Key takeaways

  • Flipkart pilots food delivery in Bengaluru around 15 August 2026 on ONDC at about 10 per cent commission, against the 16 to 30 per cent incumbents charge.
  • Rapido Ownly went citywide in Bengaluru in March 2026 at zero commission and now claims close to 10 per cent of the city market.
  • Bengaluru restaurant bodies say a restaurant billing Rs 1 lakh through an app keeps about Rs 40,000, and have called a boycott from 15 August.
  • Eternal booked Rs 3,100 crore of food delivery revenue in the June 2026 quarter, up 37 per cent, adjusted EBITDA Rs 606 crore. JM Financial pegs the take rate near 21.5 per cent.
  • Swiggy lost Rs 791 crore at group level on revenue of Rs 6,812 crore in the same quarter. It has no profit cushion for a price fight.

Order growth was never the question in Indian food delivery. The question was whether two players could keep raising what they charge restaurants without anyone undercutting them. That changed this quarter. Flipkart comes in near 10 per cent, Rapido at zero. Neither has to beat Zomato on product. They only have to be cheap enough that an angry restaurant lists on two apps at once. When a challenger runs at 10 per cent, the incumbent margin is the thing being competed away.

The attack is on the take rate, not the market

Rapido Ownly, Bengaluru0%Flipkart on ONDC, pilot10%Swiggy and Zomato16-30%0%10%20%30%Commission charged to the restaurant, per cent of order value
What the platform charges the restaurant — Three players, three price points. Swiggy and Zomato quote one 16 to 30 per cent band, drawn here as a single range rather than two separate prices. Flipkart pilots in Bengaluru around 15 August 2026
Source: Business Today, Inc42, Medianama

Flipkart builds on ONDC, so the network layer is shared and onboarding cost is not its own to carry. Rapido uses bike taxi riders already on the road, which makes the marginal cost of a food run genuinely low. That is a structural advantage, not a discount, and it is why I do not dismiss this the way the market dismissed Uber Eats in 2020.

Where Rs 100 of an order actually goes

Order value: Rs 10040%Restaurant keepsRs 40 (40%)25%Platform commissionRs 25 (25%)20%Restaurant-funded discountsRs 20 (20%)15%Ads, gateway, packagingRs 15 (15%)
Where Rs 100 of a food order goes — Modelled. The Rs 40 residual and the 25-35 per cent commission are reported; the rest is illustrative
Source: Bengaluru restaurant bodies via Business Standard; NRAI

Treat the Rs 40 carefully. It is not commission. Restaurant bodies get there by adding commission, the discounts the platform expects the restaurant to fund, ad spend, gateway charges and packaging. NRAI puts headline commission at 25 to 35 per cent, platforms describe 16 to 30. The all-in number is higher than either, which is why the anger is real even though headline commission has barely moved.

The mechanism

1Challenger lists at 10 per centIt only needs restaurants angry enough to run two apps.2Incumbent defends with discountsA commission cut is visible. Funding more discount is not.3Take rate holds, margin does notAdjusted EBITDA per order compresses before the take rate moves.
How a cheap entrant squeezes an incumbent — The mechanism, each step visible in quarterly disclosure
Source: Author’s analysis

This is the part investors miss. A commission cut is a headline. Funding more discount is a line item. Eternal can protect a 21.5 per cent take rate for several quarters while food delivery adjusted EBITDA per order erodes underneath it. Watch the rupee EBITDA, not the percentage. It is the same trap as chasing order frequency in quick commerce, where negative unit economics scale with volume.

Why Magicpin has not dented them

Here is the honest counter-argument. Magicpin is the largest food delivery app on ONDC: 1.5 lakh orders a day, above 10 per cent share in Delhi and Bengaluru, 70,000 restaurants, 15 million users. Two years in, Swiggy and Zomato have not lost a quarter of growth to it.

Copied in a quarterNot copied in a quarter0 per cent commission, set on day one297,000 Zomato partners, all India, FY25ONDC rails, so no network to build70,000 Magicpin partners, all IndiaBike taxi riders already on the road20,000 Ownly restaurants, Bengaluru onlyAn estimate that holds, a refund when it fails
Price is a launch decision. Distribution is inventory. — The three partner counts sit on different footprints and are not a ranking: two are national, one is a single city
Source: Eternal FY25 annual report; Business Standard; Inc42

Those three counts are not a league table and I will not present them as one. Zomato reported 297,000 active restaurant partners across all of India in FY25, Magicpin claims about 70,000 nationally, and the 20,000 on Ownly are in Bengaluru alone. Put them on one axis and you have measured nothing. The comparison that does hold is that a challenger sets its price on day one and buys its coverage over years. Distribution and reliability are the moat, not price. A user opens one app and expects every restaurant, an estimate that holds, and a refund when it does not. Rapido has cancelled rides on me repeatedly, and that is the failure mode of any model running on borrowed supply. Price buys a trial. Reliability buys the second order.

How this built, month by month

Aug 2025Rapido starts testing food delivery in BengaluruMar 2026Ownly goes citywide: zero commission, 20,000 restaurantsJul 2026Ownly claims nearly 10 per cent of Bengaluru; boycott called for 15 AugAug 2026Flipkart pilots food delivery on ONDC at about 10 per cent
Twelve months that repriced distribution
Source: TechCrunch, Medianama, Inc42, Business Today

What has to be true for the take rate to hold

One. The challengers stay single city. If neither Flipkart nor Ownly holds double digit share in a second city by mid 2027, the moat argument stands.

Two. My call: Eternal does not cut headline commission in FY27. It funds more discount instead. That is falsifiable either way.

Three. Restaurants do not go direct at scale. If the 15 August action collapses in a week, the leverage story is dead for two years.

Swiggy is the more exposed name. It is still loss making, its quick commerce arm only hit contribution break even in May 2026, and it is raising money into a market where India took just 5.2 billion dollars of a record global venture year. Eternal has a cushion. Swiggy does not.

What to do about it

If you hold either stock, stop leading with gross order value. Track take rate on net order value and food delivery adjusted EBITDA in rupees, every quarter. Volume growth with a softening take rate is a warning, not a good quarter.

If you run a restaurant, do the arithmetic before joining a boycott. Work out what a direct channel costs per order once you pay a rider and a gateway. Most single outlets find the platform still cheaper. Chains with their own demand hold the real leverage.

If you are building here, the lesson is adjacency, not discounting. A company with an existing fleet or an existing app enters food delivery at close to zero marginal cost. The reverse also holds: adjacency fails when nothing expensive transfers, which is what happened when Pocket FM tried to bolt a micro drama product onto an audio business.

Read next: the dark store profit and loss that explains why Zepto deferred its IPO, and why a record global venture year never reached India.

Sources

  • Business Today and India TV News, Flipkart food delivery on ONDC at 10 per cent, July to August 2026
  • Inc42, Bengaluru restaurant pushback, NRAI commission range and Ownly share, 30 July 2026
  • Medianama, Storyboard18 and TechCrunch, Rapido Ownly pilot and citywide launch
  • Business Standard and Deccan Herald, the 15 August deadline and the Rs 1 lakh to Rs 40,000 claim
  • Republic World and Business Standard, Eternal Q1 FY27 and the JM Financial take rate estimate
  • Inc42, Swiggy Q1 FY27 results; Eternal FY25 annual report for restaurant partner count

Figures are as reported by the sources named above at the time of writing.

Get the day’s top stories in your inbox

One concise email. No spam, unsubscribe anytime.