Ghost kitchens and virtual brands: does the economics actually work?

A few years ago the ghost kitchen was going to eat the restaurant industry: no dining room, no servers, no prime retail rent — just a kitchen, the delivery apps, and pure margin. Then reality showed up. The pioneers downsized, pivoted to software, or closed, and 2023 got called "the year ghost kitchens died." That's too tidy — the model isn't dead, but the version that was sold to operators mostly was. If you're weighing a delivery-only play, here's the honest economics.

The definitions, quickly

• Ghost kitchen (dark kitchen): a delivery-only production facility with no dine-in — all resources point at the kitchen, delivery tech, and marketing.
• Cloud kitchen: roughly interchangeable, often the shared, multi-tenant version of the facility.
• Virtual brand: a delivery-only brand that exists only on the apps, frequently run out of an existing restaurant's kitchen under a different name and menu.

The pitch that's actually true: lower startup

The upfront-cost advantage is real. A ghost kitchen runs roughly $30,000–$200,000 to open — a shared commissary space at the low end, a custom independent buildout at the high end — versus $175,000 to $750,000-plus for a traditional restaurant, and $750k–$1M+ for a full dine-in build. Call it 30–50% less capital, because you're not paying for a dining room, front-of-house construction, or a prime retail lease. If the whole case were startup cost, ghost kitchens would have won.

The pitch that isn't: the margin gets eaten

Here's what the low startup cost hides. A ghost kitchen has no walk-in traffic — every order comes through a third-party app, which means the delivery commission isn't a cost on part of your business, it's a cost on all of it. At a 25% commission a well-run kitchen nets maybe 5–7% on a delivery order; at 30% that falls to 0–2%, and once processing, promotions, and refunds are counted the effective take often hits 30–40% of the order. Between commissions and the marketing you must buy to be visible at all, the platforms can consume 25–30%+ of revenue, permanently.

So the trade is stark: you save 30–50% on the one-time cost of opening and then pay 25–30% of every order, forever to the apps that own your customers. The low buildout is a one-time discount; the commission drag is a lifetime tax. That's the math that closed the pioneers.

The shakeout you should learn from

This isn't theory — it already happened, and the names matter:

• Kitchen United sold or closed its physical units and pivoted to software in late 2023, after shuttering all its Kroger locations.
• CloudKitchens facilities were reportedly running around half-empty; at five locations, 41 of 71 restaurants closed within a year — a ~58% failure rate.
• Reef was pared back by Wendy's for underperformance and wound down its ghost kitchens.
• The apps themselves culled the virtual brands: in some cities they'd become one in five delivery listings, so Uber Eats removed thousands of near-duplicates and DoorDash now requires virtual brands to be meaningfully distinct (50%+ menu differentiation, a minimum item count, a 4.0+ rating).

The lesson isn't "never" — it's which version failed. The standalone, dedicated ghost kitchen that depends entirely on paid app visibility had the weakest economics and the highest failure rate.

When it still makes sense

The model that survived is the capital-efficient add-on, not the standalone bet:

• Extending an existing kitchen's off-peak capacity. You already have the kitchen, the staff, and the food cost under control — a virtual brand run out of it during slow hours is incremental revenue on infrastructure you've already paid for. That's a very different equation from renting a commissary stall and starting from zero.
• Testing a concept cheaply before committing to a storefront lease.

The through-line: a ghost kitchen works when it rides on a business that already controls its costs and doesn't live or die by the app algorithm — the same "match the model to your economics" logic I applied to choosing a system by concept. It fails when the delivery commission is the business model.

Disclosure: I work at Katalyst, and we build the online ordering that lets a virtual brand take commission-free direct orders instead of handing every sale to an app — so read the bias in. But the economics aren't ours to spin: the low startup cost is real and the commission drag is realer, and the operators who make delivery-only work are the ones extending a kitchen they already run profitably, not the ones betting a standalone facility against a 30% lifetime tax. Run the margin math before the buildout math. The cheap part to open is the expensive part to operate.

Related Katalyst products

• See Katalyst online ordering