decide in advance what a failing number triggers. Kroger had benchmarks and still ran two more years.
• Model payback with the ramp and the displacement rate written down, not assumed [11].
• Pilot in your weakest candidate market, not your strongest. The strong one will succeed regardless and teaches you nothing.
• Negotiate the exit before you need it. Leaving cost Kroger $350 million in cash.

Sources

1. The Kroger Co., Form 8-K filed November 18 2025, Item 2.06: approximately $2.6 billion of impairment and related charges. Form 8-K/A filed December 5 2025: the charges include a cash payment to Ocado of approximately $350 million. sec.gov EDGAR, CIK 56873
2. The Kroger Co., Form 10-Q for the quarter ended November 8 2025, filed December 12 2025. Note 11 states: 'The recoverability testing indicated the carrying value of each asset group exceeded the sum of undiscounted cash flows expected to result from the use and eventual disposition of the assets.' The same filing records charges of $2,585 million, $1,968 million net of tax, 'related to the Company's fulfillment network not meeting operational or financial expectations, the planned closing of three automated fulfillment facilities and the cancellation of a planned site'. sec.gov EDGAR
3. Same 10-Q: 'Kroger identified opportunities to optimize its automated fulfillment network by closing facilities in Pleasant Prairie, Wis.; Frederick, Md.; and Groveland, Fla. in January 2026, which have not met operational or financial expectations, and canceling plans for the site in Charlotte, N.C.' and 'In geographies where Kroger sees higher density of demand, the Company will continue to utilize automated fulfillment'. sec.gov EDGAR
4. The Kroger Co., Form 10-K for fiscal 2025, filed March 31 2026: the full-year charge was $2,497 million, $1,908 million net of tax. The MD&A reports eCommerce sales growth 'excluding the effect of fulfillment center exits in markets where Kroger does not operate stores'. sec.gov EDGAR
5. Ken Fenyo, former Kroger executive and managing partner of Pine Street Advisors, quoted in Grocery Dive, November 19 2025: 'Kroger's decision to locate the Ocado centers outside of cities turned out to be a key flaw'; 'You didn't have enough people ordering, and you had a fair amount of distance to drive to get the orders to them. And so ultimately, these large centers were just not processing enough orders to pay for all that technology investment you had to make.' grocerydive.com
6. Grocery Dive, November 19 2025, on the demand model: Kroger 'bet that consumers would be willing to trade delivery speed for sensible prices', while 'U.S. consumers have shown they value speed of delivery'. The same reporting notes Kroger paused fulfillment center development in September 2023 and closed three spoke facilities in March 2024, a spokesperson saying they 'did not meet the benchmarks we set for success'. grocerydive.com
7. Ocado Group FY2025 results, published February 26 2026: 'Resetting our Kroger and Sobeys partnerships; now a combined base of 7 live CFCs; following 4 site closures'. CEO Tim Steiner refers to partners' decisions 'to close underutilised sites'. The results note 'closure fees from Kroger of £261m' received in January 2026. ocadogroup.com
8. Kroger and Ocado identified 20 potential locations when the partnership launched in 2018 and ultimately opened eight facilities. Supermarket News, September 2025 and January 2026. supermarketnews.com
9. PwC 2026 Digital Trends in Operations Survey (767 US operations and supply chain leaders): 89% say technology investments have not fully delivered expected results. pwc.com
10. DHL Supply Chain Insight 2030 survey, November 2025 (350 North American supply chain and C-level executives): only 34% of VP and Director level leaders are fully satisfied with warehouse robotics deployments. warehouseautomation.ca/news/dhl-report
11. Robot Eval, Labor math 2026 (this site), on payback assumptions, and Which robot should you buy? on treating vendor viability as part of the specification