United States Quick Commerce & Instant Delivery Market Report, Size, Share and Forecast 2026–2033: USD 44 Billion, and the Seventeen-Billion-Dollar Purchase of a Shelf
Executive Summary
United States instant retail transacted USD 37.2 billion in 2025 and USD 44.0 billion in 2026. Pheonix Research forecasts USD 110.0 billion by 2033, a 14.51% CAGR, on orders rising from 496 million to 1,250 million at 12.25%.
Those two growth rates are the story, and they run the opposite way to the rest of the world. Value grows faster than volume in the United States, which means the basket widens: from USD 78 an order in 2026 to USD 88 in 2033. In the global market the basket is flat to the cent. The American consumer is not ordering more often so much as ordering more, and that is the single most important difference between this market and every other one in this report.
The market is store-picked, and that decides who wins
Only about USD 4.5 billion of the 2026 figure — roughly one dollar in ten, against one in seven globally — moves through a dedicated dark store. The rest is picked off a shelf that was already trading. Instacart, DoorDash, Uber Eats, Amazon and Walmart's on-demand services are all, in the main, demand and dispatch layers attached to somebody else's inventory.
That architecture is why the United States has produced the first genuinely profitable large instant-retail business. Instacart's Q3 2026 disclosure, for the quarter ended 30 September:
| Instacart, Q3 2026 | Value | Change YoY |
|---|---|---|
| Gross transaction value | USD 10,528m | +15% |
| Orders | 91.4m | +10% |
| Average order value | USD 115.19 | +4.8% |
| Total revenue | USD 1,095m | +17% |
| Advertising and other revenue | USD 331m | +23% |
| Adjusted EBITDA | USD 342m | +23% |
| GAAP net income | USD 147m | +2% |
A 10.4% revenue take on somebody else's groceries, a third of it advertising, converts into a 3.2% adjusted EBITDA margin on GTV and USD 307 million of free cash flow in a single quarter. No dark-store operator anywhere in the world has produced that combination.
And then the shelf was bought
On 29 October 2026, the same day it reported those numbers, Instacart agreed to be acquired by DoorDash in a cash-and-stock transaction with an equity value of approximately USD 17 billion. Against Instacart's annualised USD 1,368 million of adjusted EBITDA, that is 12.4 times run-rate earnings — a mature multiple for a company growing GTV at 15%.
The logic is not consolidation for scale. DoorDash owns restaurant-delivery demand and a courier fleet; Instacart owns grocery demand, retailer relationships and a retail-media business that grew 23% against transaction revenue's 14%. Neither can build the other's in a reasonable period. The deal is a statement that in a mature instant-retail market the asset worth owning is the shelf and the advertising attached to it, not the warehouse.
Why the American number is the weakest in the global build
Instacart's annualised GTV of USD 42,112 million is 95.7% of the entire modelled US instant-retail perimeter of USD 44.0 billion. That cannot be right as a matter of arithmetic if DoorDash, Uber Eats, Amazon and Walmart also operate in the market, and the resolution is a perimeter problem rather than a data problem. Instacart's GTV includes pickup orders and scheduled delivery windows that fall outside a two-hour promise. On our reading approximately USD 26 billion of that GTV is genuinely sub-two-hour delivery, with roughly USD 18 billion attributable to the other operators.
This is stated plainly because it is the largest reconciliation issue in the report and because it illustrates the sector's central analytical trap: the most-quoted number for a company is not the same perimeter as the market it is being compared to.
What the forecast assumes
That the store-picked architecture continues to dominate, that the DoorDash–Instacart combination completes, that retailer-owned apps continue to take share of their own delivery volume, and that dark-store operators remain a small minority. Under those assumptions the American market grows at 14.51% and its basket widens at 1.51% a year — the only large market in this report where value growth outruns volume growth.
The risk to that is not demand. It is that the retailers whose shelves are being monetised decide they no longer need the intermediary, which is the one substitution that would reprice the whole American market.
Market Overview and the American Perimeter
The United States is the second-largest instant retail market in the world and a market where the basket is still getting bigger. It is also the market where the industry's first genuinely profitable operator of scale emerged, and the two facts are related: the American market is not built on warehouses.
What the number covers
Instant retail here means retail goods ordered online and fulfilled from local inventory inside a sub-two-hour promise. That is a narrower perimeter than the gross transaction value most American platforms publish, and the difference matters more in this market than anywhere else in the world.
Instacart's annualised GTV, taken from its Q3 2026 disclosure of USD 10,528 million, is USD 42,112 million. This report sizes the entire United States instant retail market at USD 44.0 billion in 2026. Instacart's GTV alone is therefore 95.7% of the whole market, which cannot be true if DoorDash, Uber Eats, Gopuff, Amazon and Walmart are also trading in it.
The resolution is that GTV is not the instant perimeter. It includes pickup, and it includes scheduled delivery windows of a day or more. Those are real transactions on the same app and they are outside this report entirely. On our reading, roughly USD 26 billion of Instacart's GTV is genuinely sub-two-hour delivery, with the remaining USD 18 billion of the market attributable to the other operators.
| US reconciliation, 2026 | USD bn |
|---|---|
| Instacart annualised GTV (Q3 2026 × 4) | 42.1 |
| Less: pickup and scheduled volume outside the perimeter (Pheonix estimate) | −16.1 |
| Instacart inside the instant perimeter | 26.0 |
| DoorDash, Uber Eats, Gopuff, Amazon, Walmart on-demand (Pheonix estimate) | 18.0 |
| Total US instant retail, 2026 | 44.0 |
That table is the most useful thing in this report, because it converts the sector's most-quoted number into the number that is actually comparable to a market total. It also explains why published market sizes for the United States differ by a factor of two depending on whether the author counted GTV or orders fulfilled inside two hours.
The architecture split
Of the USD 44.0 billion, approximately USD 4.5 billion — roughly one dollar in ten — is fulfilled from a dedicated dark store. The global figure is one dollar in seven; the Indian figure is one dollar in one. The United States is the most thoroughly store-picked large market in the world.
| Architecture, 2026 | USD bn | Share of US market | Representative operators |
|---|---|---|---|
| Store-picked instant | 39.5 | 89.8% | Instacart, DoorDash, Uber Eats, Walmart, Amazon |
| Dark-store instant | 4.5 | 10.2% | Gopuff, Amazon Now, DashMart-format nodes |
| Total | 44.0 | 100% |
The consequence is a cost structure with almost no fixed block. A store-picked operator leases no dark store, holds no grocery inventory, employs no picker and carries no waste. It buys dispatch capacity, rents demand through marketing, and monetises somebody else's shelf twice — once through a transaction fee, once through advertising. That is why the American market produced a profitable operator four years before the Indian market did, and why the profitable operator's margin is a take-rate margin rather than a throughput margin.
Forecast 2025–2033
| 2025 | 2026 | 2033 | CAGR 2025–33 | |
|---|---|---|---|---|
| Value, USD bn | 37.2 | 44.0 | 110.0 | 14.51% |
| Volume, million orders | 496 | 564 | 1,250 | 12.25% |
| Derived basket, USD | 75.00 | 78.00 | 88.00 | 1.51% |
Read the gap between 14.51% and 12.25%. In the global market that gap is 0.16 percentage points — value growth is almost entirely volume growth. In the United States it is 2.26 percentage points, which means more than a seventh of American value growth comes from a larger basket.
Two things produce that. The first is grocery price inflation being passed through a channel that is priced off shelf prices, so a rise in the underlying basket shows up immediately and completely. The second is assortment: as the platforms added pharmacy, beauty, alcohol and general merchandise, the average order widened. Both are real, and both are also the reason the American market is less exposed to the subsidy dynamic that dominates Asia — a market whose growth comes partly from basket is a market whose growth does not have to be bought with discount.
What is excluded
- Scheduled and next-day grocery e-commerce. The largest part of Amazon's and Walmart's grocery volume sits here, and it is deliberately outside this report.
- Pickup. A material share of Instacart's GTV and a growing share of Walmart's. The consumer travels; there is no delivery promise to test.
- Restaurant food delivery. DoorDash's and Uber Eats' restaurant businesses are counted only where they move retail goods.
- Meal-kit and subscription-box delivery. A different demand curve and a different cost base.
The boundary between the first two exclusions and the market itself is where every American market-size disagreement lives, and no operator discloses enough to draw it precisely. This report draws it conservatively and says so in the reconciliation above rather than presenting a clean number that would not survive contact with a company filing.
Table des matières
Table of Contents
- Market Overview and the American Perimeter
- Market Sizing, Structure and Forecast 2025–2033
- Value Chain: Demand, Dispatch and Somebody Else’s Shelf
- Technology and the Retail Media Layer
- Competitive Landscape
- Regulatory Landscape
- Investment Activity and the DoorDash–Instacart Transaction
- Market Risk and Substitution
- Regional and Metropolitan Analysis
- Research Methodology and Sources
All figures USD. Value is retail value of goods transacted, not platform revenue. Volume is orders. Base year 2025, forecast year 2033.
Paysage concurrentiel
Competitive Landscape
Five operators matter and only one of them publishes enough to be analysed. That is the defining feature of American instant retail: the market is oligopolistic, the participants are large listed companies with substantial disclosure obligations, and yet almost none of them breaks out instant grocery in a way that permits a share calculation.
The disclosed operator
Instacart is the only participant that reports the instant grocery business as a business. In Q3 2026 it ran USD 10,528 million of GTV across 91.4 million orders at USD 115.19 an order, converted that into USD 1,095 million of revenue and USD 342 million of adjusted EBITDA, and generated USD 307 million of free cash flow while buying back USD 230 million of its own shares and ending the quarter with USD 1 billion of cash.
Those are the numbers of a mature business, not a growth company. GTV up 15%, orders up 10%, advertising up 23%, and a 3.2% adjusted EBITDA margin on GTV that has been stable at or above 3.0% for several quarters. Instacart has finished the transition from a venture-funded marketplace to a profitable intermediary, and it did so at a scale — roughly USD 42 billion of annualised GTV — that no dark-store operator outside China has approached.
The acquirer
DoorDash is the largest restaurant-delivery platform in the United States and, since 29 October 2026, the buyer of Instacart at approximately USD 17 billion of equity value. It brings the two assets Instacart does not have at scale: a courier fleet that is already economically deployed for restaurant orders, and a consumer base that opens the app several times a week for reasons other than groceries.
The strategic arithmetic is about utilisation. A courier fleet built for a dinner peak has idle capacity between the peaks, and grocery orders fill it. A grocery platform built on somebody else’s shelf has no fleet of its own and buys capacity at the market price. Combining them means the merged entity can fill its own troughs at its own marginal cost, which is the only durable cost advantage available in this architecture.
The retailers who are also competitors
The two largest grocery retailers in the United States operate their own on-demand delivery services and are, simultaneously, the shelves that third-party platforms pick from. That dual role is the most important competitive fact in the market and it is entirely absent from share tables.
A retailer running its own delivery channel has a structural advantage on cost — no 7.3% transaction fee to pay — and a structural disadvantage on demand, because its app is opened less often than a platform’s. The equilibrium has held because the platform brings incremental orders. It will stop holding when a retailer concludes that a material share of platform orders would have arrived through its own app, and every large grocer is now testing that proposition market by market.
The dark-store minority
Gopuff is the survivor of the American dark-store experiment and operates at a scale that is small against the store-picked platforms but large against any comparable non-Indian operator. Amazon’s quick-commerce format and DashMart-style nodes add further capacity, and together the dark-store segment accounts for roughly USD 4.5 billion of the USD 44.0 billion market.
Its constraint is the same everywhere: fixed cost per node against a catchment that may not support the throughput required. Its advantage is control of assortment and therefore of substitution quality, which is the operational failure that most damages store-picked platforms. In dense urban catchments a dark store can beat a supermarket pick on both speed and reliability; outside them it cannot be justified at all.
What the structure implies
Three propositions follow from the disclosed evidence:
- The store-picked architecture has already won on scale. Nine of every ten dollars of American instant retail moves through a shelf the platform does not own, and the one large profitable operator is store-picked.
- Consolidation is at the demand layer, not the asset layer. The sector’s largest transaction buys customers, retailer relationships and an advertising business. Nobody is buying warehouses.
- The retailer is the swing vote. The market’s future share between platforms and grocers will be decided by grocers’ own arithmetic about incrementality, not by anything a platform does.
What is not yet visible is a fifth entrant of consequence. The capital required to build demand at national scale is now large enough that the plausible entrants are the companies that already own either a shelf or a customer, which is why the next move in this market is more likely to be a retailer buying a platform than a new company building one.
Chaîne de valeur
Value Chain: Demand, Dispatch and Somebody Else’s Shelf
The American instant retail chain has five links instead of six, because the pick happens inside somebody else’s building. That single omission removes lease, inventory, shrinkage, waste and packing labour from the platform’s cost base and moves them onto the retailer — who accepts them because the order is incremental volume it would not otherwise have seen.
| Link | Who performs it | Who bears the cost |
|---|---|---|
| Demand generation | Platform | Platform (marketing), retailer (promotions) |
| Assortment and inventory | Retailer | Retailer |
| Pick and pack | Retailer staff or platform shopper | Retailer or platform |
| Dispatch | Platform | Platform |
| Last mile | Independent courier | Platform, consumer fee, membership |
The chain is worth reading as a map of who has leverage. The retailer owns the inventory and therefore owns the alternative if the platform’s economics stop working. The courier owns the capacity and is engaged on terms that can be renegotiated by regulation. The platform owns only the customer relationship and the ranking algorithm — which is precisely why it monetises the ranking.
Where the money actually goes
Take USD 100 of groceries bought through a store-picked platform. Roughly USD 92 goes to the retailer as the wholesale or shelf value of the goods. The platform’s revenue is the remaining USD 10.40 — of which about USD 7.30 is a transaction fee and about USD 3.10 is advertising, on Instacart’s Q3 2026 ratios.
Out of that USD 10.40 the platform pays for the courier, the shopper in some models, customer support, payment processing itself and the technology estate. It retains USD 3.20 as adjusted EBITDA. That decomposition is the whole American business in five numbers, and it explains why the sector’s margins are structurally thin even where they are positive: the platform is taking a tenth of the transaction and spending two thirds of that tenth on delivery.
| Per USD 100 of GTV, Q3 2026 | USD |
|---|---|
| Goods value to the retailer | 89.60 |
| Platform transaction fee | 7.30 |
| Platform advertising revenue | 3.10 |
| Platform adjusted EBITDA | 3.20 |
| Cost to serve, support and technology | 7.20 |
The gap between USD 10.40 of revenue and USD 3.20 of EBITDA — USD 7.20 — is the cost of delivering somebody else’s groceries. It is also the largest single number a store-picked platform can influence, and the lever it uses is batching.
Batching is the whole margin
A courier who delivers one order earns a per-drop fee that has to cover the whole trip. A courier who delivers three orders from the same store on one route covers the same trip three times and the platform’s cost per order falls by roughly two thirds on the marginal drops. Every American platform’s profitability therefore tracks two things: order density in a catchment and the proportion of orders that can be batched without breaching the promise.
That is why the American market is geographically lopsided. Dense metropolitan catchments are profitable; suburban and rural catchments are not, and are usually served on scheduled windows rather than instant ones — which is exactly the perimeter distinction this report draws in its first chapter, arriving from the operating side rather than the definitional one.
Membership and the subscription of demand
Both large American platforms bundle free delivery into a paid membership. The economics are not primarily about the membership fee. A member orders more often, batch density rises, cost per order falls, and the platform acquires a customer whose switching cost is a subscription rather than a habit.
This is the most durable structural advantage available in the American market and it is unavailable to a dark-store operator in the same form, because a dark-store operator’s cost per order is dominated by a fixed block that does not fall as frequency rises — whereas a store-picked operator’s cost per order falls sharply with batching. In the United States the subscription therefore compounds a genuine operating economy. In India it would mostly discount a fixed cost.
The retailer’s side of the bargain
The retailer accepting a 7.3% transaction fee is not obviously winning. It gives up a tenth of the ticket to reach a customer who did not come to the store, and it hands the platform the customer relationship and the data. The reason it accepts is capacity utilisation: the shelf, the staff and the building are already paid for, so an incremental order at 92% of shelf value is profitable if the alternative is no order.
That bargain holds as long as the platform brings incremental demand. It breaks the moment a retailer concludes that a material share of those orders would have come through its own app anyway — at which point it is paying 7.3% for the acquisition of a customer it already had. Every large American grocer is now running that arithmetic, and it is the single most consequential open question in the chain.
Activité d'investissement
Investment Activity and the DoorDash–Instacart Transaction
Capital in American instant retail is doing something unusual for this sector: it is buying earnings rather than building capacity. That is the direct consequence of a market where the architecture is store-picked and the cheapest way to grow is to route more orders through shelves that already exist.
The transaction, priced
DoorDash’s agreement to acquire Instacart, announced 29 October 2026, values the grocery platform at approximately USD 17 billion of equity in cash and stock. Three multiples can be derived from Instacart’s own Q3 2026 disclosure:
| Instacart valuation, 29 October 2026 | Value |
|---|---|
| Equity value | USD 17.0bn |
| Q3 2026 adjusted EBITDA | USD 342m |
| Annualised adjusted EBITDA | USD 1,368m |
| Multiple of run-rate adjusted EBITDA | 12.4x |
| Q3 2026 GAAP net income | USD 147m |
| Annualised GAAP net income | USD 588m |
| Multiple of run-rate net income | 28.9x |
Twelve times EBITDA and twenty-nine times earnings, for a business growing GTV at 15%, is a fair price for a mature intermediary and a poor price for a growth asset. That is not a criticism of the buyer. It is the clearest statement the sector has made about what it now believes: the value in American instant retail sits in the durability of the take rate and the advertising line, not in the possibility of a step change in volume.
What Instacart’s own capital allocation said first
Before the sale was agreed, Instacart’s Q3 2026 cash flow told the same story. The company generated USD 307 million of free cash flow and spent USD 230 million buying back its own shares, ending the quarter with USD 1 billion in cash and similar assets. A company buying back stock rather than investing in capacity is telling its shareholders that it cannot find a higher return inside the business than its own valuation.
That is rational in a market with no dark-store estate to build and no new geographies to enter. It is also the reason the acquisition happened: a business that generates cash and has nowhere to put it is a business that gets bought by one that has somewhere to put it.
The capital that was destroyed
The American dark-store experiment preceded the store-picked consolidation and cost more than it returned. The 2021 vintage of rapid-delivery platforms raised capital at valuations that assumed national coverage could be bought ahead of unit economics, and contracted sharply when the cost of capital rose. The survivors are few and the largest is a fraction of the size of either of the two store-picked platforms now combining.
The lesson the second wave drew, and the reason no American platform is now building dark stores at scale, is arithmetic rather than sentiment. A dark store requires a lease, a fit-out, an inventory position and a staffed pick operation before it serves a single order. A store-picked platform requires a contract with a retailer who has already paid for all four. In a market with a supermarket within two miles of most customers, the second is simply cheaper.
Where capital goes next
Three directions, in order of visibility:
- Courier fleet and dispatch capacity. The merged entity’s advantage depends on filling a fleet’s troughs with grocery orders, which requires capital in dispatch technology rather than in buildings.
- Retail media. Advertising at 3.1% of GTV and growing 23% is the fastest-compounding revenue line in the market, and it requires data infrastructure rather than property.
- Retailer software. Selling fulfilment and storefront technology to grocers deepens relationships and creates a revenue line that is insensitive to order volume.
Notably absent from that list is any investment in the physical estate. The American market is forecast to add USD 66 billion of value between 2026 and 2033 without requiring its participants to build a significant number of new fulfilment nodes. They will route the additional orders through shelves that already exist, and the capital required to do that is software and marketing rather than property.
What that means for the 2033 number
A market that grows by utilisation rather than by construction is a market that can compound at 14.51% without a capital cycle. It does not need a funding window to open, it does not depend on the cost of debt, and it does not carry the risk that an estate built in one interest-rate environment has to be paid for in another.
That is a structurally better position than the Indian market, which needs primary equity to fund leases, and a structurally worse one than the Chinese market, which reached national coverage on somebody else’s capital. It is the single most important reason to treat the American forecast as lower-risk than its growth rate would normally imply.
Technologie et innovation
Technology and the Retail Media Layer
American instant retail does not invest in buildings. It invests in three software problems and one advertising market, and the advertising market is where the money is.
Problem one: search ranking, which is a revenue line
When a customer types “milk” into an app that can see 30,000 lines in the nearest store, somebody has to decide what appears first. That decision is worth money, and the American platforms were the first to price it explicitly. Instacart’s advertising and other revenue reached USD 331 million in Q3 2026, up 23% year on year, or 3.1% of gross transaction value and nearly a third of total revenue.
That is not a sidebar banner business. It is a sponsored-placement market attached to search intent at the moment of purchase, and it carries gross margins close to total because the inventory being sold is a position on a screen. It is also the reason the American model is profitable: transaction revenue of USD 764 million grew 14% while advertising grew 23%, and adjusted EBITDA of USD 342 million grew exactly in line with the advertising line.
Every dark-store operator in the world is trying to build the same thing and none has reached a comparable share of revenue. Zepto’s advertising reached 7.23% of operating revenue in FY26, which is a higher percentage than Instacart’s — but Zepto’s operating revenue includes the full value of goods sold under a first-party model, and its absolute advertising revenue of INR 1,636 crore is a fraction of Instacart’s. Two different denominators, two different businesses.
Problem two: the pick
In a store-picked model the pick is performed by a human walking aisles designed for a browsing shopper rather than a professional picker. The platform’s technology contribution is to sequence the list against the store’s actual layout, route the shopper through the aisles in a sensible order, and handle substitution when an item is missing.
Substitution is the hardest of the three and the least discussed. A dark-store operator controls its own assortment and therefore its own stockouts. A store-picked operator does not: the supermarket’s own replenishment decides whether the order can be filled, and the platform’s reputation absorbs the failure. The technology answer is a substitution model that proposes the closest acceptable alternative and prices it so the customer does not feel cheated — and the reason it matters commercially is that a failed substitution is the single most reliable predictor of a lost repeat customer in grocery.
Problem three: dispatch against somebody else’s capacity
The American platforms largely do not employ their couriers. Dispatch therefore has to solve a marketplace problem rather than a scheduling problem: matching orders to independent couriers who choose when to work, at a price that clears the market in each neighbourhood at each hour.
That is a pricing problem before it is an optimisation problem, and it is where the American model absorbs its largest variable cost. Batching — one courier, several orders from the same store — is the main lever, and it works best in dense urban catchments. It is also the lever most exposed to the regulatory risk discussed later, because the economics of batching assume a courier paid per drop.
The fourth business: selling the technology back to the retailer
The most under-appreciated part of the American model is that the platforms sell fulfilment and e-commerce software to the retailers whose shelves they monetise. Storefront, fulfilment, in-store technology and connected-store systems are sold as a platform, which converts the retailer from a supplier of inventory into a customer of software.
That has two effects. It deepens the relationship beyond a commission negotiation, and it makes the platform’s revenue less sensitive to order volume alone. It also creates a strategic tension: a technology platform that also runs a competing consumer marketplace is selling tools to businesses it competes with for the same consumer, and the retailers know it.
What is genuinely not being built
The United States has deliberately not built the automated micro-fulfilment centre at scale. The technology exists and is proven in grocery elsewhere; the economics do not compel it here, because American retail already has a shelf within two miles of almost every customer and the cheapest way to get goods off that shelf is a person with a phone.
That is the correct decision for this market and it is the reason the American forecast grows at 14.51% without any requirement for capital investment in dark stores. A market that can add volume by routing more orders through shelves that already exist does not need to build anything, and therefore does not need to be financed to build it.
Risque de marché
Market Risk and Substitution
Four risks matter in the United States and only one of them is a market risk. The other three are structural, regulatory and commercial, and in each case the exposure is to the revenue line rather than to demand.
Risk one: the retailer decides it does not need the platform
This is the largest risk in the American market and it is entirely commercial. The platform charges roughly a 7.3% transaction fee to route an order to a shelf owned by a grocer, and the grocer accepts because the order is incremental. The moment a grocer concludes that a material share of those orders would have arrived through its own app, the fee stops being a customer-acquisition cost and becomes a tax on demand it already had.
Every large American grocer is now running that calculation, and the evidence that it matters is that they all operate their own on-demand channels. A coordinated shift of volume to retailer-owned apps would compress the transaction take across the market without reducing demand at all — the orders would still happen, at a lower revenue per order for the intermediary. In a market where adjusted EBITDA is 3.2% of GTV, a two-percentage-point reduction in the transaction take would eliminate most of the profit.
Risk two: growth depends on the basket
More than a seventh of American value growth comes from a larger basket rather than more orders. That is unusual in this sector and it is the market’s greatest strength and its most under-priced vulnerability.
A basket that widens because grocery prices are rising is not a structural gain. If grocery inflation normalises toward zero, the 2.26-percentage-point gap between value and volume growth closes, and the American market grows at something closer to its 12.25% volume rate. The difference between 14.51% and 12.25% over seven years is roughly USD 16 billion of 2033 value — material, but not existential, which is precisely why it tends to be ignored in market commentary.
Risk three: the courier is reclassified
The per-delivery cost that makes batching profitable assumes an independent contractor paid per drop. Reclassification in a state large enough to set the national benchmark would convert part of that variable cost into scheduled labour, and the effect on cost per order would be immediate and disproportionate because the model’s margin is only 3.2% of GTV.
The American position is unsettled rather than resolved: a state-by-state patchwork, with some jurisdictions imposing minimum pay standards and a smaller number attempting reclassification. That uncertainty is itself a cost, because it discourages exactly the long-lease commitments a dark-store strategy would require and therefore forecloses one route out of the model’s thin margins.
Risk four: substitution from the shelf and the club
The strongest substitutes for American instant retail are not other delivery platforms. They are the store the consumer drives to, the club format that sells in bulk at a discount, and the same grocer’s own pickup option.
Pickup deserves specific attention because it is close to free for the retailer — the customer performs the last mile — and it is growing inside the same apps that sell delivery. Every order that converts from delivery to pickup removes a courier cost and a delivery fee from the system. In a market whose total value is defined by goods transacted, that conversion does not show up as a decline in GTV, which means the risk is invisible in the headline number and visible only in the platform’s cost to serve.
Scenario range, 2033
| Scenario | 2033 value, USD bn | CAGR 2025–33 | Assumptions |
|---|---|---|---|
| Retailer-led contraction | 96 | 12.58% | Grocers shift material volume to their own channels, transaction take compresses, grocery inflation normalises. |
| Base case | 110 | 14.51% | DoorDash–Instacart completes, store-picked architecture holds share, basket widens at 1.51% a year. |
| Platform consolidation pays off | 124 | 16.24% | Merged fleet lowers cost per order, retail media expands faster than GTV, dark-store niche grows in dense metros. |
Note where the scenarios diverge. They differ by about 29% of 2033 value, which is a wide band for a market with no construction requirement — and the band is driven almost entirely by the transaction take rather than by order volume. In every scenario Americans order more groceries for delivery in 2033 than they do in 2026. What varies is how much of that the intermediary keeps.
What would change the forecast
- A retailer acquiring or building a national platform. The single event that would move the market toward the low case, because it would prove the shelf can disintermediate the app.
- A merger remedy separating Instacart’s advertising business. This would protect margin while capping the fastest-growing revenue line — a different shape of outcome from a volume shock.
- Courier reclassification in California or New York. Either would reset the national cost benchmark and force a repricing of delivery fees, which would propagate into basket and then into value.
None of these is a prediction. They are the three points at which the American model is most sensitive, stated so that a reader can price them rather than accept them.
Paysage réglementaire
Regulatory Landscape
The United States does not regulate instant retail as a category. It regulates the courier, the alcohol licence, the subscription and the merger, and those four regimes in combination do more to shape the market’s economics than any consumer-protection rule could.
The merger is the live question
The DoorDash–Instacart transaction announced on 29 October 2026 will be reviewed as both a horizontal and a vertical matter. The horizontal question is straightforward: two of the largest platforms delivering retail goods to American homes combining into one.
The vertical question is the more consequential and the less precedented. Instacart’s advertising business depends on being a neutral intermediary between retailers, brands and consumers. Owned by a competing delivery platform, that neutrality is contestable, and a competition authority that treats retail-media neutrality as a relevant dimension of competition would extract structural remedies — data separation, or an advertising arm operated at arm’s length — that change how the combined business earns its margin.
Note what is not at stake. No authority is examining whether grocery delivery should exist, or whether a platform may pick from a retailer’s shelf. The regulatory risk in this market is a risk to the structure of the revenue line, not to the format.
Courier classification: a state-by-state patchwork with a federal vacuum
The American platforms engage couriers as independent contractors, and that classification is the foundation of the per-delivery cost that makes batching profitable. It is protected in some states by statute and contested in others, and the result is a national market operating under fifty different labour regimes.
Two developments matter for the operating model. Several states and cities have imposed minimum per-delivery or per-hour pay standards for app-based couriers, which converts part of the cost base from a market-clearing price into a floor. And a small number of jurisdictions have reclassified or attempted to reclassify couriers as employees, which would move the model from a variable cost per drop to a scheduled labour cost with hours, contributions and shift obligations.
The distinction is the same one that decided the European market in the opposite direction. A model whose cost per delivery depends on a self-employed courier paid per drop has a materially different scaling profile from one whose cost depends on employed labour. In the United States the question is unsettled rather than resolved, and that uncertainty is itself a cost, because it discourages the long-lease commitments a dark-store estate would require.
Licensing: alcohol and the assortment question
Alcohol is the highest-value category a rapid delivery platform can add and the most heavily licensed. Licences are issued at state and often municipal level, they are frequently non-transferable, and third-party delivery of alcohol requires a chain of authorisation through the retailer that holds the licence. The result is that alcohol availability varies street by street and the category’s contribution to basket — the single biggest lever on American order value — is decided by licensing boards rather than by merchants.
Prescription and over-the-counter pharmacy adds a parallel set of constraints. Both categories are worth more than the grocery that anchors the order.
Consumer protection: subscriptions and fees
Membership is central to the American model and is therefore the natural target for consumer-protection activity. Rules requiring straightforward cancellation of recurring subscriptions, and disclosure requirements on fees presented late in checkout, directly affect the two levers that hold average order value up: bundled free delivery and the delivery fee itself.
Neither is existential. Both reduce the effective price of an order at the margin, and in a market whose growth depends more on basket than on volume, a marginal price change propagates into the forecast faster than it would in a volume-led market.
What regulation does to this forecast
The 14.51% base case assumes no reclassification of couriers in a state large enough to reset the national cost benchmark, and assumes the DoorDash–Instacart transaction completes without a remedy that separates the advertising business.
Those are the two sensitivities that matter, and they point in opposite directions on value and on margin. A reclassification would raise cost per order and slow volume; a structural remedy on retail media would protect margin while capping the fastest-growing revenue line. Neither is a demand event, which is characteristic of a market whose regulator has so far been interested in the participants rather than the category.
