Ghost Kitchens and Dark Stores: Where the Models Converge—and Where They Don’t

Last Updated on October 4, 2026 by Craig Allen Keefner

The first wave of ghost-kitchen hype is over.

A few years ago, it seemed like every restaurant-tech startup had a plan to put a shipping-container kitchen in a parking lot, load it with virtual burger brands, plug into DoorDash and Uber Eats, and scale to hundreds of locations. It was going to be cheaper than a restaurant, faster than a restaurant, and somehow more profitable than a restaurant.

That was the pitch.

Reality had other ideas.

Delivery commissions, food quality, customer discovery, labor, packaging, order accuracy, courier wait times, and basic unit economics all turned out to matter. A kitchen without a dining room still needs volume. A virtual brand still needs a real reason for customers to order it again. And an operator that depends entirely on third-party marketplaces is giving up a large part of the customer relationship.

That does not mean ghost kitchens disappeared. It means the business has moved past the easy hype cycle and into something more familiar: operators looking for ways to use capacity better, reduce delivery friction, and serve customers through a mix of delivery, pickup, drive-thru, lockers, and self-service.

Dark stores are part of the same broad fulfillment shift, but they are not the same thing.

Ghost kitchens make food. Dark stores pick and stage retail inventory. The operational details are different. The customer handoff problem is increasingly similar.

That is where kiosks, lockers, order-status systems, identity verification, and unattended pickup become relevant.

The Ghost-Kitchen Reset

A ghost kitchen is a food-production operation built mainly for off-premise orders. There may be no dining room, no counter service, and little or no walk-in traffic. Orders come through a restaurant app, web ordering system, marketplace, delivery platform, or in some cases a pickup workflow.

The early promise was straightforward: eliminate the expensive dining room, operate from lower-cost real estate, run multiple brands out of one kitchen, and let delivery platforms provide the demand.

The issue was that reducing front-of-house expense does not automatically create a good restaurant business.

Food still has to be prepared correctly. Packaging still has to hold up. Delivery drivers still have to arrive on time. Orders still have to be accurate. Brand discovery still costs money. A restaurant that appears only as one more icon in a delivery marketplace can have a hard time developing repeat business or controlling its customer relationship.

The more durable version of the model is usually less glamorous.

It may be an established restaurant using excess kitchen capacity for delivery-only menus. It may be a host kitchen preparing a partner brand’s menu. It may be a shared commercial kitchen serving caterers, food entrepreneurs, regional brands, or delivery-focused operators. It may be a restaurant that uses digital ordering, staging, and pickup technology to make an existing location more productive.

That is a very different model from “we will create 500 delivery-only restaurants from scratch.”

Four Market Examples

Kitchen United: The high-profile reset

Kitchen United became one of the most visible names in the shared-kitchen and virtual-brand business. It built delivery-oriented kitchen facilities and later expanded into food-hall concepts, including locations associated with Kroger.

The model ran into the same problems facing much of the early ghost-kitchen market: real-estate cost, operational complexity, demand generation, and the difficulty of making a standardized multi-brand kitchen platform work consistently across many markets.

Kitchen United closed its Kroger food-hall operations in late 2023 and shifted its stated focus toward software. In 2024, Sam Nazarian of C3 acquired selected Kitchen United assets and intellectual property for a new digital-restaurant effort.

The point is not that off-premise ordering went away. It did not.

The point is that the original “build a large network of shared kitchens and let virtual brands fill it” model proved harder than the investor decks suggested. The market did not vanish. It reset.

Nimbus: The more practical infrastructure model

Nimbus is a useful counterexample because it shows that commercial kitchen capacity can still be valuable even when the larger platform story fails.

Rather than trying to own the entire restaurant technology stack, generate demand, build virtual brands, manage delivery channels, and operate a national restaurant platform, Nimbus has concentrated on providing flexible kitchen infrastructure. It expanded by taking over former Kitchen United facilities that were already designed for commercial food production.

That is a more modest model. It is also potentially more durable.

There is a real market for permitted, equipped, flexible commercial kitchen space. Caterers, food producers, growing regional brands, delivery-focused operators, and specialty food businesses all need access to it. The operator does not necessarily need to pretend it has solved the entire restaurant business.

The practical lesson is simple: the kitchen asset can have value even when the original ghost-kitchen platform does not.

Kroger and Ocado: Dedicated grocery fulfillment

Kroger’s relationship with Ocado illustrates the more capital-intensive side of dark-store economics.

This is not a ghost-kitchen model. It is grocery fulfillment at scale. The work is not cooking food; it is managing inventory, picking individual items, handling substitutions, staging orders, loading delivery vehicles, and keeping the operation moving fast enough to support online grocery demand.

Ocado’s automated fulfillment systems are designed around those tasks. In 2024, Kroger ordered additional automation technologies for multiple fulfillment centers, including robotic-picking and automated loading capabilities.

The lesson is that dedicated fulfillment can make sense when there is sufficient order density, strong inventory discipline, capital to invest, and a real logistics operation behind it.

It is not a casual add-on.

Dark-store economics depend on volume. They depend on accurate inventory. They depend on delivery routing. They depend on labor productivity. And they depend on whether the operator can keep expensive automation busy enough to justify it.

Walmart: The store as fulfillment node

Walmart is probably the clearest example of the more scalable U.S. model: use the existing store network as a local fulfillment network.

The company reported pickup availability at roughly 4,600 U.S. stores and store-based delivery at about 4,300 locations. It has also developed automated fulfillment capabilities within selected stores to support faster pickup and delivery operations.

That matters because it avoids one of the biggest challenges of a standalone dark-store model: creating a separate real-estate footprint and a separate operating system before demand is proven.

A Walmart store already has inventory. It already has employees. It already has customer traffic. It already has parking, loading access, and brand awareness. It can become a distributed fulfillment node for pickup, delivery, returns, and local inventory without becoming a completely invisible warehouse.

That does not mean store fulfillment is easy. In-store picking can interfere with shoppers. Staffing is difficult. Substitutions are unpopular. Curbside handoff can become congested. But the basic model has scale because it uses assets that already exist.

Dark Stores Are Not Just Grocery Warehouses

A dark store is generally a retail facility that fulfills online orders without serving normal walk-in shoppers.

It may occupy a converted supermarket, an urban retail space, a micro-fulfillment facility, a backroom operation, or a dedicated customer fulfillment center. The layout is designed around picking and staging customer orders, not browsing, merchandising, or impulse buying.

That changes the operating equation.

In a conventional store, the retailer has to balance the needs of shoppers, merchandising, replenishment, checkout, customer service, and fulfillment. In a dark store, the operation can focus almost entirely on availability, pick paths, inventory accuracy, cold-chain management, order staging, courier loading, and pickup.

But a dark store does not automatically mean a robotic warehouse. Many are labor-driven operations. Some use automation. Some are built around delivery only. Others support customer pickup, curbside collection, or a combination of channels.

The right model depends on order density, geography, real estate, labor availability, assortment size, delivery radius, and how much customers are willing to pay for speed and convenience.

Where Ghost Kitchens and Dark Stores Actually Converge

Ghost kitchens and dark stores share a digital-first fulfillment architecture, but they solve different production problems.

Ghost kitchens have to manage:

• Food preparation and cook time.

• Menu availability.

• Order sequencing.

• Food quality and temperature.

• Packaging.

• Driver arrival and handoff.

• Peak-period capacity.

• Remakes, refunds, and customer complaints.

Dark stores have to manage:

• Inventory accuracy.

• Picking productivity.

• Product substitutions.

• Bagging and staging.

• Cold-chain handling.

• Delivery density.

• Courier routing.

• Pickup timing.

• Shrink, spoilage, and exceptions.

The overlap comes at the handoff layer.

In both models, somebody has to receive the order accurately and conveniently. That could be a delivery driver. It could be a customer at a pickup window. It could be a curbside associate. It could be a staffed service desk. Or it could be an unattended locker bank.

This is the part of the workflow that matters most to the self-service industry.

Where Kiosks and Lockers Fit

There is no reason to put a kiosk in front of every ghost kitchen or dark store. In many cases, the customer never comes to the facility. The order goes directly from the kitchen or fulfillment operation to a courier.

The opportunity appears when the operator wants customer pickup.

That may be because the operator wants to reduce last-mile delivery expense. It may be because customers prefer pickup. It may be because a suburban location has parking and easy vehicle access. It may be because a grocery or restaurant operator wants to extend pickup hours without keeping a full counter staffed. It may be because the location handles enough volume that a manual handoff becomes a bottleneck.

A locker-based pickup workflow can be straightforward:

1. The customer places an order through an app, website, restaurant platform, or retailer system.

2. The order is prepared and staged.

3. The customer receives a message that the order is ready.

4. The customer verifies identity using a QR code, order code, app credential, card, or other approved method.

5. A locker compartment or controlled pickup door opens.

6. The system records collection and flags orders that are late, incomplete, or require intervention.

The technology is not difficult to describe. The deployment is more complicated.

For groceries and prepared food, the real questions are:

• Is the order ambient, refrigerated, frozen, or hot?

• How long can it sit before quality becomes a problem?

• What happens if the customer is late?

• What happens if an item is substituted or missing?

• How are refunds, remakes, and age-restricted items handled?

• Is the system accessible to wheelchair users, people with limited reach, and customers using screen readers or other assistive technology?

• What happens when a locker door jams, the network fails, or the customer cannot authenticate?

• Who cleans, services, monitors, and replenishes the equipment?

• What does the labor model look like when exceptions happen?

This is where a lot of “frictionless” retail concepts run into the real world.

A locker can reduce routine handoff labor. It does not eliminate the need for operations.

The Real Self-Service Opportunity

The opportunity is not simply selling another kiosk or putting a locker wall outside a restaurant.

The opportunity is building a reliable customer-handoff system.

That system needs to connect order management, kitchen systems, retail inventory, delivery platforms, notifications, identity verification, accessibility, payment, temperature management, device monitoring, and exception handling.

For a restaurant, the goal may be to keep delivery drivers from crowding the counter while giving pickup customers a fast, predictable experience.

For a grocery retailer, the goal may be to support pickup outside normal staffed hours, reduce curbside congestion, or improve the efficiency of a high-volume staging area.

For an unattended-retail provider, the goal may be to turn a pickup location into a secure, self-service point that works for food, grocery, retail goods, returns, and possibly multiple merchants.

That is the direction worth watching.

Bottom Line

Ghost kitchens and dark stores are not interchangeable business models.

Ghost kitchens are about food production. Dark stores are about retail order fulfillment. Both are part of the broader move away from the traditional idea that every transaction must end at a staffed counter inside a conventional store or restaurant.

Their point of convergence is the physical handoff: getting the right order to the right person, at the right time, at the right temperature, with minimal labor and minimal customer friction.

That is where kiosks, lockers, pickup automation, order-status technology, and unattended collection become strategically important.

The hype phase may be over. The operational work is just beginning.