Ocado Put the Supermarket on a Grid. Now It Has to Prove the Grid Travels.

The most revealing object in an Ocado warehouse is not a robot. It is a bag of groceries. A robot can move a crate quickly; the bag insists that a bottle of bleach stay away from the grapes, that the ice cream remain frozen and that the parsley arrive looking like parsley. Grocery is a fiendish automation problem disguised as an errand. Ocado has spent a quarter-century turning that errand into software, steel and recurring fees.

To British shoppers, Ocado is the supermarket with brightly coloured vans and no shops. To its corporate customers, Ocado Group is something else: a technology and logistics supplier whose Ocado Smart Platform, or OSP, runs the chain from webpage to warehouse to doorstep. It forecasts demand, decides where stock should sit, directs robots and pickers, assembles orders, plans delivery routes and helps a retailer operate the whole contraption. The grocery business now called Ocado Retail is a 50:50 economic venture with Marks & Spencer. The listed group increasingly sells the plumbing.

Three bankers and a problem nobody wanted

Tim Steiner, Jason Gissing and Jonathan Faiman left Goldman Sachs and founded the business in 2000. Their thesis was unfashionably physical. Instead of asking supermarket employees to roam aisles for online orders, Ocado would build a store that customers never entered. Goods would come to workers inside a dedicated customer fulfilment centre, or CFC; vans would leave on tightly planned routes. Pilot deliveries began in 2001, and the first Hatfield CFC helped the company reach 2.2 million households in 2002.

That choice cost more and taught more. Ocado had to write ecommerce software, model demand, manage expiration dates, design warehouses and run the last mile. In 2009 it launched a transactional grocery iPhone app. In 2010 it floated in London. By 2015 annual sales had passed £1 billion. The important turn came in 2013, when Morrisons signed as the first strategic customer. The apparatus Ocado had built for itself could be sold.

The product is the choreography

The camera loves Ocado's grid. Thousands of bots skim across its top, lower grippers into stacks of bins and bring the right tote to a picking station. More than 17,000 bots now travel the equivalent of 153 million kilometres in a year. Ocado says the system can retrieve a 50-item order in about five minutes. Yet the robot is only one musician. The commercial product is the score.

OSP is sold to grocers including Bon Preu, Monoprix, Kroger, Sobeys, ICA, AEON, Coles, Alcampo, Auchan, Lotte and Panda. In Britain, Ocado Logistics serves Ocado Retail and Morrisons. Outside grocery, Ocado Intelligent Automation offers mobile robots and orchestration software to customers such as McKesson Canada. Revenue comes from technology fees charged to partners, logistics fees and cost recharges, and capital sales or robotics-as-a-service arrangements. In FY2025, Technology Solutions produced £561.2 million of revenue and Logistics £800.3 million, for a group total of £1.3615 billion.

This integration distinguishes Ocado from a warehouse-equipment vendor. A retailer can buy an AutoStore cube, contract a systems integrator, let employees pick in stores, or hand demand to marketplaces such as Instacart. Ocado's wager is that grocery works better when forecasting, fulfilment and routes share one nervous system. Its 3,000-plus patent portfolio protects pieces of that system; its harder-to-copy asset is operating history. A simulation can model a jam. Twenty-five years of live baskets tells you which jam ruins Friday dinner.

“We have largely completed a very significant phase of investment in our robotics and automation capabilities.”Tim Steiner, reporting FY2025 results

The £88 million alarm bell

What failed first was painfully literal. In February 2019, an electrical fault in a charging robot at Ocado's Andover CFC started a fire. The building, machinery and stock were destroyed. Ocado recorded £111.8 million of costs and recognized £23.8 million of insurance proceeds that year, leaving a net exceptional charge of £88 million. The incident exposed how tightly coupled a centralized, automated operation can be: one site represented about a tenth of fulfilment capacity.

Ocado did not abandon the grid. It changed the engineering brief. Rebuilt systems added fire detection and suppression measures, while the newer product family attacked weight, construction time, human handling and flexibility. The lesson was not merely to add sprinklers. A machine expected to run a country's dinner plans needs graceful failure, serviceable parts and spare capacity designed in.

A robot goes on a diet

Ocado Re:Imagined, unveiled in 2022, is the clearest evidence of the change. Its 600 Series bot is roughly five times lighter than the previous generation, with more than half its parts 3D printed. A lighter bot needs a lighter grid; a lighter grid uses less material and can be installed in weeks instead of months. This is not cosmetic engineering. It reduces the capital and time required before a retailer earns its first delivery fee.

On-Grid Robotic Pick adds arms with computer vision, sensing and reinforcement learning. Their job is comically difficult: identify and grip tens of thousands of unfamiliar objects, then pack them densely without turning yogurt, herbs and watermelon into soup. By July 2026, the arms had made more than 350 million picks across 14 CFCs and eight partners. At the Luton site, roughly half of order volume was being picked robotically across a 45,000-item range.

Automated Frameload handles the heavy totes that leave for vans. Swift Router holds scheduled orders in the grid, combines them with later two-to-six-hour orders and sends both on denser routes. Ocado reports that advanced sites can reach 40% same-day volume; one 40-item basket went from click to delivery in 73 minutes. Meanwhile, In-Store Fulfilment and Store Based Automation let retailers use existing shops and dark stores. The once-monolithic pitch is becoming a menu.

The warehouse cannot invent customers

North America supplied the second hard lesson. Kroger's 2018 agreement once contemplated 20 automated sites. Eight were built. In late 2025, Kroger chose to close three CFCs and take a $2.6 billion charge across its automated fulfilment network, arguing that stores could serve most customers faster and more cheaply. Sobeys also closed an underused Canadian site. Ocado received closure payments, but the strategic message was unmissable: excellent automation cannot compensate for a catchment with too few orders, too much driving or abundant stores near shoppers.

That changed the sales posture. Ocado and its partners concentrated volume in stronger sites, allowed exclusivity to end in most markets, narrowed research spending and pursued £150 million of cost reductions. It also emphasized formats that do not require every customer to begin with a giant greenfield CFC. In the first half of 2026, underlying revenue excluding closure effects grew just 1%, and underlying cash flow remained negative £147 million. This is still a capital-disciplined prove-it year, not a victory lap.

There are signs the broader pitch travels. Auchan's Warsaw CFC opened in 2025 only 12 months after ground was broken. In 2026, Lotte opened a Busan CFC with 600 Series bots and an automated freezer built for dawn delivery. Asda chose OSP software for more than 700,000 weekly ecommerce orders across stores and dark stores, with a rollout planned from 2027. Danish pure-play grocer nemlig became the newest European partner. Coles said its two Australian CFCs reached positive EBITDA in their second year while supermarket ecommerce sales rose 26.4%.

What operators can borrow

The copyable part of Ocado is therefore less glamorous than the grid. Own the problem end to end. Instrument it. Simulate failure. Design around the ugliest objects, not the demo-friendly ones. Match fixed capital to proven demand, and give customers a path that starts smaller than your architectural ideal.

The conditions matter. A centralized CFC struggles where population is scattered, online share is low, labor is inexpensive, stores already sit close to customers or shoppers expect delivery in minutes. A retailer with weak data, volatile stock records or too little organizational patience can turn integrated software into integrated confusion. Ocado's newer options acknowledge these limits. The company that once asked retailers to buy the whole machine now has to show that its pieces create value separately, then compound when joined.

From spectacle to utility

Ocado remains unusual because it has been the retailer, the warehouse operator, the software vendor and the robot maker. That breadth has produced both its moat and its bills. FY2025 adjusted EBITDA reached £178 million, yet underlying cash outflow was £213 million. The group says it can generate positive cash flow in the second half of 2026 and for the full 2027 financial year. Investors have heard versions of the destination before. This time the evidence will come from partner economics, not a beautiful video of bots.

The robots will keep whizzing; that is their job. Ocado's job is now quieter: make a grocer's existing network more useful, deploy faster, charge sensibly and prove that a basket earns its journey. If it succeeds, the grid becomes infrastructure. If it does not, it remains one of industrial technology's most watchable spectacles.