What is a dark kitchen: the before and after of the real 2026 numbers

What a dark kitchen is, and why its price looks nothing like a restaurant's

A dark kitchen is a production kitchen with no dining room and no walk-in service, where every sale arrives through aggregators, your own channel or the phone, and setting one up in 2026 runs between 8,000 and 45,000 USD depending on the format. Hidden kitchen, ghost kitchen, dark kitchen: three labels for one thing, used interchangeably. That the range comes out so wide is not vagueness, it follows from a structural decision. A rented box inside a hub starts near 8,000 USD because the hood, the grease trap and the electrical service arrive already solved; an empty warehouse with permits of your own climbs to 45,000 USD before you buy the first kilo of chicken. GlobeNewswire projects the global ghost kitchen market at 204 billion USD by 2030, and Statista counted more than 20,000 operating locations in the United States alone in 2023.

What each investment tier actually includes, as of September 2026?

Three tiers, and you read them by what sits inside rather than by the headline figure. The first, 8,000 to 14,000 USD, buys a 12 to 20 square meter box inside a hub:

shared hood and extraction, grease trap installed, the operator's umbrella license, while your own money covers the hot line, refrigeration and roughly 1,500 USD of smallwares. Step up to the second, 15,000 to 26,000 USD, and now you pay for three-phase electrical work, ducted extraction to the roof, a walk-in cooler and another 3,000 to 5,000 USD in permits and health approvals. Above 27,000 and up to 45,000 come two or three production lines for running several virtual brands, a blast chiller, cold room and multichannel dispatch software. Nobody skips a tier without paying for it later.

Five factors that move the price, with their impact measured in cash

Five variables explain almost the entire spread in these budgets, and kitchen equipment, the first thing everybody quotes, is not among them. Extraction rules: with no duct to the roof in the property, add 4,000 to 9,000 USD and two months of construction. Behind it comes electrical capacity, because moving from single-phase to three-phase with a new service line runs 2,500 to 6,000 USD across most Latin American cities. Health and fire permits swing between 800 and 5,000 USD depending on the municipality, and no line item delays an opening more. Each additional virtual brand asks for 2,000 to 3,500 USD of dedicated equipment: it multiplies the hot line, not the square footage. That leaves the lease deposit, three to six months of rent that almost nobody budgets.

The cost that never shows up in the quote and decides whether the model survives

The opening budget is the cheap part of this business, and that confusion sinks more hidden kitchens than any equipment mistake. An operator in Chapinero sent me his June 2026 P&L with one line on top: I sell 42 million and nothing sticks. He was selling. Except 26% went to aggregator commission, 11% to packaging plus the in-house courier on direct orders he handled badly, and 4.5% to a geotargeted campaign he switched on nine months earlier and never opened again. Building that kitchen cost 9,800 USD. Keeping it alive, far more than the plan admitted. The trade is structural: dropping fixed cost from the 32% to 40% a dining room carries down to the 14% or 18% of a box has a price, and that price is 21 to 34 fresh points paid to the channel on every order.

How much volume your delivery radius needs before the numbers close?

Without enough volume inside the delivery radius no cost structure rescues a hidden kitchen, and that calculation belongs before the lease.

At a 12 USD average ticket, 26% commission, 30% food cost and 16% fixed cost, break-even lands somewhere around 38 to 45 sustained daily orders, not the Friday peak. Users exist: Statista puts Latin America at 147 million online delivery users in 2026, and Rappi reported 35 million active users with 150 million downloads as of August 2024. Whether they order on your block is another matter. Should your radius produce only 22 orders a day, no commission negotiation repairs it: cutting 26% to 22% hands back roughly 10 USD daily. You are short 20 orders, not four points.

How to negotiate and cut the real cost, in that order?

Negotiate rent and deposit with the hub first, commission second, never the reverse: rent gets paid whether the sale happens or not, while commission only applies to money that already came in.

Hubs running below 70% occupancy will grant three months of stepped rent and a two-month deposit instead of six, provided you show up with a credible projected P&L. With aggregators, the discount does not come from asking. It comes from offering exclusivity on one of your virtual brands, or from entering their advertising program with a capped budget, which usually shakes loose two or three points. Build the owned channel from week one as well: Simon-Kucher & Partners warns that lost first-party data is a cost that surfaces late, the day you want to run a promotion and find your customer base living on somebody else's server.

What happens if the aggregator raises your commission two points next year?

Say you sell 30,000 USD a month with 78% of volume through an aggregator, and tomorrow the commission moves from 26% to 28%.

That is 468 USD less every month, 5,616 a year, against a business whose healthy net margin sits around 9% to 13%. One single move swallows 14% to 20% of your annual profit, and you have no table, no bar, no after-dinner dessert to cover it with. There sits the fragility this model hides: with no dining room absorbing margin, commission bites twice as hard. When 78% of revenue depends on a third party that sets the price of access and orders the storefront, you do not own a business, you own a contract-manufacturing deal with an app. Push that dependency down to 55% before somebody pushes it for you.

The diagnostic mistake that costs the most before you invest

A dark kitchen does not fix a badly costed menu, it accelerates it. I got this wrong for years: I recommended the format as an escape route for units already struggling with profitability in a physical location, and in several cases what I did was move the problem into a channel where the owner held even less control over price, over customer data, over the algorithm that orders the storefront. With food cost at 38%, delivery bills you for it inside the first two weeks, not in month six. Diego F. Parra reverses at Masterestaurant the order the market asks for: cost every dish first until food cost lands at 32% as a ceiling, then choose the kitchen format. Rebuilding your menu engineering costs about 400 USD. Committing to a warehouse, 20,000.

Five differences that reshape your cost structure, not just your storefront

The swap between fixed and variable defines everything else. Where the dining room carried 32% to 40% fixed cost against revenue, a well-built box drops to 14% or 18%; in exchange, 21 to 34 points of variable cost show up that were never in the P&L before. Whether the net result improves or worsens comes down to one variable: the volume your delivery radius can actually produce. You stop owning the customer relationship. On-table the guest is yours; inside an aggregator the guest belongs to the platform, and what reaches the kitchen is an anonymous order with no name and no phone number. Simon-Kucher & Partners has argued that losing first-party data is the strategic cost delivery operators discover late, once they want to build repeat purchase and have nobody to write to. Visibility is bought or earned, never free.

Five differences that reshape your cost structure, not just your storefront — in practice

In a physical unit, Google Business Profile and 5★ reviews bring traffic at no marginal cost; inside an aggregator, listing position answers to an algorithm rewarding prep time, acceptance rate, rating and, without much disguise, ad spend. The same dish sells three times more from position 4 than from position 19. The brand becomes elastic, and that opens an interesting door: one kitchen feeds three or four independent listings in the app, each with its own menu and audience. It is the commercial argument that lands best and also the one that has sunk the most operations, because multiplying brands without multiplying mise en place turns the kitchen into a bottleneck at 8 pm. Break-even is measured in orders per day, not covers. With a 15 USD ticket, 24% commission and 30% food cost, each order contributes roughly 5.10 USD; if your fixed costs add up to 4,200 USD a month, you need 28 orders a day to avoid losing money. That number —28— is what belongs painted on the kitchen wall, not the sales target.