Let me start by disagreeing with the title of this article. Vending machines in China are not an unexplored channel. They were unexplored around 2015. Then billions of yuan of venture money poured in, thousands of operators launched, and most of them died. What is left is a real channel, but a hard one: low margin, capital hungry, and almost entirely decided by where the machine stands. I still recommend it to some brands. I talk many others out of it. This article explains which one you are.
I am Olivier Verot, founder of GMA. I have lived in Shanghai since 2012 and I have sat in the negotiation room with Chinese vending operators for food, drink and personal care brands, on both the machine-ownership side and the slotting-fee side. What follows is what those contracts actually look like, not what the trade press says about them.
The channel is real. The gold rush is over.
The mechanics that made everyone excited a few years ago are still true and still impressive. QR codes and mobile wallets removed coins from the equation completely. You scan, you pay, you take. Some machines read your palm. Restocking is routed by software, and the newer smart cabinets use cameras to charge you after you close the door instead of dropping a can down a chute. Compared to a European vending machine that eats your two-euro coin, this is another world.
But look at the numbers instead of the technology. China’s vending machine market was worth around RMB 24.5 billion in 2024, up roughly 19% year on year, according to Zhiyan Consulting. That sounds healthy until you notice the size. RMB 24.5 billion is a rounding error next to Chinese e-commerce, and it is split between hundreds of regional operators.
Then look at the market leader. UBOX (友宝在线, 2429.HK) is the biggest vending operator in the country and the only listed pure player. For the full year 2025 it reported around 67,840 points of sale, up just 1.0% year on year, revenue of about RMB 2.76 billion, down 5.5%, and it was still loss-making at roughly RMB 68.7 million, per its results reported by Sina Finance. Read that again. The category leader, after more than a decade, with tens of thousands of machines, is growing its footprint by one percent and shrinking its revenue.
That is not a dying channel. It is a mature, consolidated, thin-margin channel. Which is a very different sales pitch from the one this article originally made.
Instant retail ate the emergency purchase
Here is what nobody predicted in 2020. The core use case of a vending machine is the unplanned purchase: I am thirsty, I am here, I want it now. In China, that job has been taken over by delivery.
Instant retail, the 30-minute delivery of ordinary supermarket goods from local stores and front warehouses, reached about RMB 780 billion in 2024 and was expected to pass RMB 1 trillion in 2025, according to People’s Daily. Meituan Instashopping, Taobao Shangou and JD’s on-demand service now fight to put a cold drink in your hand faster than you can walk downstairs. When a consumer can get any of forty thousand SKUs delivered in half an hour, at platform-subsidised prices, the machine in the lobby with sixty facings loses a big part of its reason to exist.
This is the single most important change since this article was first written, and it reframes everything. A vending machine no longer wins on convenience in general. It only wins where delivery cannot reach or is too slow to bother with. That is a much smaller map, and it is a map you can actually draw.
The places where machines still print money are closed environments. Office towers where nobody wants to wait fifteen minutes for a rider at the security desk. Factory floors and campuses where phones and deliveries are restricted. Gyms, where the purchase happens sweating, two metres from the treadmill. Hospitals. Airports and train stations past security, where the captive-audience premium is enormous. Schools and university dorm blocks. Notice the pattern: friction between the consumer and the alternative. No friction, no machine.
Who actually owns the machine, and who gets paid
Most foreign brands I meet have the model completely wrong in their head. They picture buying a hundred machines with their logo on them and placing them around Shanghai. Let me walk through how the money really moves, because this is where the decision gets made.
There are three parties. The operator owns the machine, stocks it, maintains it, and holds the payment account. The site owner, meaning the building manager, the mall, the gym chain, the metro concession holder, controls the square metre where the machine stands. The brand supplies the product. In the vast majority of deals, these are three different companies.
The site owner gets paid in one of two ways, sometimes both. Either a fixed monthly fee, which for a good office tower in a tier-1 city is real money, or a share of gross sales taken off the top. In the deals I have seen, that revenue share sits in the low double digits for ordinary locations and climbs steeply for the trophy spots. Airports, metro platforms and hospitals are usually awarded by tender, and the winning bid can absorb most of the gross margin on the machine. That is why the same brand of water costs twice as much behind airport security.
The brand pays in a different currency. You sell your product to the operator at a wholesale price, and on top of that you very often pay a listing fee per SKU, per machine, per month, to hold a slot. A machine has somewhere between forty and eighty facings. Every one of those slots has an opportunity cost for the operator, who knows exactly how many yuan per day per slot a Coca-Cola or a Nongfu Spring bottle generates. Your unknown foreign brand is competing against that number. If your product turns slower, you are removed within a quarter. No conversation, no warning, the route driver simply stops loading you.
Now, the question everybody asks: should we buy our own machines? Almost never. Owning the fleet means you take on the hardware capital cost, the site negotiation in Chinese with hundreds of separate landlords, the restocking logistics, the payment licence questions, the cash flow of stock sitting in metal boxes across a city, and the maintenance of a machine that is out of order and losing money in a building you have never visited. Operators exist because that work is a business of its own, run on route density. One driver serving forty machines in one district is profitable. One driver serving your six machines is not.
Owning your own machines makes sense in two cases only. First, when the machine is the product experience rather than the distribution: a branded cabinet in your own flagship store, a temporary installation at an event, a fresh-food or made-to-order machine that no operator can service. Second, when you already have a physical network in China with staff who can restock, which is to say when you are Luckin Coffee and you have tens of thousands of stores. Which brings me to the case study this article used to celebrate.
The Luckin lesson, told honestly this time
The original version of this article held up Luckin Coffee’s 2020 vending push, the Luckin Coffee Express and Luckin Pop Mini machines, as proof that vending was the future. A few months after that announcement, Luckin was exposed for fabricating a large part of its sales, was delisted from Nasdaq, and the vending expansion quietly stopped being the story. Luckin later rebuilt itself into the largest coffee chain in China, but it did so with physical stores, an app and aggressive pricing, not with machines in bus terminals.
I am keeping the example because the useful part survives, it just is not the part we thought. What Luckin got right was that the machine was never a standalone shop. It was a node attached to an app that already had the customer, the payment method, the coupon history and the order data. The machine did not acquire anyone. It served people the brand had already won elsewhere, and it did so at locations chosen from that app’s own data. That is the correct mental model for a vending machine in China in 2026: a fulfilment point for demand you created somewhere else. If you have not created the demand, the machine is a very expensive shelf.
Read more on that story: why Luckin Coffee did not outrank Starbucks in China.
Where a foreign brand should still say yes
Despite everything above, I do push some clients into this channel. Four situations justify it.
- Sampling and trial, not sales. Treat the machine as paid media with a product attached. A cabinet in twenty gyms that puts your energy drink in eight hundred hands a week, each scan captured into your WeChat account, is cheaper per trial than most advertising and infinitely more convincing. Judge it on cost per trial and on how many of those scans you keep, not on the RMB it rings up.
- Genuine impulse products. Single-serve, consumed on the spot, under roughly RMB 20, no explanation needed. Drinks, functional shots, snacks, ice cream. If your product needs a paragraph of education, it will not sell from a machine. Skincare in a machine is mostly a photo opportunity.
- Closed venues where you already have a relationship. If you supply a gym chain, an international school group or a corporate campus, a machine is an easy add-on to a contract that already exists, and the site fee negotiation is a conversation instead of an auction.
- Proof of offline traction for a distributor. This is the underrated one. Chinese distributors and buyers want evidence a foreign product moves. Sell-through data from a few hundred machines, by location type, by hour, by price point, is exactly the kind of evidence that ends a stalled negotiation with a supermarket buyer.
And the flip side, when I say no: if you want national coverage fast, if your unit price is above RMB 50, if your margin cannot survive a double-digit site cut plus a listing fee, or if nobody in China has heard of you yet. That last one is not negotiable, and it is the whole point of the next section.
Nobody buys an unknown brand from a machine
This part of the original article was right and I am keeping it, sharpened. A vending machine gives a consumer about two seconds and no salesperson. There is no shop assistant, no sampling girl, no packaging story to read. The purchase is decided by recognition alone. Either the buyer has seen the brand before, or their thumb keeps scrolling to the row with Nongfu Spring in it.
The brands sitting in Chinese machines today did not get there by emailing distributors. They built recognition first and the operator came to them, or at least took their call. So the marketing work is not something you do after you win the slot. It is the thing that wins the slot, and then the thing that makes the slot pay.
Xiaohongshu is where the verification happens
Watch a Chinese consumer standing in front of a machine with an unfamiliar product in it. A good number of them will search the name on Xiaohongshu before paying, or right after. Xiaohongshu now works as a search engine for consumer goods, not just a feed. If your brand name returns nothing, or returns three posts from your own official account, the purchase does not happen and the repeat purchase certainly does not.
The mechanism to fix that is volume of ordinary posts, not one big campaign. Fifty to a hundred KOC notes from real small accounts, written with the words people actually type, including your Chinese name and the category term, published over weeks rather than one day. The platform ranks by engagement density and freshness, so a slow steady stream beats a spike. Budget for that before you budget for machines.
The QR code on the machine is the real asset
Every payment scan is a chance to convert a stranger into a contact you own. This is where the WeChat mini-program earns its keep, and it is the one piece of the original article that has become more important, not less.
The setup is simple. A sticker on the machine offers something small and immediate: one yuan off this purchase, a chance at a free second unit, a coupon for the same machine tomorrow. Scanning opens your mini-program, which asks for nothing more than a tap to follow your official account. Now you know which building that person buys in, at what time, and what they chose. You can push a coupon on Thursday afternoon at 3pm to everyone who buys in that tower. You can invite them into a WeChat group tied to their office complex. That private domain contact list stays yours when the machine contract ends, which the sales in the machine do not.
If you have not built that layer yet, start with the basics: how to launch a WeChat mini-program, and what has changed with WeChat Mini Shop for foreign brands.
Douyin, and the geo-targeting nobody uses properly
Douyin remains the cheapest way to reach a young Chinese audience at scale, and its recommendation engine will find the demographic you describe. For vending specifically, the useful feature is location targeting. You are not trying to reach all of China. You are trying to reach the twelve thousand people who work in three office towers. Running short videos to a radius around the buildings where your machines actually stand, at the hours people get thirsty, costs very little and closes the gap between awareness and the machine in the lobby. Most brands run national campaigns and wonder why the machines do not move. More on the platform here: Douyin marketing trends.
AI answers now sit between you and the operator
Something new since this article was written. Procurement managers and site operators in China increasingly ask DeepSeek or Doubao about a supplier before replying to it. The answer these models give is assembled from Chinese-language sources they can read: Baidu-indexed pages, Xiaohongshu, Zhihu, WeChat articles, trade media. If everything about your brand exists only in English on your own website, the model says it has little information, and that reads as risk to the person asking.
Fixing this is unglamorous. Publish Chinese-language pages that state plainly what you make, where, since when, which certifications you hold and who else distributes you. Get mentioned in Chinese trade media and on Zhihu. Keep your Chinese brand name consistent everywhere, because a model cannot connect three different transliterations. This is the same work as SEO, aimed at a different reader.
Case study: Zsofia, and the hundred machines she did not buy
Zsofia is Hungarian. She founded a protein and functional snack brand in Budapest, strong in Central European gyms and pharmacies, and she came to us convinced that vending was her way into China without the cost of a distributor.
She had already started before we met. She had signed with a small Shanghai operator for forty machines of her own, branded in her colours, placed in a mix of mall corridors and residential compounds. Nine months in, the average machine was selling a handful of units a day. She was paying site fees on every one of them, plus restocking, plus the machines themselves. She was losing money every month and had roughly RMB 400,000 sunk into hardware sitting in the wrong buildings.
Two things had gone wrong, and they were not the machines. First, the locations. Malls and residential compounds are exactly the places where instant retail delivery wins, and where a convenience store is thirty metres away. Second, nobody in Shanghai had ever heard of her brand. A twenty-eight yuan protein bar with Hungarian text on the wrapper, in a machine, next to a familiar Chinese brand at nine yuan. There was no reason for a thumb to stop.
What we changed. We got her out of machine ownership and into a supply agreement with an established operator that already held contracts with gym chains and office towers, paying a listing fee for slots instead of carrying hardware. Fewer than eighty machines, all of them in gyms and corporate buildings, no malls, no compounds. In parallel we ran a Xiaohongshu programme with fitness KOCs over four months, Chinese name fixed and used consistently, plus Douyin video targeted to a radius around the specific buildings. Every machine carried a sticker: scan, join, get the second bar at half price. The mini-program captured the contact and the location.
Six months later, sell-through per machine was roughly four times what her own fleet had managed, the gym locations were carrying the corporate ones, and she had built a private domain list in the low thousands with the building each person buys in. She still does not make meaningful profit from the machines themselves, and I want to be clear about that. What she got was a WeChat customer base she owns, real sell-through data by location type, and, on the strength of that data, a conversation with a national sports-nutrition distributor that had ignored her for a year. That conversation was the actual return.
If distribution is your real goal, read this next: the distribution mistakes to stop making in China.
What I would tell you in a meeting
Vending in China is a distribution tactic, not a market entry strategy. It works when it is attached to demand you already built, in venues where delivery cannot compete, through an operator who already owns the locations. It fails when a foreign brand treats it as a shortcut around brand building or around distributors.
So the honest sequence is boring. Get known on Xiaohongshu and Douyin first. Set up the WeChat layer so the scans are worth something. Then, and only then, go to an operator with an argument they care about: here is the demand in your buildings, here is the price point, here is the turn rate we expect. Do it in that order and the machine becomes a useful piece of your setup. Do it backwards and you buy forty metal boxes, like Zsofia did.
FAQ
Should we buy our own vending machines in China or go through an operator?
Go through an operator, in almost every case. Owning machines means you take on hardware cost, site negotiations in Chinese with every individual landlord, restocking logistics, maintenance and stock sitting idle across a city. Operators are profitable because of route density, one driver covering dozens of machines in one district. You will never reach that density with a small fleet. Own machines only when the cabinet is part of the experience, in your own store or at an event, or when you already run a physical network in China with staff on the ground.
How much does it cost to get a product into vending machines in China?
There is no single price, and be suspicious of anyone who quotes one before seeing your product. You sell to the operator at wholesale, and you typically pay a listing fee per SKU per machine per month on top. The site owner takes either fixed rent or a cut of gross sales, and premium locations such as airports or metro stations are awarded by tender at levels that can absorb most of the margin. Budget separately for the marketing that makes the product turn, because an operator who sees weak sell-through will drop you within a quarter.
Working with GMA on this channel
We are a Shanghai agency and we handle the part that decides whether a machine pays: making the brand recognisable before the slot is negotiated, then converting every scan into a contact you keep.
Concretely, that means Xiaohongshu and Douyin visibility around the buildings your machines stand in, the WeChat mini-program and private domain layer behind the QR code, and the Chinese-language content that makes operators and distributors take your call.
If you are weighing this channel against a classic distributor route, tell us the product, the unit price and the city. We will tell you honestly which one we would do. Contact us here.
Written by Olivier Verot, founder of GMA, Shanghai.
