Most dark stores should never have been built

A data-led operator's view on quick-commerce unit economics · July 2026

Unpopular opinion from someone who runs quick commerce for a living: most dark stores should never have been built. I say this as someone who has personally launched twenty of them — every one inside an existing hypermarket, which turns out to be the entire point.

What follows
  1. The graveyard, priced
  2. The product was never the problem
  3. Didn't India crack it?
  4. The number that should worry everyone
  5. Meanwhile, the boring incumbents
  6. What this looks like in the Gulf
  7. So are dark stores dead?
  8. The operator's test
  9. What is not public
  10. Sources

The graveyard, priced

The money that went in is easier to count than the money that came out. PitchBook put $39.3 billion into foodtech in 2021, of which more than $10 billion went to grocery startups alone.[1][2] Very little of the instant-grocery share of it came back.

Getir is the headline case. It raised a $768 million Series E at an $11.8 billion valuation in March 2022.[38] It exited France, Italy, Spain and Portugal in 2023, then announced its withdrawal from the US, UK, Germany and the Netherlands in April 2024, winding down the German business by mid-May and laying off around 1,800 warehouse staff.[3] In February 2026 Uber agreed to buy the Turkish food-delivery business for $335 million cash, plus $100 million for a 15% stake in the grocery, retail and water arm.[4][5]

Worth pausing on that last transaction, because it has been widely misreported. The circulating "$435 million" headline is the sum of two different things: an outright purchase of the food-delivery business and a minority stake in the grocery business. Uber did not buy Getir's grocery operation. From $11.8 billion to that, in under four years.

Gorillas reached a $3 billion valuation in 2021[6] and was acquired by Getir in December 2022 in a deal reported at $1.2 billion — much of it in stock, and the brand was shut down entirely eighteen months later.[7][3] Jokr hit a $1 billion valuation eight months after launch, exited the US in June 2022, sold its American assets to competitors, and later raised $50 million at a sub-billion valuation.[8][9] Buyk and Fridge No More both ceased operations in the same week of March 2022.[10]

China ran the same experiment with more capital and more discipline. Missfresh listed on Nasdaq in June 2021, raising $273 million at a valuation of around $3 billion.[39] It lost RMB 2.91 billion, RMB 1.65 billion and RMB 3.85 billion in 2019, 2020 and 2021 respectively.[11] It received a minimum-bid-price deficiency notice from Nasdaq in June 2023 and a delisting notice in November 2023.[11][40]

Peak valuation vs. what it was eventually worth (US$bn) Getir 11.8 0.335 — Türkiye food delivery sold to Uber, 2026 Gorillas 3.0 1.2 — largely stock; brand later closed Missfresh ~3.0 delisted from Nasdaq, November 2023 036912 Peak valuation Outcome
Peak valuations are drawn to scale. Outcomes are not equivalent events — a business-line sale, a mostly-stock acquisition and a delisting — and are shown for magnitude, not as a like-for-like comparison. Sources [3], [4], [6], [7], [11], [38], [40].

The product was never the problem

Here is the thing nobody wanted to admit at the time: customers loved fifteen-minute groceries. They still do. Demand was never the failure point. Rent was.

A dark store opens with three liabilities on day one — a lease, a fit-out cost, and zero customers. Every order it will ever fulfil has to climb out of that hole. When the cost of capital was near zero, you could pretend the hole did not exist. When rates went up, the hole ate the industry.

That is not a story about execution. Getir and Gorillas were not badly run in any obvious operational sense; they hit their delivery promises far more often than they missed. They were structurally short of a fixed-cost base that somebody else was already paying for.

Didn't India crack it?

India is the counter-argument everyone raises, and it deserves a serious answer — because in one important case, the answer is yes.

Blinkit has genuinely turned. In the quarter to June 2026 it reported net order value of ₹17,132 crore, up 86% year on year, from 2,443 dark stores, with adjusted EBITDA of ₹102 crore — positive 0.6% of NOV, its fifth consecutive quarter of margin improvement.[12][13] Five quarters earlier that number was negative 1.8%. This is a real operating turnaround at real scale, and anyone arguing dark stores can never work has to explain it.

Blinkit: adjusted EBITDA as % of net order value 0 +1.00.0−2.0 −1.8%−1.3%+0.03%+0.3%+0.6% Q1 FY26Q2 FY26Q3 FY26Q4 FY26Q1 FY27 1,544 stores1,8162,0272,2432,443
Adjusted EBITDA as a percentage of net order value, five quarters to June 2026, with dark-store count beneath each period. Sources [12], [13], [14], [15], [16].

But read the rest of the disclosure, because the same results contain three numbers that describe exactly how narrow the path is.

Capex per store went up 150%. On the Q1 FY27 earnings call, Blinkit raised its steady-state capital-expenditure guidance per dark store from ₹1 crore to ₹2.5 crore.[13] It simultaneously raised its steady-state assumption for net order value per store per day from ₹7 lakh to ₹11 lakh. In other words: the stores now have to do substantially more volume, because they now cost substantially more to build. The model did not get easier as it scaled. It got heavier, and the throughput assumption had to be raised to match.

Shrink is 1.8% of NOV. That is roughly ₹308 crore in a single quarter, roughly three times the adjusted EBITDA it earned in the same period, and management attributed it largely to perishables — fruit and vegetables.[13][17] This is the cost line that store-based fulfilment structurally attacks, because walk-in customers keep stock turning. A dark store has no second channel to sell down its ageing produce.

Average order value is flat. Net AOV was ₹518, against ₹521 a year earlier.[13] Nearly all of that 86% growth is volume and store count, not basket economics improving.

And Blinkit is the best of them

The comparison is what makes the point. Swiggy Instamart, in its most recent published quarter, ran a contribution margin of negative 1.8% of gross order value, improving to negative 1.1% on a March exit rate, with adjusted EBITDA of negative ₹858 crore.[18] Zepto lost ₹5,905 crore in FY26 on revenue of ₹22,623 crore, with its best quarter still at negative 15.3% adjusted EBITDA as a share of net revenue[19][20] — against management guidance a year earlier of reaching EBITDA profitability by exactly that quarter. That is a miss of roughly 1,500 basis points against its own public target.

OperatorDark storesMarginBasis / period
Blinkit2,443+0.6%Adj. EBITDA % of NOV, Q1 FY27
Swiggy Instamart1,143−1.8%Contribution margin % of GOV, Q4 FY26
Zepto1,139−15.3%Adj. EBITDA % of net revenue, best quarter FY26

These are not like-for-like margin definitions — see What is not public. Sources [12], [13], [18], [19], [20].

Blinkit needed roughly 2,400 stores and four years of investment reported at around ₹3,000 crore[13] to reach a margin of 0.6%. That is the industry's success story. It is a genuine achievement and a very thin reward, and it is available only to an operator with the balance sheet to fund a decade of losses first.

The number that should worry everyone

Here is the finding that changed how I read this whole sector.

Between April and July 2026, India's five major quick-commerce platforms opened roughly 900 new dark stores. Those 900 stores added 152 new pin codes.[21][22]

Bernstein's own framing of what that bought is the sharper number: the share of metro pin codes served by all five players rose from 26% to 44% in a single quarter.[22] The stores were not opening where quick commerce was absent. They were stacking on top of each other where it was already present.

Adding stores is no longer adding coverage (India, April–July 2026) Dark stores added 900 New pin codes served 152 Metro capacity Stores in metros ~4,300 Est. profitable capacity ~3,600
Roughly 700 metro dark stores above estimated profitable capacity — a modelled figure, not observed closures. Sources [21], [22], [23].

The same analysis put around 4,300 stores in India's metros against an estimated profitable capacity of about 3,600 — roughly 700 stores of oversupply.[21] An earlier note in April 2026 found that 100% of metro pin codes already had quick-commerce coverage, and nearly 80% were served by three or more players.[23]

This is what late-stage dark-store expansion actually looks like. It is no longer reaching new customers. It is subdividing existing ones, and the marginal store is increasingly cannibalising the network's own volume rather than adding to it.

The gap worth staring at Analysts model roughly 700 metro stores of oversupply. No Indian grocery operator has disclosed a single dark-store closure. Modelled overcapacity and zero reported rationalisation cannot both be right indefinitely. One of those two numbers is going to move.

There is one honest counter-signal, and it comes from the operator with the weakest margin. Swiggy added a net seven Instamart stores in its most recent quarter — effectively a full stop on expansion. On the earnings call management said it did not "see the necessity to add stores… over at least the next few quarters considering the current utilisation," and described densifying only once a store approaches 80–85% of capacity.[41] That is the correct response to the densification data. It is also, tellingly, the operator that has the least room to be wrong.

Meanwhile, the boring incumbents

While the pure plays were raising and dying, the supermarkets were quietly doing the same job out of buildings they already owned.

Tesco Whoosh grew sales 51% to over £400 million in the year to February 2026, reaching 73% of UK households, and added a further 34 large stores to the service in the following quarter.[24][25] Every one of those is an existing shop that now also picks rapid orders. Tesco's UK online business overall reached £7.5 billion, 14.3% of UK sales.[24] No new leases were signed to build any of it.

Kroger ran the natural experiment in the other direction. In November 2025 it announced the closure of three Ocado-powered automated customer fulfilment centres, took an impairment of roughly $2.6 billion, and said it expected about $400 million of e-commerce profitability improvement in 2026 — explicitly pivoting toward capital-light, store-based automation.[26] In the quarter to May 2026 it reported e-commerce sales up 19% with improved profitability.[27]

Ahold Delhaize reported that more than 90% of its online sales are now fulfilled through same-day options, with US online sales up 14.3% in constant currency.[28]

None of this is glamorous. All of it is being paid for by rent that walk-in customers already cover.

What this looks like in the Gulf

The region I work in is running a compressed version of the same cycle, and the published numbers are unusually clear about the trade-off.

talabat — the largest quick-commerce operator in the GCC — reported first-quarter 2026 GMV of $2,685 million, up 19%, of which $2,122 million came from the GCC. Adjusted EBITDA was $130 million, or 4.8% of GMV, down from 6.3% a year earlier.[29] In the same disclosure it set out a $120 million investment programme for 2026, roughly $75 million operating and $45 million capital and lease costs, of which dark-store density and supply chain for talabat mart is one of three named pillars, alongside its subscription programme and new retail offerings.[30]

Read those two facts together. The margin compressed by 150 basis points in the year the company started spending heavily on dark-store density. That is not a scandal — it is a deliberate investment decision, and grocery and retail GMV grew 47% in 2025 to $2.8 billion, 29% of the platform total.[31] But it is a clean, publicly disclosed illustration of the trade: dark-store density is bought with margin.

Not everyone is buying. Careem wound down most of its consumer services in Saudi Arabia in May 2026, including what its co-founder described as a "strategic pause of our Quick Commerce verticals in KSA" — roughly thirteen months after launching grocery in Riyadh.[32] Rabbit, the Egyptian operator, expanded into Riyadh in 2025 and subsequently exited.[33] Meanwhile Amazon Now launched in the UAE in October 2025 with a fifteen-minute promise across 25 locations, and opened a third MENA market in Egypt in July 2026.[34][33]

And then there is the rent

The variable that killed the Western cohort is moving in the wrong direction here. Knight Frank's H2 2025 UAE industrial and logistics research recorded Dubai warehouse rents at AED 100 per square foot in Al Quoz and AED 58 in Dubai Industrial City — the latter up 32% year on year, the highest ever recorded there. Dubai South rose around 25%, JAFZA around 22%.[35]

The pairing that matters Dubai industrial rents rising up to 32% year on year, against a regional quick-commerce leader earning 4.8% of GMV at the EBITDA line. A model whose principal fixed cost is inflating six times faster than its margin has very little tolerance for error.

Knight Frank expects rental stabilisation only toward late 2026, with 6.6 million square feet of new Dubai industrial supply arriving that year.[35] Even if that holds, anyone signing a dark-store lease in Dubai today is underwriting a decade of cost at close to a cyclical peak.

So are dark stores dead?

No — and this is where the nuance lives.

Dark stores earn their place where there is no store nearby: catchments with real demand and no retail asset to piggyback on. Used that way, as infill rather than as the default, the dark store is a precision tool. Blinkit's results prove the model can clear breakeven with enough density, enough discipline and enough patience.

China offers the most instructive survival case, and it needs its caveat stated. Dingdong Maicai has now posted nine consecutive GAAP-profitable quarters and fourteen consecutive non-GAAP-profitable ones, with Q1 2026 revenue of RMB 5,892.7 million.[36] But roughly RMB 138 million of its RMB 165 million GAAP net income that quarter came from ceasing depreciation on assets classified as held for sale — because Meituan is acquiring its China business for $717 million.[36][37] The honest reading is not "profitable and independent." It is "survived long enough to be bought at a real price," which is a genuinely good outcome next to being written to zero, and a very different one from what investors underwrote.

What I would want to see a business case for is the other pattern: signing leases for sheds next to your own shops. If you already own the box, the shelf and the customer, building a second box across the street is not a strategy. It is a subsidy to your landlord.

The operator's test

Five questions, in order. If you cannot answer the first three with numbers, you are not ready to sign a lease.

  1. Do you already own retail assets in the catchment? Convert them first. Capex-light beats capex-heavy when the service level is identical — and the published evidence says the service level is identical. Tesco reaches 73% of UK households from existing shops.
  2. Is there demand density but no asset? That is a genuine dark-store catchment. Test it against the marginal case, not the average: what does the tenth-best hour of the week look like, not the best one?
  3. What does rent do to your cost per order at realistic volume — and what happens at lease renewal? Model it at the rent you will pay in year five, not year one. In Dubai that means underwriting rents that rose up to 32% last year.
  4. Is your catchment already served by three or more players? In India's metros, nearly 80% of pin codes are.[23] Adding a store there is a share fight against incumbents with a cost base you cannot yet match, not a growth opportunity.
  5. Is the dark store there because a competitor opened one? That is not a business case. That is FOMO with a lease attached.
Speed is a product feature. Rent is a business model. The winners of the next phase of quick commerce will be the ones who never confuse the two.

What is not public

An argument this dependent on numbers should be honest about where the numbers run out.

Nobody publishes cost per order. Not one operator in this article discloses it. Every margin figure here is a different construct — Blinkit's adjusted EBITDA as a percentage of net order value, Instamart's contribution margin as a percentage of gross order value, Zepto's adjusted EBITDA against net revenue, talabat's adjusted EBITDA against GMV. These are not interchangeable, and the comparison table above should be read as an ordering, not a like-for-like ranking.

Store-based operators disclose least of all. This cuts against my own case. Tesco publishes Whoosh sales growth but not Whoosh unit economics; no European grocer publishes a per-order cost split by pick method. Kroger's $400 million improvement figure is the closest thing to a clean read on store-pick versus warehouse-pick, and it is a forward guidance number, not an audited outcome. If store-based fulfilment is as structurally superior as I believe, the people running it have not proven it in public either.

The densification figures are broker estimates. Bernstein's store counts, pin-code coverage and profitable-capacity numbers are modelled, and I have read them via press summaries rather than the notes themselves. The "700 stores of oversupply" figure in particular is an estimate of capacity, not a count of anything.

Some collapse-era figures are contested. Getir's total funding is reported at $1.8 billion by one source and $2.4 billion by another.[3][4] The Missfresh loss figures here are press restatements of SEC filings rather than the filings themselves. Where a number is doing real work in the argument, I have tried to use the primary disclosure; where I could not, I have said so.

And one correction to my own earlier writing: I have previously used a figure of $14 billion for quick-commerce venture funding. I could not source it, and I could not source a clean quick-commerce-only total for 2021 either — the published PitchBook figures cover foodtech and grocery startups more broadly, which is why those are the numbers used above. If you have seen the $14 billion figure attributed to a primary source, I would like the reference.

Read next Store-based quick commerce: what it is and when it beats dark stores — the reference explainer behind this argument. A working definition, how the cost structure decomposes into site, pick, drop and waste, what is genuinely hard about running it, a six-step sequence for converting an existing estate, and what breaks at scale.

Sources

  1. Food Dive, citing PitchBook: foodtech saw $39.3bn in VC investment in 2021.
  2. The Spoon, citing PitchBook: more than $10bn invested in grocery startups, 2021.
  3. tech.eu, "Getir is exiting Europe and closing Gorillas", 29 April 2024.
  4. TechCrunch, "Uber to buy delivery arm of Turkey's Getir", 9 February 2026.
  5. Turkish Minute, "Uber to acquire Getir's delivery operations in Turkey for $335 million", 10 February 2026.
  6. PitchBook, "Rapid grocery delivery startup Gorillas valued at $3B", 2021.
  7. CNBC, "Getir acquires embattled rival Gorillas", 9 December 2022.
  8. FreightWaves, "Ultrafast grocery firm Jokr pulling out of US", 17 June 2022.
  9. Locate2u, "Jokr raises $50m, valuation now under a billion dollars".
  10. Supermarket News, "Bankrupt Buyk sells itself off bit by bit", March 2022.
  11. Produce Report, "Missfresh mired in delisting risk as stock price plunges", June 2023.
  12. Inc42, "Blinkit's Q1 adjusted EBITDA rises to ₹102 Cr", 22 July 2026.
  13. MediaNama, "Takeaways from the Eternal Q1 FY27 earnings call", 24 July 2026.
  14. Inc42, Blinkit Q2 FY26 results, 16 October 2025.
  15. Inc42, Eternal Q3 FY26 results, 21 January 2026.
  16. Entrackr, "Eternal's reality check: Blinkit's thin margins", 29 April 2026.
  17. Apparel Resources, "Blinkit loses ₹308 Cr inventory, earns ₹102 Cr profit", July 2026.
  18. Swiggy Q4 FY2026 Shareholder Letter (primary), May 2026.
  19. Entrackr, "Zepto doubles revenue in FY26, losses widen to ₹5,905 Cr", 9 June 2026.
  20. Outlook Business, "How far Zepto fell short of its breakeven goals", 21 June 2026.
  21. Bernstein research on Indian dark-store saturation, summarised, 16 July 2026.
  22. Whalesbook, "Quick commerce firms open 900 dark stores in 3 months", July 2026.
  23. Business Standard, "India's quick commerce boom hits saturation in metros, says Bernstein", 10 April 2026.
  24. Tesco PLC Preliminary Results 2025/26 (primary), 16 April 2026.
  25. Tesco PLC Q1 Trading Statement 2026/27 (primary), 18 June 2026.
  26. Kroger, "Kroger evolves eCommerce offerings" (primary), 18 November 2025.
  27. Kroger First Quarter 2026 Results (primary), 18 June 2026.
  28. Ahold Delhaize Q1 2026 results (primary), 6 May 2026.
  29. talabat Q1 2026 results press release (primary), 12 May 2026.
  30. talabat Q1 2026 results presentation (primary), 12 May 2026.
  31. talabat Q4 2025 results press release (primary), 13 February 2026.
  32. fwdstart, "Careem is winding down most of its consumer services in Saudi Arabia", 3 May 2026.
  33. fwdstart, "Amazon brings 20-minute delivery to Egypt", 10 July 2026.
  34. Gulf News on the Amazon Now UAE launch, October 2025.
  35. Knight Frank, UAE Industrial & Logistics Report H2 2025, 9 February 2026.
  36. Dingdong (Cayman) Limited Q1 2026 financial results (primary), 21 May 2026.
  37. Caixin, "Meituan to buy Dingdong's China business for $717 million", 6 February 2026.
  38. Getir, "Getir raises $768 million in Series E funding at $11.8 billion valuation" (primary), March 2022.
  39. Nasdaq/Reuters, "Tencent-backed Missfresh raises $273 million in U.S. IPO", 25 June 2021.
  40. Missfresh, "Missfresh receives delisting notice from Nasdaq" (primary), 17 November 2023.
  41. Swiggy Q4 FY2026 earnings call transcript (primary), May 2026.

Dipankar Biswas leads quick commerce and online supermarket operations for a major grocery retailer in the UAE, running a multi-country, multi-channel fulfilment network across own rapid delivery, scheduled online grocery and marketplace partnerships. He previously built an integrated dark-store network across two countries, and led operations at noon and Swiggy. He writes about store-based quick commerce at dipankarbiswas.me.