How to start a cloud kitchen

Why start a cloud kitchen instead of a dine-in room?

Because a cloud kitchen deletes the two most expensive things about a restaurant: the dining room, and the months you wait for it to fill. No high-street frontage, no interiors, no front-of-house payroll, no 60-cover fit-out. What you keep is the kitchen. What you inherit is delivery economics, and that trade decides everything else in this guide.

Read the last row twice. A cloud kitchen does not escape the cost of reaching customers, it converts that cost from fixed rent into a percentage of every single order. A dine-in owner pays for the room whether it fills or not; you pay the aggregators whether the order was profitable or not. The operators who make cloud kitchens work treat that commission line as the enemy from day one, not as a fact of life.

What does a cloud kitchen cost in 2026? The ₹5–15 lakh budget, itemised

Honest ranges below, quotes vary widely by city, menu and how much refurbished equipment you are willing to buy. Get three quotes for every line and treat the top of each band as a metro number.

Sum the minimums and you land just under ₹6 lakh; sum the maximums and you overshoot ₹15 lakh, which is exactly why the honest headline is ₹5–15 lakh, not a single number. Nobody hits every maximum, and most launches settle in the middle of the band.

Three rules survive every budget. First, the deposit is not the budget, a ₹40k/month kitchen with a six-month deposit eats ₹2.8 lakh before you own a single burner. Second, hold roughly 20% of the total as working capital; cloud kitchens die of cash flow in month two, not of bad biryani. Third, do not economise on the photoshoot, customers cannot smell your kitchen, they can only see it.

Which licences does a cloud kitchen need?

The good news: the stack is shorter than dine-in. No liquor, no music licence, no signage permission, usually no eating-house licence. The core four:

1. FSSAI licence, a state licence for most cloud kitchens. Aggregators verify the FSSAI number before your listing goes live, so this is the true gate on your launch date. File it the week you sign the lease.
2. Municipal trade licence, from your city corporation. Timelines vary by city; start it in parallel with FSSAI, never after.
3. GST registration, mandatory from day one if you sell on aggregators, regardless of turnover.
4. Shops & Establishment registration, the state labour registration, needed once you hire and often for the current account.

A fire NOC may apply depending on the premises, commercial kitchens with large LPG banks attract attention, so ask before the landlord's last tenant's history becomes your problem. The full stack, issuer by issuer, is in the 12-licence hub.

Where should the kitchen be?

You are not buying footfall, so stop thinking like a dine-in scout. You are buying delivery-radius density: the number of households and offices reachable inside the platforms' delivery window, divided by the rent you pay to reach them. In practice that means a back lane or first floor 500 metres from the expensive market, same radius, a fraction of the rent.

The checklist, in order of how often each one kills a kitchen: residential plus office density within the delivery radius; rent inside the ₹25–60k/month band (paying high-street premiums for a kitchen nobody visits is the classic first-timer error); adequate power and water, in writing; ventilation and neighbours who will tolerate a tandoor at 11 pm; and clean rider access, if pickup is painful, your preparation times look bad to the algorithm and your ranking pays for it.

One more input most first-timers skip: open the aggregator apps at your shortlisted address at 8 pm and count competitors in your cuisine. A dense zone means demand exists, and that you will need a sharper brand to rank in it.

What will Zomato and Swiggy really take from every order?

More than the commission slide says. The quoted base commission is only the start, add the per-order platform fee, payment gateway charges, GST on the platform's services, visibility ads and your share of co-funded discounts, and restaurants effectively surrender 25–35% of order value. On a ₹300 delivery order, ₹75–105 leaves before you have paid for a single ingredient. The full fee anatomy, and nine ways to claw it back, is in the commission playbook.

Run your own numbers before you sign the onboarding forms, not after:

The counter-strategy is channels, not complaints. ONDC runs at 3–5% commission, treat it as a margin-rich third channel, not a replacement for aggregator volume. And WhatsApp/QR direct ordering runs at 0%: every repeat customer you move there is the single highest-margin act available to a cloud kitchen. Print the QR on every box from order one. The customers already know your food; the box is the only marketing surface you own.

How does the multi-brand playbook actually work?

One kitchen, one team, one inventory, three brands on the apps. The biryani brand, the curry-bowl brand and the rolls brand can share a hot line, 70% of a purchase list and a single evening prep. Each brand gets its own listing, its own photos and its own shot at category rankings, so the same fixed cost gets amortised across more order volume. This is not a trick; it is how most profitable cloud kitchens are actually structured, and it is why the format can reach profit at volumes a single menu never would.

The trap is accounting, not cooking. When three brands share one stock of chicken, rice and oil, most operators lose the ability to say which brand makes money, and a brand that loses ₹15 on every order scales its losses beautifully. The fix is recipe-linked, per-brand costing: every dish mapped to ingredient quantities, every order depleting shared stock, every brand throwing its own P&L. CountStand is an AI-native restaurant operating system for India, offline-first billing, KDS, inventory, GST & compliance, and an autonomous AI manager, in one platform, from ₹999/mo per outlet. Its cloud-kitchen mode runs multiple brands from one screen with shared recipe-linked inventory, so per-brand food cost is a report you read, not a guess you argue about. Book a demo before you design brand number two.

Packaging, ratings and the wrong-item tax

Your packaging is your dining room, it is the entire physical experience of your brand. Leaky dal and soggy rolls do not get second chances; they get one-star ratings, and ratings drive ranking, which drives the order volume your break-even depends on. Travel-test every dish in its actual container for 30 minutes on a bike before it goes on the menu. If it does not survive, re-engineer the dish or the container, not the review section.

Then there is the wrong-item tax: missing items and wrong orders are refunded out of your payout, and repeat offenders sink in the rankings. The fix is boring process, a packing checklist read against the kitchen ticket, one person sealing every order, and a kitchen display rather than paper slips so nothing gets lost between fryer and bag. Boring is the point; the winners in this format are the operators who make fewer unforced errors per hundred orders.

When does a cloud kitchen break even?

The model is simple: fixed costs (rent, salaries, subscriptions, that ads kitty) divided by what each order actually contributes after food cost and commission. Two things surprise first-timers, how much the 25–35% commission compresses per-order contribution, and how violently the answer improves as direct-channel share grows. Run it honestly, with commission included in variable cost:

Model three scenarios before you commit capital: all-aggregator, 20% direct, 40% direct. The spread between the first and third is usually the difference between a kitchen that grinds and one that compounds. Put the resulting numbers straight into your business plan, a lender or partner will ask for exactly this page, and sanity-check the output against what margins actually look like by format.

The 45-day launch checklist

Days 1–7, paper. Lease signed and registered; FSSAI, trade licence, GST and Shops & Establishment all filed; kitchen layout frozen; menu engineered for travel, not for a dining room, short, high-margin, and every dish container-tested.

Days 8–21, build. Equipment ordered against the frozen layout; branding and photoshoot done; packaging ordered after travel tests; POS, KDS and inventory stack configured with every recipe costed per dish; aggregator onboarding submitted the moment the FSSAI number arrives.

Days 22–35, people and process. Cooks and packer hired and trained on the actual systems; recipes standardised with yields; supplier accounts opened with two vendors per critical ingredient; test orders through the full flow, app to KDS to sealed bag.

Days 36–45, launch. Listings live; the launch-ads kitty deployed for the first ratings; QR-to-WhatsApp inserts in every single order from day one; and a daily day-close that ties orders, payouts and stock together, because in this business, the operator who knows their numbers nightly beats the one who finds out monthly.