Live shopping outside China and South East Asia: buzzword or the future of ecommerce?

01

The thesis: live shopping share depends overwhelmingly on median income

In China, a third of online shopping happens in a live room. In South East Asia, a fifth. In the United States, 0.9%. In Europe, zero point nothing.

Why?

It is not the platforms. The company that owns live selling in Jakarta runs the same app in America. Its US live line: $2 billion out of $1.2 trillion of online shopping. 0.2%. America's biggest live seller is Whatnot, an auction app that started with trading cards.

I went through twenty-one markets. The answer is one number: median income. Tell me what a country earns in a month and I will tell you how much of its shopping happens in a stream.

The reason is supply. Live shopping is a person selling forty items in four hours. The feed is worth opening only when thousands of those hours run every day. In a tier-3 Chinese city, the hour costs less than a coffee in Oslo. In London, it costs the seller's day job. And nobody trained them.

So no rich country ever reached critical supply. Demand there was never tested. The crowd might show up if the choice existed. Nobody knows.

The scoreboard: no live shopping app has gone mainstream in a rich country. Live-first apps do niches. Mainstream apps run live as a side feature.

Coresight forecast $68 billion of US live commerce for 2026. The measured 2025 number: $10 to 12 billion.

Below: where live is actually big, who gets large anyway, and the one switch that changes every number. None of it is what I would prefer to be true.

02

Twenty-one markets: the richer the country, the smaller the live share

Twenty-one markets in four groups. Live is GMV closed in a live room; clips are GMV attributed to shoppable short video; the shop tab and search sit in neither. Online retail is B2C goods GMV, latest full year. Wages are average gross monthly (wage.is; a true median is unpublished in half these countries). Every figure carries a tag: A published, B derived from published components, C my estimate where nothing is published.

Color guide: markets where live is a real sales channel, all with wages under $1,000 a month the exceptions and the open seats wages above $2,400, where live never clears 2%. The Live % and Wage columns carry the same colors because they move together; that correlation is the argument of this piece. Amber in the Clips % column marks rich markets where clips already outsell live. Live = GMV closed in a live room; Clips = GMV attributed to shoppable short video; the Shop tab and search sit in neither. Content splits: TikTok Shop ~50% video / 14%→8% live in the US, ~38% video / 40% live in SEA (Momentum Works); Shopee Video ~10% of GMV; Douyin ~10% short video (industry est). FX: EUR 1.13, GBP 1.34, JPY 150, CNY 7.25, KRW 1,400, MXN 19.3, BRL 5.4. European figures include services; on goods only the shares roughly double and the ranking holds.

The cleanest test: Singapore and Indonesia. Same apps, same year, same TikTok Shop team. Indonesia: 17% live. Singapore: 4% and falling. The wage gap: 23 to 1.

Korea tried hardest. Naver at 3% commission, Coupang, Kakao, thirty years of TV home shopping. Result: 1.7%, against a 2021 forecast of $18 billion. That is what full infrastructure buys at $36,000 a head.

China is not live everywhere either. Hangzhou has 50,000 professional hosts. Yiwu, 1.2 million people, employs 60,000 in live commerce. 60% of new live buyers come from tier-3 cities and below. The streamers live where the hour is cheap.

One trap: share and size are different numbers. The US is 0.9%, a failure as a format, and $10 to 12 billion, five times the big European markets combined. 1% of $1.2 trillion is a business. 1% of Italy is a hobby.

Look at the Clips column. In Germany, Italy, France and Spain the recorded clip outsells the live room. Rich consumers buy from video. Not at a fixed hour.

03

India and Africa: the biggest live markets that do not exist yet

India: $70 billion of online retail, a $334 monthly wage, TikTok banned since 2020. Nigeria and Kenya: no TikTok Shop. Sellers stream on TikTok Live and close on WhatsApp and M-Pesa, unmeasured. The cheap hour is everywhere. The platform is nowhere. When one arrives, the share goes where Indonesia's went.

The arithmetic: 25% of India is $17.5 billion, ten times Europe's live market today. Europe at 2 to 5% of $565 billion is $11 to 28 billion. Europe's failure and India's success are the same number. India's comes with a $334 wage.

The industry sells the share. Buy the absolute. Live shopping converts cheap hours into small baskets, and rich countries have no cheap hours left to feed it.

04

Two ways to sell live: monetize a platform you already own, or go standalone

Every player in this industry runs one of two models.

Model one: monetize an existing platform. Live as a feature inside an app that already owns the audience. The feed delivers 300 strangers at 8pm for free; the ad side pays for them. TikTok Shop began as a way to monetize TikTok's time. Where the hour was cheap, the feature became the mainstream. Douyin is the same story at $550 billion.

The same feature in rich markets stays a tab. TikTok Shop's continental Europe: $565 million per quarter, two thirds of it short video. eBay Live, Amazon Live, Naver: tabs. Same product, opposite outcomes. The difference is the wage around it.

Model two: the standalone. Whatnot, Tilt. No feed, no free strangers. The seller brings the crowd. It matters only where the market is enormous, or where several markets become one room. Whatnot's Europe runs up to 60% international, mostly its US audience buying. Tilt ships domestically.

The players. One table, two families: the feature (in cheap-hour and expensive-hour markets) and the standalone. Year is when the live product started, not the company. The band colors carry the verdict.

One model, two fates, plus the outsiders: the feature where hours are cheap, which became the market the feature where hours are expensive, which stays a tab the standalone, real businesses, structurally small.

05

Why no standalone exists outside community shopping: demand is supply-driven

Notice what is missing from that table: a standalone womenswear live app. A standalone beauty live app. A standalone anything-mainstream live app. They do not exist, anywhere. Why?

Because a standalone's funnel runs on supply. The seller brings the crowd: audiences built on Instagram, on Twitch, at card shows. In community categories the crowd pre-exists, so supply arrives with demand attached. GMV concentrates in a thin top layer of sellers, and the app is small by construction.

Outside those categories, growth needs paid ads. And paid ads hit synchronicity, the core constraint of marketplace theory: buyer and seller must be online at the same minute. A product page is asynchronous; ads can buy its clicks. A live room needs 200 synchronous people; ads cannot buy that. Meta tried with 3 billion users and shut both its live products.

An ad converts only if the app always has a room worth landing in. That takes enormous supply. But live supply is humans, and rich-country humans are expensive. Supply never reaches the level where ads convert, and without converting ads there is no demand to pay for more supply. The loop never closes. The market does not tick.

And synchronicity is only half of the category story. The deeper variable is supply concentration: how many SKUs a category needs before the experience is good. Cards and sneakers are concentrated. A few hundred grails carry a room, and the community knows every one. Skincare, sport and discounted brands sit in the middle: thousands of SKUs. Womenswear is the long tail. The user needs a million SKUs to find her size, her style, her price. The critical mass that makes a paid funnel tick scales with the tail.

Even China obeys: the only standalones that survived Douyin sell concentrated one-of-one goods. Weipaitang, $5.6 billion of jade and watches in 2021.

Will clips change this? Tilt's Clips, Whatnot's Video Listings? Maybe. My answer: no for the long tail. Clips remove synchronicity. They do not remove the SKU requirement. In mid-concentration categories, clips might be enough. In womenswear, where the real volumes are, the critical mass is so large that even clips will not close the loop.

06

Clips: selling without synchronicity, and the two bets it opens

A clip is asynchronous. Nobody must be online at the same minute. The algorithm matches it at any hour. Ads can buy its distribution. One recorded hour sells for weeks.

TikTok shows what that is worth. In Germany, France, Italy and Spain, clips outsell live rooms two or three to one. Same app, same sellers. Nobody has to show up at 8pm.

So Whatnot launched Video Listings and Tilt launched Clips. Two possible payoffs, two different bets.

Bet one: same categories, better matching. Same card sellers and buyers, minus the synchronicity. Fast growth, then a plateau in six to nine months: you match the same two sides better, you do not add a side. A no-brainer to ship, hard to size. Nobody has proven that cards and sneakers, built on the ritual of the room, convert as recorded video.

Bet two: the mid-concentration categories. Discounted brands, luxury, sport, the top skincare sellers: thousands of SKUs, not millions, and no seller could ever feed their live hour. Clips can carry them. A longer growth curve, open to the feature players too, and further from what Whatnot and Tilt are good at. They might cross. They might not.

The long tail stays closed either way. Womenswear needs a million SKUs, a critical mass clips cannot assemble.

Either way, every player ships clips, because clips attack the constraint that money cannot. Does the clip feed the room or replace it? Douyin's drift, live falling while video rises, says replace.

07

Chapeau

Two data points defy the theory and win against my analysis.

Whatnot built a $20 billion company inside an industry going nowhere: $8 billion of GMV in 2025, roughly $1 billion of revenue, H1 2026 bigger than all of 2025. They did not make live work in the West. They found the 1% that works, cards and sneakers, in the one market big enough to forgive it, and took all of it.

Korea attacked the labour cost. Lotte put its counter staff in front of cameras, 60 shows a month. Naver lets any store go live for 3%. The hour was free. The result is 1.7%. A free hour does not fill a room when the viewers earn $2,700 a month and have somewhere else to be.

08

The 2026 rush: incumbents bolt on live, and why it stays a tab

eBay Live in five countries. eBay buying Depop for $1.2 billion. Catawiki Live. Vinted's cheque into Tilt. All correct. All smaller than advertised.

Catawiki streams the last 90 minutes of auctions it already runs; one in four objects sells in-stream. No new supply, no new hour. eBay Live claims 7 to 8 times growth on an undisclosed base; if the base mattered, they would have told you. Depop hands eBay 7 million buyers under 34 and no live tab. Switching one on costs nothing. That is the point.

Vinted has no live product; its fund joined Tilt's round and said nothing. A feature takes the share the grid gives it, which in France rounds to 0.1%. Features monetize behaviour that exists. They do not create the hour.

09

AI hosts: the only real fix to supply, if the technology ever arrives

One thing takes live to 20 or 30% in a rich country: removing the human from the hour.

The technology is not ready, and will not be for a few years. Even in China, humans sell the overwhelming majority of live GMV. The avatars are exceptions.

The exceptions show the direction. JD made its digital humans free in December; 70,000 sellers by March, at a tenth of the cost. Baidu's avatar of Luo Yonghao sold $7.6 million in six hours, more than the man, to an audience the man had built. The avatars convert 30% better at 3am. They are awake.

IF avatars arrive, the chain is mechanical. The hour goes free, so supply explodes: every store runs rooms around the clock, ten times the supply, in every category. The feed becomes worth opening. Demand finally gets its test.

Then two things that never worked start working. Paid acquisition: an ad can land on a live room at any hour, so growth becomes buyable. Retention: a browser always finds a room, so they come back. Supply first. Demand follows.

The threshold is the open question. Thousands of rooms a night, or hundreds of thousands, China's number? Nobody knows. If AI supply clears the bar, a standalone finally has a marketing funnel in mainstream categories, and the feature's one advantage, the free feed, stops being decisive.

If AI does not develop, growth in rich countries stays marginal and clips are the only help. The proof is in the table: the US, with clips fully deployed by TikTok, sits at 1 to 2% of ecommerce. Developing countries sit at 20 to 30%. Fixing synchronicity moves a market by a point. Fixing supply moves it by twenty.

I cannot tell you the switch fires. I can tell you it is the only path to a big number in the West.

10

The forecast: under 5% without AI, Indonesia-like with it

The whole forecast in one table. The detail and the rules follow.

Three rules, read off the market table, applied forward. No vendor CAGRs; those are sold, not measured.

Rule 1, the ceiling. Live share settles at a level set by the cost of the hour. The collectibles exception adds 1 to 2 points in any market large enough to hold a Whatnot.

Rule 2, the speed. Live growth starts at the observed 2025 rate and halves roughly every two years until the ceiling binds. Speed depends on whether a feed platform with live as default is present. Rule 3, the company. Company GMV = market live GMV × model share; revenue = GMV × take rate (TikTok Shop 10–15%, Whatnot ~12%, eBay ~13%); value = revenue × 8–10 growing, × 4–5 once growth is under 20%. The AI switch. If an avatar ever costs under $3 an hour and converts at or above a human (JD claims it off-peak; treat that as marketing until replicated), Rule 1 stops applying and every band's ceiling moves to 20–30% by 2035. The technology is not there today, so I give the switch 15% before 2028 in a rich market, 50% before 2031. The green columns below are a scenario, not a schedule, and they assume it fires by 2028.

Market sizes, 2025 to 2035

rich markets stuck under 5% without AI the swing rows: one platform decision moves $65B of India's 2035 GMV conditional on the AI switch firing by 2028. Online retail growth: China 6%/4%, SEA 13%/8%, US 7%/5%, Europe 4–5%/3–4%, Brazil and Mexico 12%/8%, India 22%/13% (Redseer), Africa block 12%/10%. AI-half shares: China 45%/55%, SEA 35%/45%, US and Korea 8%/25%, UK 6%/22%, Germany and Japan 5%/20%, France, Italy, Spain 4%/18%, Brazil 15%/35%, Mexico 12%/30%, India 25%/40%.

• Without AI, Europe's five big markets reach $17B in 2035, 1.7%: what Indonesia does in 2028. Clips soften these numbers; they do not change them.
• Without AI, the US reaches 4.4% and $97B by 2035: still the second-largest live market on earth, still a rounding error as a share.
• With AI the rich world does in 2035 what Indonesia does today. The AI column is the size of the prize if the hour disappears, and it is why everyone keeps funding a format that doesn't work.

11

Who captures it: company sizes in 2030

The market table above says where the GMV will be; this one says who captures it. The companies that matter: 2025 GMV, my 2030 GMV without AI, the revenue and value that implies, and 2030 with AI (green, conditional on the switch). Whatnot's growth decays from +125% in 2026 to +12% in 2030; TikTok Shop's from +92% to +15%.

12

The deals: who buys whom by 2031, with probabilities

The transactions I expect or consider live, with window, my probability (green ≥50%, amber 25–49%, red <25%), price band and logic.

13

Four things to watch, in the order I expect them to move

• TikTok Shop's US live share stops falling. Human supply cannot reverse it at US wages, so the first quarter it turns up, AI made the room cheap. The switch, visible in public data.
• Whatnot or Tilt ships an avatar that runs a room alone for an hour. The day a standalone gets infinite supply, it finally gets a marketing funnel.
• Amazon puts a talking host on a listing. The day it streams, the US forecast doubles.
• Norway. When a $90,000-a-head country sells a third of its online retail live, this article is wrong, and I will write the retraction myself.

If two of the four move, the switch is firing. Supply explodes, demand finally gets its test, and by 2035 the rich world does what Indonesia does today: the United States at $550 billion and 25% of its ecommerce, Europe at $190 billion and 19%, ten new billion-dollar companies.

If none move, this is the future: Europe at 1.7% in 2035, the US at 4%, India at 25% if someone is allowed to build it, and clips as the only help. Live shopping stays what it is today, a niche that every year gets called the future of shopping.