Coupang vs Gmarket: South Korea's E-commerce Models Compared
Executive Summary
Coupang's vertically integrated Rocket Delivery faces Gmarket, now being reshaped by the Shinsegae-Alibaba joint venture Grand Opus Holdings. We compare delivery promises, membership economics, seller fees, and the data signals each platform exposes.
The short answer
Coupang is South Korea's dominant, vertically integrated marketplace—NYSE-listed (CPNG since March 2021), with revenue of $30.6 billion in 2024 and roughly $34.5 billion in 2025 per its Wikipedia entry and market-data aggregators. Gmarket, a Korean e-commerce pioneer founded in 1999-2000, has spent the last several years being restructured: eBay bought it for about $1.2 billion in 2009; Shinsegae's emart paid 3.44 trillion won (~$3.4 billion) for it in June 2021; and on December 26, 2024 Shinsegae and Alibaba International announced a 50:50 joint venture valued at roughly 6 trillion won (~$4.5 billion), tentatively named Grand Opus Holdings, combining Gmarket with Alibaba's AliExpress Korea.
The JV is real but its final shape deserves epistemic care. Korea's antitrust regulator granted conditional approval in September 2025, Shinsegae Chair Chung Yong-jin took the board chairmanship of Grand Opus Holdings in November 2025, and reporting at announcement time suggested emart might inject Gmarket capital into the venture in a way that cut its stake toward 40%—i.e., Alibaba potentially holding the majority. We treat ownership percentages as unsettled until confirmed filings appear.
At a glance
| Dimension | Coupang | Gmarket |
|---|---|---|
| Founded | 2010, by Bom Kim (Seoul) | 1999 within Interpark; standalone from 2000 |
| Ownership today | Public (NYSE: CPNG) | Shinsegae/emart-controlled; Grand Opus Holdings JV with Alibaba announced Dec 26, 2024 |
| Regulatory status | — | Korea FTC conditional approval, Sept 2025 |
| Scale marker | $30.6B revenue 2024; ~$34.5B 2025 | No current GMV disclosure; operating losses of 65.4B won (2022) and 32.1B won (2023) |
| User reach | ~32 million users (Mobile Index, cited Jan 2025) | ~5.07 million MAU, last among top five (Mobile Index, cited Jan 2025) |
| Fulfillment model | Vertically integrated: ~200 fulfillment centers, Rocket Delivery | Marketplace + Smile Delivery next-day service |
| Membership | Wow membership, KRW 4,990/month (since Dec 2021) | Smile Club subscription tied to Smile Delivery free-shipping benefits |
| Reach beyond Korea | Taiwan operations; Farfetch acquired Jan 2024 | JV explicitly aims to use AliExpress' ~180-country network for exports |
Ownership: a twenty-year divergence
| Year | Coupang | Gmarket |
|---|---|---|
| 1999-2010 | Not yet founded | Launched within Interpark (1999); standalone from 2000; NASDAQ listing 2006 (GMKT) |
| 2009-2010 | Founded July 2010 | eBay acquires Gmarket for ~$1.2B; NASDAQ delisting |
| 2021 | NYSE IPO, March 11, 2021 (~$4.6B raised) | emart buys eBay Korea for 3.44 trillion won (announced June 24); renamed Gmarket Global (Jan 2022), Shinsegae 80.01% / eBay 19.99% |
| 2023-2024 | First profitable year (2023); Farfetch acquired Jan 2024 | Chung Hyung-kwon (ex-Alibaba Korea, ex-Coupang) joins Gmarket (July 2024) |
| Dec 2024 | — | Shinsegae-Alibaba International announce 50:50 JV, tentatively Grand Opus Holdings |
| 2025 | ~$34.5B revenue, ~$214M net income | FTC conditional approval (Sept); Chung Yong-jin chairs JV board (Nov) |
The pattern is hard to miss: Gmarket has changed controlling hands roughly every decade, while Coupang has compounded under continuous leadership since founding. For data continuity, that means Gmarket series carry more structural-break risk and need more explicit segmentation.
Rocket Delivery versus Smile Delivery
The two platforms embody opposite answers to the same question: should a marketplace own its logistics? Coupang answered yes at extraordinary scale—on the order of 200 warehouses covering roughly 1.9 million square meters, overnight delivery for orders placed before midnight, and Rocket Fresh grocery arriving by 7am. Its own disclosures put 99.3% of Rocket orders arriving within one day. Gmarket runs Smile Delivery, where customers order before 8pm for next-day shipping—a meaningful promise, but one built around carrier partnerships rather than owned end-to-end infrastructure. In June 2024 Shinsegae partnered with CJ Logistics, which handles around 80% of Korean parcel deliveries for AliExpress, a signal of how the JV intends to close the speed gap without replicating Coupang's capex.
- Coupang Wow: KRW 4,990/month bundles free Rocket Delivery, 30-day returns, discounts, and Coupang Play streaming.
- Gmarket Smile Club: subscription free-shipping economics layered on Smile Delivery; historically it drove more than half of payments across Gmarket and Auction within four years of launch.
- Data consequence: Coupang surfaces delivery-promise and freshness fields that can be collected and compared city-by-city; Gmarket exposes marketplace-native signals—seller ratings, fee structures, affiliate mechanics—that Coupang's first-party-heavy model obscures.
Categories and assortment
Coupang spans general merchandise, appliances, fresh food (up to ~8,500 food types via Rocket Fresh), plus streaming, food delivery (Eats), and fintech (Coupang Pay); it added luxury fashion through the Farfetch and Stadium Goods acquisitions in January 2024. Gmarket began as a C2C auction-and-bazaar platform and remains a broadline marketplace: appliances, computers, furniture, collectibles, vehicles, and general goods, with a deep long tail of third-party sellers. For intelligence work this means Coupang data skews toward first-party-consistent pricing and availability, while Gmarket data captures seller-side dynamics—fee sensitivity, rating systems, assortment churn—at greater resolution.
Seller fees, advertising, and the Item Winner wrinkle
- Coupang monetization: commissions plus advertising, with the “Item Winner” system designating the cheapest identical product as the sole listed offer—a mechanism criticized for enabling near-zero pricing and reviewed by Korea's Fair Trade Commission as far back as May 2021.
- Gmarket monetization: seller fees based on selling and starting prices plus advertising; it also runs a significant affiliate program. Its fee increases have historically been contentious—a 2005 store-fee hike drew public backlash.
- Advertising opacity: neither platform breaks out retail-media revenue in the sources reviewed (checked August 21, 2026). Sponsored-share must be measured from observed placements, not disclosures.
Seller economics compared
| Seller consideration | Coupang | Gmarket |
|---|---|---|
| Demand concentration | Highest single-platform reach in Korea (~32M users) | ~5M MAU but strong category-specific intent |
| Logistics burden | Optional fulfillment by Coupang; Rocket eligibility drives visibility | Seller-fulfilled norm; Smile Delivery participation varies |
| Pricing pressure | Item Winner pushes identical items to the floor | Auction/marketplace heritage allows wider price dispersion |
| Export optionality | Limited; Taiwan focus | JV pitch: AliExpress network as an export channel for Korean brands |
Which platform to track for which intelligence use case
- Price benchmarking and availability SLAs: Coupang—first-party consistency makes observations comparable across cities and time.
- Seller-side economics and fee effects: Gmarket—marketplace structure surfaces commission, rating, and assortment-churn signals.
- Quick-commerce and delivery-promise evolution: both, with matched windows; the CJ Logistics tie-up means Gmarket promises may shift quarter to quarter.
- Cross-border flows into Korea: watch the JV entity—AliExpress sourcing plus Gmarket shelf space is precisely the combination regulators conditioned their approval on.
- Korean-brand exports: the JV's stated 1-trillion-won cross-border sales target makes outbound listings a leading indicator worth collecting.
What disciplined collection has to handle
- Entity ambiguity during the JV transition: record whether an observation came from Gmarket, AliExpress Korea surfaces, or shared inventory.
- Item Winner-style deduplication: resolve identical products before comparing prices, or Coupang's collapsed offers will bias spreads.
- Membership-state dependence: Wow and Smile Club change displayed prices and fees; capture session state.
- Structural breaks: segment series around December 2024 (JV announcement) and September 2025 (conditional approval).
- Language and localization: Korean-first interfaces mean parsing fixtures should be built against Korean-language pages, with translated labels kept as analysis-time metadata only.
Bottom line
Coupang wins on scale, logistics depth, and data comparability; Gmarket's value lies in its seller-side transparency and in whatever the Alibaba JV turns it into—which is exactly why it should be tracked now, while the outcome is unresolved. Our Coupang marketplace page and Gmarket marketplace page detail the data points we track on each platform.
Related Articles
Coupang Rocket Delivery Economics: What the Filings Show
August 21, 2026
How Coupang's Rocket stack is priced for shoppers, what sellers and brands should track, and what the unit economics in Coupang's filings do and do not reveal about Korea's dominant e-commerce platform.
Trendyol vs Hepsiburada: Comparing Turkey's Two E-commerce Giants
August 21, 2026
Trendyol (Alibaba-controlled) and Hepsiburada (now under Kazakhstan's Kaspi.kz) define Turkish e-commerce from opposite directions. We compare scale, categories, delivery networks, seller economics, and which platform to track for which intelligence use case.
Meesho vs Flipkart: Two Models of Indian E-commerce
August 21, 2026
Newly public Meesho and Walmart-backed Flipkart represent India's value-segment and mainstream e-commerce theses. We compare business models, seller economics, tier-2/3 reach, and which data signals matter on each platform.
Show us the data you wish existed
Name the websites or apps, fields, locations, and frequency. We'll scope a representative sample and the production feed behind it.
Request a sample
Tell us the sources you need and what decisions the data should support