The platform layer is one of the most concentrated markets in the digital economy measured by revenue captured, and one of the most fragmented measured by participants. Three companies disclose 53.01 per cent of global platform rent. Below them sit dozens of regional marketplaces, several large commerce-software vendors, and a long tail of vertical platforms that will never split out a third-party line.
Consolidated net sales of USD 716,924m in 2025, up 12 per cent, with net income of USD 77,670m. Third-party seller services reached USD 172,162m, up 10.26 per cent, and advertising services USD 68,635m, up 22.10 per cent. First-party online stores grew 9.01 per cent to USD 269,287m. The competitive fact inside those numbers is that Amazon's seller-facing revenue is growing faster than its own retail revenue, which is the definition of a platform becoming more platform-like.
Revenue of RMB 1,023,670m, or USD 148,401m, up 3 per cent in fiscal 2026 after excluding the disposed Sun Art and Intime businesses; on a like-for-like basis growth was 11 per cent. Customer management revenue rose 5 per cent to RMB 343,867m, which management attributes to improvement in take rate. Quick commerce revenue rose 47 per cent to RMB 78,520m, and Cloud Intelligence Group rose 34 per cent to RMB 158,132m. The company is simultaneously raising rent on its core marketplace and buying frequency in immediate delivery.
Merchant GMV of USD 378,441m, up 29.48 per cent, and revenue of USD 11,556m, up 30.14 per cent, with free cash flow of USD 2,007m and a USD 2 billion repurchase authorisation. GMV is three times the 2020 level. Shopify's own estimate puts its share of the US e-commerce market above 14 per cent, derived from Census retail data rather than from a survey.
PDD Holdings, JD.com, MercadoLibre, Sea Limited, Coupang, eBay and Etsy each hold a defensible national or category position and none of them splits a comparable third-party line. They are carried in the modelled residual at USD 225,532m and they are the main reason that residual is not a rounding error.
Commission is the least defensible revenue line on a platform's income statement and services are the most defensible. That is not a preference, it is what the disclosed numbers show: the fastest-growing lines across the three disclosing platforms are advertising, payment attach and cloud, none of which is the commission rate.
The second implication concerns scale. Fixed compliance cost, fulfilment density and machine-learning infrastructure all reward size, which is why the layer consolidates at the top while fragmenting at the merchant level. A regional platform with 2 per cent of global rent can still be dominant nationally, and usually is.
Expect three movements to 2033: consolidation of commerce-software vendors into payment and fulfilment owners; a durable split between capital-heavy marketplace networks and asset-light software platforms; and rising regulatory scrutiny of the services a platform sells to the sellers who depend on its ranking.
" value_chain_content: "A platform's six cents per dollar of GMV is not collected in one place. It is collected at four points in a single order: the commission at checkout, the spread on the payment, the fee for storing and shipping the item, and the advertising the seller buys to be found in the first place. Each point has a different gross margin and a different ceiling, and the mix explains why two companies with similar take rates can have very different economics.
Mapping named companies to the stage they are actually paid at, rather than to a category, is the only way to read the layer.
| Stage | Who is paid here | 2025 evidence |
|---|---|---|
| Storefront and merchandising software | Shopify Inc., Salesforce, Adobe | Shopify: USD 11,556m of revenue on USD 378,441m of merchant GMV, 3.05 per cent |
| Marketplace commission | Amazon, Alibaba Group, PDD Holdings | Amazon third-party seller services USD 172,162m, up 10.26 per cent; Alibaba customer management revenue RMB 343,867m, up 5 per cent |
| Cross-border platform | Alibaba International Digital Commerce, Sea Limited | Alibaba international commerce USD 20,900m on RMB 144,170m, up 9 per cent |
| Fulfilment and logistics services | Amazon, Alibaba, Coupang | Amazon's related fulfilment and shipping fees are booked inside third-party seller services; Alibaba direct sales and logistics revenue RMB 105,518m |
| Payments | Shopify, MercadoLibre, Sea | Shopify Gross Payments Volume up 37 per cent and Shop Pay GMV up 62 per cent in 2025 |
| Discovery and seller advertising | Amazon, Alibaba, Etsy | Amazon advertising services USD 68,635m, up 22.10 per cent - faster than any other Amazon line except AWS |
| Immediate and quick commerce | Alibaba, Coupang, MercadoLibre | Alibaba quick commerce RMB 78,520m, up 47 per cent |
The same six cents can be ninety per cent gross margin or thirty. A subscription and commission business with a hosted storefront carries software gross margins. A business that stores inventory in its own buildings, employs couriers and pays for the return leg is running a logistics operation with a software front end. Amazon's 2025 accounts make the split visible: USD 142,352m of net additions to property and equipment in a single year, of which USD 96,496m went to AWS and the balance largely to the fulfilment and transportation network that serves the marketplace.
This is the structural reason a platform that owns fulfilment can raise its take rate without touching its commission schedule. The commission is published and competitive; the fulfilment fee is inside the platform's own service catalogue and priced against a benchmark the seller cannot easily reconstruct.
For a platform, the defensible position is the stage where the platform holds information the merchant cannot reproduce: where the parcel physically is, whether the card was authorised, whether the buyer returned the last three orders. For a merchant, the correct measure is total cost of channel per order, not commission rate - and total cost of channel includes the advertising needed to be discovered inside the platform's own search results.
Through 2033 the value chain consolidates around two poles: catalogue and identity on one side, physical and payment rails on the other. Platforms that own neither become interchangeable at the margin, which is why the commerce-software cohort is being repriced against payment attach and fulfilment partnerships rather than against seat count.
" investment_activity_content: "Capital allocation in this layer has bifurcated. Asset-light software platforms are converting operating leverage into cash returns. Marketplace networks are converting it into buildings, robots and delivery density. Both are rational, and they produce very different risk profiles for anyone valuing the layer.
Net additions to property and equipment were USD 142,352m in 2025, against USD 85,752m in 2024 and USD 48,344m in 2023. USD 96,496m of the 2025 figure went to AWS, and the North America segment added USD 35,919m. Total property and equipment, net, reached USD 357,025m. This is the cost of guaranteeing delivery promises and of hosting the inference that increasingly sits behind the storefront.
Free cash flow of USD 2,007m and a USD 2 billion share repurchase programme authorised alongside the full-year results. Ten consecutive quarters of double-digit free cash flow margins. The capital is not going into warehouses because the merchants own the inventory.
Quick commerce investment drove a 47 per cent increase in that segment's revenue to RMB 78,520m, funded while the company was completing the disposal of Sun Art and Intime. Cloud revenue rose 34 per cent to RMB 158,132m. The allocation decision is explicit: pay for demand today, own the compute layer for tomorrow.
The bifurcation matters for anyone underwriting the layer. An asset-light platform is valued on revenue quality and free cash flow conversion; a network platform is valued on capital efficiency and the margin on services. Applying one framework to the other is the most common analytical error in this market.
It also determines who can afford regulation and who can afford AI. Both are fixed-cost programmes, and the platforms with the largest disclosed rent are the only ones that can carry them without diluting the merchant proposition.
Capital intensity rises in absolute terms to 2033 while falling as a share of platform revenue, because services revenue scales faster than the physical network required to serve it. The exception is immediate delivery, where density economics do not improve with software.
" technology_innovation_content: "The layer's technology question changed shape in 2025. For two decades the innovation agenda was the storefront: faster pages, better search, higher conversion. The agenda now is the transaction itself - who presents the product, who authorises the payment, and which system holds the authoritative answer about availability and delivery.
Three technologies are being deployed at commercial scale rather than in pilots.
Product discovery is moving from a search box the buyer operates to a model that assembles a shortlist. Shopify's own account of 2025 names the investments explicitly - Catalog, Sidekick and a Universal Commerce Protocol - alongside the statement that AI is reshaping how buyers discover and purchase. The commercial consequence is straightforward: when a model assembles the shortlist, the seller's ranking inside a marketplace is worth less than the machine-readable completeness of its catalogue.
Shop Pay GMV grew 62 per cent in 2025 and Gross Payments Volume 37 per cent, both far faster than the underlying GMV growth of 29.48 per cent. Payment attach is the most reliable way for a software platform to raise monetisation without renegotiating commission, because the merchant is choosing a convenience rather than accepting a price rise.
Robotics, automated storage and retrieval, and route density models are the technology that actually changes unit cost. They are also the technology that shows up in capital expenditure rather than in a press release: Amazon's net additions to property and equipment reached USD 142,352m in 2025 against USD 85,752m in 2024.
Technology spend in this layer is converging on the transaction record rather than the shop window. That favours platforms that already hold the payment authorisation and the fulfilment scan, because those two events generate the data an agent needs to complete a purchase. It disadvantages the pure storefront vendor whose differentiation was template quality.
The second implication is a measurement problem. A platform that shifts capability into an ecosystem of apps and partners books less revenue per transaction while retaining more control. Reported take rates will therefore understate true monetisation where the platform takes a share of partner revenue rather than a fee from the merchant.
By 2033 the base case is that machine-mediated discovery handles a substantial minority of product research, payment authorisation is largely tokenised and invisible, and fulfilment promises are generated by the platform's own network model rather than by the merchant's warehouse. The platforms that hold catalog data, payment tokens and delivery telemetry simultaneously will price the transaction. The ones that hold only the storefront will compete on price and lose the argument.
" market_risk_content: "The 8.60 per cent forecast rests on two variables, third-party GMV growth and take-rate drift. Both can move against the platform layer without any change in consumer demand, which makes the risk section unusually load-bearing for this report.
US e-commerce grew 12.2 per cent year on year in the second quarter of 2026 while total retail grew 6.7 per cent. China's online retail of physical goods grew 5.2 per cent in 2025 against total retail growth of 3.7 per cent. The two largest markets are growing at very different speeds, and a platform's revenue is weighted to the market where it collects.
Take-rate drift has a ceiling set by the merchant's alternative. Composable commerce, regional marketplaces and direct storefronts are all viable at sufficient scale, and every basis point of drift raises the incentive to test one.
Three companies disclose 53.01 per cent of the layer. A forecast this dependent on a handful of disclosed lines is exposed to a single accounting reclassification - a platform moving a service between revenue lines would move the measured market without changing the underlying economics.
The withdrawal of low-value duty relief in 2025 raised the landed cost of imported goods and shifted customs work into checkout. Further changes to de minimis treatment, tariffs or product-safety liability would reduce the cross-border share of platform revenue fastest, because cross-border is where the compliance service fee is largest.
Immediate delivery growth is being bought. Where a platform is paying for frequency, the revenue line grows faster than the contribution margin, and a withdrawal of subsidy would show up as a revenue slowdown rather than as a cost saving.
Readers should treat the disclosed seats as evidence and the residual as an assumption. The correct sensitivity is not to the CAGR but to the residual: a 25 per cent error there moves the 2025 base to USD 423.6 billion or USD 536.4 billion, which is wide enough to change a valuation and narrow enough that the direction to 2033 holds.
The most probable disruption is not a new competitor but a change in the surface where the purchase begins. If discovery moves decisively to machine intermediaries, the platform layer's rent does not disappear - it moves to whoever holds the catalogue, the token and the delivery promise. That outcome is closer to a reallocation than a contraction, and it is the scenario this forecast is built to survive.
" regulatory_landscape_content: "Platform regulation has moved from consumer protection to structural conduct. The obligations now attaching to large marketplaces - on self-preferencing, data access, seller terms and product safety - act directly on the revenue lines this report measures, because they constrain how a platform may charge the sellers who compete with its own retail operation.
The Digital Markets Act imposes conduct obligations on designated gatekeepers, including limits on anti-steering and requirements around data access and ranking transparency. The Digital Services Act imposes notice-and-action duties and, for very large platforms, systemic risk assessment. The General Product Safety Regulation places responsibility for the safety of goods sold through a marketplace on the economic operator, which in practice means the platform must be able to identify the seller. Separate VAT rules and the DAC7 reporting framework create a documented seller trail that tax authorities can now request directly.
The most consequential change for cross-border volume was the withdrawal of low-value duty-free treatment in 2025, applied first to China-origin parcels and then more broadly. That removes a cost advantage that had been built into the price of low-value imported goods, and it converts a customs question into a checkout question: duty calculation is now a platform service, and someone has to pay for it.
Platform obligations have been tightened around pricing conduct, seller fees and delivery commitments, with particular attention to the immediate-delivery segment where subsidy competition has been aggressive. The regulatory direction is toward fee transparency rather than toward capping take rates.
Inventory-based models remain restricted in India's marketplace rules, which shapes the country's platform layer toward a marketplace-with-logistics model rather than a first-party one. Comparable restrictions and disclosure duties are spreading across Southeast Asia and the Gulf.
Compliance is fixed cost, and fixed cost is a moat. Every new obligation - seller identification, safety evidence, tax reporting, algorithmic disclosure - costs roughly the same to implement regardless of platform size, which is why regulation has so far consolidated the layer rather than fragmented it.
The revenue effect is subtler. Obligations that restrict self-preferencing push platforms to separate retail from marketplace more cleanly in their accounts, which will make the rent line easier to see and easier to compare across companies. That is a disclosure improvement for buyers of this research and a competitive exposure for platforms whose retail and marketplace results have been blended.
To 2033 the direction is convergence on seller transparency and divergence on liability. Expect the fee schedule to become a partially public document in more jurisdictions, and expect the services attached to it - fulfilment, advertising, compliance - to become the contested ground.
" faq_question_1: "What is the size of the global e-commerce platform market?" faq_answer_1: "USD 480.0 billion in 2025, rising to USD 928.7 billion in 2033 at an 8.60 per cent CAGR. That is 6.04 cents of platform revenue on every dollar of global e-commerce sales." faq_question_2: "What counts as platform revenue in this report?" faq_answer_2: "Commissions, subscription and platform fees, and platform services sold to sellers - fulfilment, logistics-as-a-service, seller-funded advertising and payments processed by the platform. First-party product revenue is excluded, because a platform buying and reselling inventory is taking inventory risk, not earning rent." faq_question_3: "Why does the platform market grow faster than e-commerce overall?" faq_answer_3: "Because platform revenue is third-party GMV multiplied by a take rate, and take rates drift upward. Third-party GMV grows 7.01 per cent over the forecast; the platform layer grows 8.60 per cent. The difference is monetisation, not volume." faq_question_4: "Can take rates be compared across platforms?" faq_answer_4: "Only with care. Shopify monetises 3.05 per cent of merchant GMV because payments and capital products are optional add-ons. Amazon's third-party seller services include fulfilment fees that partially pass through cost. Blended take rates across different revenue mixes are not a like-for-like comparison." global_schema: organization: "@id": MIWS-root#organization "@type": Organization url: "https://www.pheonixresearch.com/" name: Pheonix Research website: "@id": MIWS-root "@type": WebSite url: "https://www.pheonixresearch.com/" name: Pheonix Research api: "@id": MIWS-root#api "@type": WebAPI url: "https://www.pheonixresearch.com/" graph: node_id: MIMR-node_f81891e graph_node_endpoint: "https://graph.pheonixresearch.com/api/v1/query/live/node/MIMR-node_f81891e" graph_snapshot: "https://graph.pheonixresearch.com/api/v1/read/live/graph" graph_exclusions: "https://graph.pheonixresearch.com/api/v1/read/live/exclusions" graph_meta: "https://graph.pheonixresearch.com/api/v1/read/live/meta" graph_bundle: "https://graph.pheonixresearch.com/api/v1/read/live/bundle" discovery: discovery_json: "https://www.pheonixresearch.com/.well-known/pheonix-discovery.json" llms_txt: "https://www.pheonixresearch.com/llms.txt" sitemap: "https://www.pheonixresearch.com/sitemap.xml" --- # Global E-Commerce Platform Market Report, Size, Share and Forecast 2026–2033: USD 480 Billion of Rent on a USD 7.95 Trillion Market - **Market Size (Base Year):** USD 480.0 Billion - **Forecast Value:** USD 928.7 Billion - **CAGR:** 8.60% - **Forecast Period:** 2026 - 2033 - **Coverage:** Global - Global ## Executive Summary The global e-commerce platform market was USD 480.0 billion in 2025 and reaches USD 928.7 billion by 2033, an 8.60 per cent CAGR and an increment of USD 448.7 billion. That is 6.04 cents of platform rent on every dollar of the USD 7,950.0 billion the world spent online in 2025. The layer is measured as revenue platforms book for someone else's transaction - commission, subscription and platform services on third-party and merchant commerce - and explicitly excludes first-party product revenue. The distinction matters because Amazon's USD 716.9 billion of 2025 net sales contains only USD 172.2 billion of platform rent. Three platforms disclose enough to be counted hard: Amazon third-party seller services at USD 172,162 million, Alibaba customer management revenue at USD 49,850 million plus USD 20,900 million of international commerce, and Shopify at USD 11,556 million. Together they are 53.01 per cent of the layer; the remaining 46.99 per cent is a named, modelled residual. The growth mechanism is take-rate drift rather than volume. Third-party GMV grows 7.01 per cent across the forecast; the platform layer grows 8.60 per cent. Shopify illustrates both the size and the limit of that drift: GMV rose 29.48 per cent while monetisation moved from 3.0382 to 3.0536 per cent of GMV, a gain of 1.54 basis points. ## Market Report # Global E-Commerce Platform Market Snapshot, 2026–2033 | Metric | | Value | | Platform revenue (2025) | | USD 480.0 billion | | Platform revenue (2033) | | USD 928.7 billion | | CAGR (2026–2033) | | 8.60% | | Increment 2025–2033 | | USD 448.7 billion | | Unit of measure | | Platform-attributed revenue: commission, subscription and platform services on third-party and merchant commerce | | Platform rent per dollar of e-commerce sales | | 6.04 cents | | Largest disclosed seat | | Amazon third-party seller services, USD 172.2 billion | | Lowest observed take rate | | Shopify, 3.05% of GMV | | Share of layer disclosed by three platforms | | 53.01% | | Growth arithmetic | | Third-party GMV growth (7.01%) plus take-rate drift | ## What a platform market is, once you stop counting sales Amazon booked USD 716.9 billion of net sales in 2025. Almost none of it was platform revenue. USD 269.3 billion was Amazon selling its own inventory as the seller of record, USD 22.6 billion came from physical stores, USD 128.7 billion was AWS. The line that describes a platform business — third-party seller services, the commissions and related fulfilment and shipping fees Amazon earns for enabling someone else's sale — was USD 172.2 billion. That distinction is the whole report. The [global e-commerce market](https://www.pheonixresearch.com/consumer-goods-retail/general-consumer-needs-products/market-report/global-e-commerce-market/) measures what buyers spent: USD 7,950.0 billion in 2025. The platform market measures what the intermediary kept. The two numbers are routinely confused in the same sentence, and the confusion produces forecasts that grow a rent line at the speed of a sales line. Here the platform layer is measured as revenue the platform books for someone else's transaction: commissions, subscription and platform fees, and platform services sold to sellers — fulfilment, logistics-as-a-service, seller-funded advertising, and payments processed by the platform itself. First-party product revenue is excluded, because a platform that buys and resells inventory is not being paid rent, it is taking inventory risk. ## Articles on this market - Two Countries Are 47.69 Per Cent of Global E-Commerce. The Other 52 Per Cent Is a Model. - Shopify Grew GMV 29.48 Per Cent and Revenue 30.14 Per Cent. That 0.66 Points Is the Platform Market. ## USD 480.0 billion: the rent on a USD 7.95 trillion market On that definition the global e-commerce platform layer was **USD 480.0 billion in 2025** and reaches **USD 928.7 billion in 2033**, an **8.60 per cent CAGR** and an increment of USD 448.7 billion. Expressed against the sales that generate it, the entire platform layer takes **6.04 cents on every dollar of global e-commerce GMV**. That single ratio is the reason this market deserves its own forecast rather than a paragraph inside a commerce report. A market growing 7.01 per cent at the GMV layer and 8.60 per cent at the rent layer is not one market with two measurements. It is two markets, and the gap between them is a policy variable that platforms control. ## Three platforms disclose 53.01 per cent of the layer The disclosed seats are unusually concentrated, because only a handful of platform companies split their third-party line out of total revenue. | Platform line, fiscal 2025 | | Revenue (USD m) | | What it measures | | Amazon third-party seller services | | 172,162 | | Commissions plus related fulfilment and shipping fees on third-party sales | | Alibaba customer management revenue | | 49,850 | | RMB 343,867m; marketing and commission revenue earned on China retail GMV | | Alibaba international commerce (retail and wholesale) | | 20,900 | | RMB 144,170m; cross-border retail and wholesale platform revenue | | Shopify | | 11,556 | | Subscription plus merchant solutions on USD 378,441m of merchant GMV | | **Disclosed seats** | | **254,468** | | **53.01 per cent of the 2025 layer** | | Modelled residual | | 225,532 | | All platforms that do not split the third-party line | | **Total platform layer** | | **480,000** | | | The residual is modelled, not disclosed, and it is named rather than blended: PDD Holdings, JD.com's marketplace line, MercadoLibre, eBay, Sea Limited, Coupang, Walmart's marketplace, TikTok Shop, Etsy and the commerce-software long tail behind Shopify. A 25 per cent error in that residual moves the 2025 base to USD 423.6 billion or USD 536.4 billion — wide enough to matter, narrow enough that the 2033 direction does not change. Read the disclosed seats as a floor, not a sample. Three companies account for more than half of global platform rent because those three are the only large platforms that publish the line. Every other figure in the table is an estimate with a stated method, and the report treats the disclosed seats as the only hard evidence in the sizing. ## The take rate is the report Shopify grew merchant GMV 29.48 per cent in 2025, from USD 292,275 million to USD 378,441 million, and revenue 30.14 per cent, from USD 8,880 million to USD 11,556 million. Those two rates are almost identical, which is the point: **monetisation barely moved, from 3.0382 to 3.0536 per cent of GMV — an increase of 1.54 basis points**. A basis point and a half is the entire pricing story of the largest independent commerce platform in the world in a year its GMV rose 29 per cent. Competitive pressure, merchant mix and the deliberate choice to leave payments and capital products as optional add-ons all hold the take rate in a narrow band. The revenue growth is a volume story; the pricing power is a rounding error that happens to compound. That is not true everywhere. Alibaba's customer management revenue rose 5 per cent in fiscal 2026 against total revenue growth of 3 per cent, and the company attributes that gap to improvement in take rate — monetisation rising faster than volume, achieved by charging sellers more effectively rather than by selling more. Amazon's third-party seller services grew 10.26 per cent while its first-party online stores grew 9.01 per cent, and its seller-facing advertising line grew 22.10 per cent. Two of the three disclose rising monetisation; the third discloses rising monetisation in advertising rather than in commission. ## Why the layer outgrows the market underneath it Platform revenue is a product of two variables: third-party GMV, and the take rate applied to it. The GMV layer grows 7.01 per cent a year over 2026–2033 in the [global e-commerce forecast](https://www.pheonixresearch.com/consumer-goods-retail/general-consumer-needs-products/market-report/global-e-commerce-market/). The platform layer grows 8.60 per cent. The 1.6-point difference is take-rate drift. The mechanism is unglamorous and it is visible in every set of results cited above. When volume growth decelerates, a platform has three ways to keep revenue growing: sell more, charge more per transaction, or capture adjacent lines the merchant used to buy elsewhere — fulfilment, advertising, payments, credit, returns. The first option is contested by other platforms. The second is constrained by merchant tolerance. The third is where the growth has actually come from, and it is why a platform's revenue mix drifts toward services as its marketplace matures. The constraint is the same everywhere: a merchant who can move to a composable stack, a regional marketplace or a direct storefront will do so once total cost of the channel crosses their own threshold. Take-rate drift is therefore a slow variable with a hard ceiling, and any forecast that accelerates it is really forecasting merchant inertia. ## Segmentation ### By platform type Marketplaces operating third-party inventory at scale (the largest single group and the source of most disclosed rent); commerce software sold to merchants who keep their own storefront; social and content-led commerce, where the discovery surface and the checkout sit in different companies; quick commerce, where the platform owns the dark store and the courier; and B2B procurement platforms, which are priced per enterprise contract rather than per transaction. ### By revenue line Commission on sale, subscription and platform fees, fulfilment and logistics services, seller advertising, and payment processing. The mix separates two economically different businesses that share the word platform: a software business with gross margins near ninety per cent, and a logistics business with gross margins near thirty. ### By merchant size Enterprise merchants negotiate commission schedules and buy platform services à la carte; small merchants accept standard rates and buy bundles. The same platform can show a materially different take rate against each cohort, which is why blended take rates are a weak comparison across companies. ## How to use this forecast Three questions separate a platform position from a sales position. **What is actually being charged per transaction, and is that rate rising or falling?** **Which revenue lines are rent and which are cost pass-through?** A fulfilment fee charged to a seller for a service the platform delivers at cost is revenue, but it is not pricing power, and a forecast that treats it as such overstates the layer. **What does the merchant's alternative cost?** Take-rate drift is bounded by the switching option, and the switching option is different in every region. ## Sources and method This edition is built on national statistics and company filings, in that order of priority. National accounts establish the size of the sales layer; filings establish the rent layer; primary work establishes the mechanisms that connect them. - US Census Bureau, Quarterly Retail E-Commerce Sales, second quarter 2026: USD 340.2 billion, 17.1 per cent of total retail. - National Bureau of Statistics of China, Total Retail Sales of Consumer Goods in December 2025: online retail of physical goods RMB 13,092.3 billion, 26.1 per cent of retail. - Amazon.com, Inc., Form 10-K for the year ended 31 December 2025: net sales by group of similar products and services. - Alibaba Group Holding Limited, March Quarter and Fiscal Year 2026 Results: revenue by segment, customer management revenue. - Shopify Inc., Fourth Quarter and Full Year 2025 Results: GMV, revenue, gross profit, free cash flow. - Pheonix Research primary programme: structured interviews with marketplace commercial teams, merchant finance leads and fulfilment operators; conference-floor interviews at commerce and retail industry events; and a survey of cross-border sellers. The primary layer establishes how take rates are set, where they are capped and which lines are cost pass-through. It carries no magnitude of its own. **Research split.** Roughly 70 per cent of the evidence in this report is secondary — national statistics, filings, regulation. Roughly 30 per cent is primary research conducted by Pheonix: interviews, conference attendance, open-ended merchant conversations and surveys. Where a primary finding contradicts a published figure, the published figure is reported and the contradiction is stated. ## Table of Contents Table of Contents Executive Summary Platform Revenue and GMV, 2025 Six Cents of Rent on Every Dollar of Sales The Disclosed Seats and the Modelled Residual Take-Rate Drift as the Growth Mechanism Strategic Outlook to 2033 Market Introduction and Overview Definition: Rent Versus Sales What Is Included and What Is Excluded Why First-Party Revenue Is Not Platform Revenue The Four Platform Models How the Layer Relates to the E-Commerce Market Beneath It Research Methodology Secondary Sources and Their Priority Primary Research Programme Disclosed-Seat Plus Residual Sizing Forecasting Assumptions, 2026-2033 Triangulation and Sensitivity Market Dynamics Structural Drivers Restraints on Take-Rate Drift Opportunities in Services Attached to Third-Party Volume Challenges: Merchant Switching and Regional Fragmentation Market Size and Forecast, 2026-2033 Platform Revenue Analysis CAGR and Increment Decomposition into Volume and Take Rate Sensitivity to the Residual Segmentation By Platform Type By Revenue Line By Merchant Size By Region Regional Analysis China North America Europe Latin America Asia-Pacific ex-China Middle East and Africa Competitive Landscape Market Structure and Concentration Disclosed Platform Revenue Benchmarking Take-Rate Comparison and Its Limits Strategic Developments, 2025-2026 Company Profiles Shopify Inc. Amazon.com, Inc. Alibaba Group Holding Limited Salesforce, Inc. Adobe Inc. MercadoLibre, Inc. PDD Holdings Inc. Sea Limited eBay Inc. Etsy, Inc. Value Chain and Fulfilment Economics The Order-to-Cash Chain Company-to-Stage Mapping Value Chain Signals Strategic Implications Forward Outlook Technology and Innovation Landscape Current Scenario Innovation Signals Strategic Implications Forward Outlook Regulatory and Policy Environment Current Scenario Regulatory Signals Strategic Implications Forward Outlook Investment and Funding Dynamics Current Scenario Investment Signals Strategic Implications Forward Outlook Risk Factors and Disruption Threats Current Scenario Risk Signals Strategic Implications Forward Outlook Appendix, Sources and Method About Pheonix Research ## Competitive Landscape Competitive landscape: concentrated where the rent is booked The platform layer is one of the most concentrated markets in the digital economy measured by revenue captured, and one of the most fragmented measured by participants. Three companies disclose 53.01 per cent of global platform rent. Below them sit dozens of regional marketplaces, several large commerce-software vendors, and a long tail of vertical platforms that will never split out a third-party line. Current scenario: the disclosed positions Amazon.com, Inc. Consolidated net sales of USD 716,924m in 2025, up 12 per cent, with net income of USD 77,670m. Third-party seller services reached USD 172,162m, up 10.26 per cent, and advertising services USD 68,635m, up 22.10 per cent. First-party online stores grew 9.01 per cent to USD 269,287m. The competitive fact inside those numbers is that Amazon’s seller-facing revenue is growing faster than its own retail revenue, which is the definition of a platform becoming more platform-like. Alibaba Group Holding Limited Revenue of RMB 1,023,670m, or USD 148,401m, up 3 per cent in fiscal 2026 after excluding the disposed Sun Art and Intime businesses; on a like-for-like basis growth was 11 per cent. Customer management revenue rose 5 per cent to RMB 343,867m, which management attributes to improvement in take rate. Quick commerce revenue rose 47 per cent to RMB 78,520m, and Cloud Intelligence Group rose 34 per cent to RMB 158,132m. The company is simultaneously raising rent on its core marketplace and buying frequency in immediate delivery. Shopify Inc. Merchant GMV of USD 378,441m, up 29.48 per cent, and revenue of USD 11,556m, up 30.14 per cent, with free cash flow of USD 2,007m and a USD 2 billion repurchase authorisation. GMV is three times the 2020 level. Shopify’s own estimate puts its share of the US e-commerce market above 14 per cent, derived from Census retail data rather than from a survey. The regional champions PDD Holdings, JD.com, MercadoLibre, Sea Limited, Coupang, eBay and Etsy each hold a defensible national or category position and none of them splits a comparable third-party line. They are carried in the modelled residual at USD 225,532m and they are the main reason that residual is not a rounding error. What drives competitive intensity The take-rate ceiling. Commission is the most transparent price a platform charges and therefore the easiest to compare. Every basis point above a competitor is a merchant conversation. Subsidised frequency. Quick commerce competition has moved from convenience to price, and the subsidy is funded from platform revenue, which compresses margins before it compresses take rates. Payment attach. The cheapest competitive weapon available to a software platform, because it raises monetisation without raising the headline commission. Machine-mediated discovery. If a model assembles the shortlist, the incumbent’s search advertising inventory is worth less, and the platform’s defensibility shifts to catalogue completeness and delivery truth. Strategic implications Commission is the least defensible revenue line on a platform’s income statement and services are the most defensible. That is not a preference, it is what the disclosed numbers show: the fastest-growing lines across the three disclosing platforms are advertising, payment attach and cloud, none of which is the commission rate. The second implication concerns scale. Fixed compliance cost, fulfilment density and machine-learning infrastructure all reward size, which is why the layer consolidates at the top while fragmenting at the merchant level. A regional platform with 2 per cent of global rent can still be dominant nationally, and usually is. Forward outlook Expect three movements to 2033: consolidation of commerce-software vendors into payment and fulfilment owners; a durable split between capital-heavy marketplace networks and asset-light software platforms; and rising regulatory scrutiny of the services a platform sells to the sellers who depend on its ranking. - **Structure:** concentrated - **Tier 1 Players:** 3 - **Intensity:** high ## Value Chain Value chain and fulfilment economics: where the six cents come from A platform’s six cents per dollar of GMV is not collected in one place. It is collected at four points in a single order: the commission at checkout, the spread on the payment, the fee for storing and shipping the item, and the advertising the seller buys to be found in the first place. Each point has a different gross margin and a different ceiling, and the mix explains why two companies with similar take rates can have very different economics. Current scenario: company-to-stage mapping Mapping named companies to the stage they are actually paid at, rather than to a category, is the only way to read the layer. Stage Who is paid here 2025 evidence Storefront and merchandising software Shopify Inc., Salesforce, Adobe Shopify: USD 11,556m of revenue on USD 378,441m of merchant GMV, 3.05 per cent Marketplace commission Amazon, Alibaba Group, PDD Holdings Amazon third-party seller services USD 172,162m, up 10.26 per cent; Alibaba customer management revenue RMB 343,867m, up 5 per cent Cross-border platform Alibaba International Digital Commerce, Sea Limited Alibaba international commerce USD 20,900m on RMB 144,170m, up 9 per cent Fulfilment and logistics services Amazon, Alibaba, Coupang Amazon’s related fulfilment and shipping fees are booked inside third-party seller services; Alibaba direct sales and logistics revenue RMB 105,518m Payments Shopify, MercadoLibre, Sea Shopify Gross Payments Volume up 37 per cent and Shop Pay GMV up 62 per cent in 2025 Discovery and seller advertising Amazon, Alibaba, Etsy Amazon advertising services USD 68,635m, up 22.10 per cent – faster than any other Amazon line except AWS Immediate and quick commerce Alibaba, Coupang, MercadoLibre Alibaba quick commerce RMB 78,520m, up 47 per cent Where the margin actually sits The same six cents can be ninety per cent gross margin or thirty. A subscription and commission business with a hosted storefront carries software gross margins. A business that stores inventory in its own buildings, employs couriers and pays for the return leg is running a logistics operation with a software front end. Amazon’s 2025 accounts make the split visible: USD 142,352m of net additions to property and equipment in a single year, of which USD 96,496m went to AWS and the balance largely to the fulfilment and transportation network that serves the marketplace. This is the structural reason a platform that owns fulfilment can raise its take rate without touching its commission schedule. The commission is published and competitive; the fulfilment fee is inside the platform’s own service catalogue and priced against a benchmark the seller cannot easily reconstruct. Value chain signals worth tracking Wholesale moving onto consumer platforms. Shopify reported B2B GMV growth of 96 per cent in 2025 off a small base. The rent line is following the order, not the channel. Offline attached to online. Shopify’s offline revenue grew 27 per cent, because the point-of-sale terminal is a rent line on a physical transaction that happens to share a merchant record. Fulfilment built ahead of demand. Amazon’s capital additions doubled between 2024 and 2025. Capacity is going in before the seller invoice that justifies it. Frequency bought with subsidy. Alibaba’s quick commerce revenue rose 47 per cent to RMB 78,520m while the segment’s economics were being rebuilt around delivery density. Subsidised frequency is a purchase of future take rate. Strategic implications For a platform, the defensible position is the stage where the platform holds information the merchant cannot reproduce: where the parcel physically is, whether the card was authorised, whether the buyer returned the last three orders. For a merchant, the correct measure is total cost of channel per order, not commission rate – and total cost of channel includes the advertising needed to be discovered inside the platform’s own search results. Forward outlook Through 2033 the value chain consolidates around two poles: catalogue and identity on one side, physical and payment rails on the other. Platforms that own neither become interchangeable at the margin, which is why the commerce-software cohort is being repriced against payment attach and fulfilment partnerships rather than against seat count. - **Model:** asset_light - **Distribution:** hybrid - **Supply Complexity:** high ## Investment Activity Investment and funding dynamics: capital is going into the parcel, not the page Capital allocation in this layer has bifurcated. Asset-light software platforms are converting operating leverage into cash returns. Marketplace networks are converting it into buildings, robots and delivery density. Both are rational, and they produce very different risk profiles for anyone valuing the layer. Current scenario Amazon: the largest capital programme in commerce Net additions to property and equipment were USD 142,352m in 2025, against USD 85,752m in 2024 and USD 48,344m in 2023. USD 96,496m of the 2025 figure went to AWS, and the North America segment added USD 35,919m. Total property and equipment, net, reached USD 357,025m. This is the cost of guaranteeing delivery promises and of hosting the inference that increasingly sits behind the storefront. Shopify: cash return rather than capacity Free cash flow of USD 2,007m and a USD 2 billion share repurchase programme authorised alongside the full-year results. Ten consecutive quarters of double-digit free cash flow margins. The capital is not going into warehouses because the merchants own the inventory. Alibaba: buying frequency and rebuilding the core Quick commerce investment drove a 47 per cent increase in that segment’s revenue to RMB 78,520m, funded while the company was completing the disposal of Sun Art and Intime. Cloud revenue rose 34 per cent to RMB 158,132m. The allocation decision is explicit: pay for demand today, own the compute layer for tomorrow. Investment signals Fulfilment capacity is being built ahead of the seller invoice that pays for it, which means take rates have room to rise without new announcements. Payment and credit attach is the cheapest acquisition channel a platform has, and it is being funded from operating cash flow rather than from external capital. Private capital has moved to tooling: catalogue enrichment, returns logistics and cross-border compliance, all of which sell into platforms rather than compete with them. Quick commerce is the one segment where capital is being spent on customer acquisition at a scale that would be unusual in marketplace software. Strategic implications The bifurcation matters for anyone underwriting the layer. An asset-light platform is valued on revenue quality and free cash flow conversion; a network platform is valued on capital efficiency and the margin on services. Applying one framework to the other is the most common analytical error in this market. It also determines who can afford regulation and who can afford AI. Both are fixed-cost programmes, and the platforms with the largest disclosed rent are the only ones that can carry them without diluting the merchant proposition. Forward outlook Capital intensity rises in absolute terms to 2033 while falling as a share of platform revenue, because services revenue scales faster than the physical network required to serve it. The exception is immediate delivery, where density economics do not improve with software. - **Trend:** rising - **Capital Intensity:** high - **Recent M&A:** yes ## Technology & Innovation Technology and innovation: the storefront is being separated from the sale The layer’s technology question changed shape in 2025. For two decades the innovation agenda was the storefront: faster pages, better search, higher conversion. The agenda now is the transaction itself – who presents the product, who authorises the payment, and which system holds the authoritative answer about availability and delivery. Current scenario Three technologies are being deployed at commercial scale rather than in pilots. 1. Machine-mediated discovery Product discovery is moving from a search box the buyer operates to a model that assembles a shortlist. Shopify’s own account of 2025 names the investments explicitly – Catalog, Sidekick and a Universal Commerce Protocol – alongside the statement that AI is reshaping how buyers discover and purchase. The commercial consequence is straightforward: when a model assembles the shortlist, the seller’s ranking inside a marketplace is worth less than the machine-readable completeness of its catalogue. 2. Payment attach as a product Shop Pay GMV grew 62 per cent in 2025 and Gross Payments Volume 37 per cent, both far faster than the underlying GMV growth of 29.48 per cent. Payment attach is the most reliable way for a software platform to raise monetisation without renegotiating commission, because the merchant is choosing a convenience rather than accepting a price rise. 3. Fulfilment and returns automation Robotics, automated storage and retrieval, and route density models are the technology that actually changes unit cost. They are also the technology that shows up in capital expenditure rather than in a press release: Amazon’s net additions to property and equipment reached USD 142,352m in 2025 against USD 85,752m in 2024. Innovation signals Agentic checkout. The first platform to make an agent-initiated purchase auditable – who authorised it, who bears fraud liability, who pays the commission – sets the default for the next decade of transactions. Composable commerce. Headless architectures move spend from platform subscription to integration and cloud, which lowers measured platform revenue while raising total cost of ownership. Seller-facing analytics as a paid product. Demand forecasting and pricing tools priced per seller are high-margin rent that does not touch the commission line. Cross-border compliance automation. Duty, tax and product-safety calculations embedded in checkout are becoming a paid platform service, particularly after the withdrawal of low-value duty relief. Strategic implications Technology spend in this layer is converging on the transaction record rather than the shop window. That favours platforms that already hold the payment authorisation and the fulfilment scan, because those two events generate the data an agent needs to complete a purchase. It disadvantages the pure storefront vendor whose differentiation was template quality. The second implication is a measurement problem. A platform that shifts capability into an ecosystem of apps and partners books less revenue per transaction while retaining more control. Reported take rates will therefore understate true monetisation where the platform takes a share of partner revenue rather than a fee from the merchant. Forward outlook By 2033 the base case is that machine-mediated discovery handles a substantial minority of product research, payment authorisation is largely tokenised and invisible, and fulfilment promises are generated by the platform’s own network model rather than by the merchant’s warehouse. The platforms that hold catalog data, payment tokens and delivery telemetry simultaneously will price the transaction. The ones that hold only the storefront will compete on price and lose the argument. - **Innovation:** high - **Patent Activity:** moderate - **Maturity:** growth ## Market Risk Risk factors and disruption threats: five ways this forecast is wrong The 8.60 per cent forecast rests on two variables, third-party GMV growth and take-rate drift. Both can move against the platform layer without any change in consumer demand, which makes the risk section unusually load-bearing for this report. Current scenario 1. Volume deceleration is not uniform US e-commerce grew 12.2 per cent year on year in the second quarter of 2026 while total retail grew 6.7 per cent. China’s online retail of physical goods grew 5.2 per cent in 2025 against total retail growth of 3.7 per cent. The two largest markets are growing at very different speeds, and a platform’s revenue is weighted to the market where it collects. 2. Merchant switching Take-rate drift has a ceiling set by the merchant’s alternative. Composable commerce, regional marketplaces and direct storefronts are all viable at sufficient scale, and every basis point of drift raises the incentive to test one. 3. Concentration Three companies disclose 53.01 per cent of the layer. A forecast this dependent on a handful of disclosed lines is exposed to a single accounting reclassification – a platform moving a service between revenue lines would move the measured market without changing the underlying economics. 4. Cross-border policy The withdrawal of low-value duty relief in 2025 raised the landed cost of imported goods and shifted customs work into checkout. Further changes to de minimis treatment, tariffs or product-safety liability would reduce the cross-border share of platform revenue fastest, because cross-border is where the compliance service fee is largest. 5. Subsidy-driven demand Immediate delivery growth is being bought. Where a platform is paying for frequency, the revenue line grows faster than the contribution margin, and a withdrawal of subsidy would show up as a revenue slowdown rather than as a cost saving. Risk signals Take-rate compression at a disclosing platform would be visible immediately and would falsify the drift assumption. Regulatory anti-steering enforcement in Europe would constrain the highest-margin rent line. Machine-mediated discovery could reduce the value of marketplace search advertising, which is the fastest-growing line at the largest platform. Returns rates rising with category mix would raise fulfilment cost per order without raising the fee charged for it. Strategic implications Readers should treat the disclosed seats as evidence and the residual as an assumption. The correct sensitivity is not to the CAGR but to the residual: a 25 per cent error there moves the 2025 base to USD 423.6 billion or USD 536.4 billion, which is wide enough to change a valuation and narrow enough that the direction to 2033 holds. Forward outlook The most probable disruption is not a new competitor but a change in the surface where the purchase begins. If discovery moves decisively to machine intermediaries, the platform layer’s rent does not disappear – it moves to whoever holds the catalogue, the token and the delivery promise. That outcome is closer to a reallocation than a contraction, and it is the scenario this forecast is built to survive. - **Overall Risk:** high - **Geopolitical Exposure:** high - **Substitution Risk:** high ## Regulatory Landscape Regulatory and policy environment: the platform is now a regulated intermediary Platform regulation has moved from consumer protection to structural conduct. The obligations now attaching to large marketplaces – on self-preferencing, data access, seller terms and product safety – act directly on the revenue lines this report measures, because they constrain how a platform may charge the sellers who compete with its own retail operation. Current scenario Europe The Digital Markets Act imposes conduct obligations on designated gatekeepers, including limits on anti-steering and requirements around data access and ranking transparency. The Digital Services Act imposes notice-and-action duties and, for very large platforms, systemic risk assessment. The General Product Safety Regulation places responsibility for the safety of goods sold through a marketplace on the economic operator, which in practice means the platform must be able to identify the seller. Separate VAT rules and the DAC7 reporting framework create a documented seller trail that tax authorities can now request directly. United States The most consequential change for cross-border volume was the withdrawal of low-value duty-free treatment in 2025, applied first to China-origin parcels and then more broadly. That removes a cost advantage that had been built into the price of low-value imported goods, and it converts a customs question into a checkout question: duty calculation is now a platform service, and someone has to pay for it. China Platform obligations have been tightened around pricing conduct, seller fees and delivery commitments, with particular attention to the immediate-delivery segment where subsidy competition has been aggressive. The regulatory direction is toward fee transparency rather than toward capping take rates. India and other markets Inventory-based models remain restricted in India’s marketplace rules, which shapes the country’s platform layer toward a marketplace-with-logistics model rather than a first-party one. Comparable restrictions and disclosure duties are spreading across Southeast Asia and the Gulf. Regulatory signals Anti-steering obligations constrain the platform’s ability to charge for access to the buyer relationship, which is where the highest-margin rent sits. Seller data access rights reduce the information asymmetry that justifies a services fee. Product-safety traceability forces seller identification at scale, raising fixed cost and favouring platforms with existing onboarding infrastructure. Customs and tax reporting turns cross-border compliance into a product, and product means a chargeable line. Strategic implications Compliance is fixed cost, and fixed cost is a moat. Every new obligation – seller identification, safety evidence, tax reporting, algorithmic disclosure – costs roughly the same to implement regardless of platform size, which is why regulation has so far consolidated the layer rather than fragmented it. The revenue effect is subtler. Obligations that restrict self-preferencing push platforms to separate retail from marketplace more cleanly in their accounts, which will make the rent line easier to see and easier to compare across companies. That is a disclosure improvement for buyers of this research and a competitive exposure for platforms whose retail and marketplace results have been blended. Forward outlook To 2033 the direction is convergence on seller transparency and divergence on liability. Expect the fee schedule to become a partially public document in more jurisdictions, and expect the services attached to it – fulfilment, advertising, compliance – to become the contested ground. - **Complexity:** high - **Approval Pathway:** standardized_commercial ## Frequently Asked Questions **Q: What is the size of the global e-commerce platform market?** USD 480.0 billion in 2025, rising to USD 928.7 billion in 2033 at an 8.60 per cent CAGR. That is 6.04 cents of platform revenue on every dollar of global e-commerce sales. **Q: What counts as platform revenue in this report?** Commissions, subscription and platform fees, and platform services sold to sellers - fulfilment, logistics-as-a-service, seller-funded advertising and payments processed by the platform. First-party product revenue is excluded, because a platform buying and reselling inventory is taking inventory risk, not earning rent. **Q: Why does the platform market grow faster than e-commerce overall?** Because platform revenue is third-party GMV multiplied by a take rate, and take rates drift upward. Third-party GMV grows 7.01 per cent over the forecast; the platform layer grows 8.60 per cent. The difference is monetisation, not volume. **Q: Can take rates be compared across platforms?** Only with care. Shopify monetises 3.05 per cent of merchant GMV because payments and capital products are optional add-ons. Amazon's third-party seller services include fulfilment fees that partially pass through cost. Blended take rates across different revenue mixes are not a like-for-like comparison.