Going overseas: don't wait until you're big to go to the US and Europe — go there first so you can get big
Wei Zhe says many companies think they'll go to the US and Europe once they're big, but it's exactly backwards — you go there so you can get big. Pick the wrong region and nothing that follows — product 3.0, the channel pyramid — can be built.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
The old cross-border mega-seller playbook is dead
Wei Zhe says the ‘cross-border mega-seller’ model popular in Shenzhen is a distribution-style mindset, running dozens or hundreds of accounts; he has seen companies with over a thousand accounts and hundreds of thousands of SKUs. Ten years ago these companies made money faster than other paths, and plenty of investors loved them. But a distribution-style company is essentially a traffic-style company: the more stores, the more likely you are to get traffic. When a platform's goods aren't plentiful enough, he doesn't object to distribution; once goods are plentiful enough and you're still using distribution to grab traffic, that's bad money driving out good. More critically, there are three red lines: goods not matching the description harms consumers, brushing orders harms other merchants, and slipping in little cards to build private-domain traffic harms the platform. Amazon's flywheel cares enormously about consumer experience, and these behaviours run straight into that protective wall.
— Wei ZheOf the four choices, region comes first
Wei Zhe says it isn't just cross-border e-commerce — all of life is about choice mattering more than effort. He gives four choices: region, channel, product category, brand positioning. On region he has always argued for seizing the US and European markets first, because paying power is strong, internet penetration is high, and more importantly a brand can build momentum there. This isn't something they invented: when Japanese and Korean brands went global back then, they all seized the US and Europe. On category they change their advice every three years: the first three years they backed products common to China and abroad, where the domestic supply chain could sell overseas without modification — Anker's power banks back then were almost identical apart from the voltage; the second three years they advised overseas-exclusive products, because China's supply chain had all gone overseas and China-abroad commonality offered no differentiation; the last three years it's product 3.0 — powered, chipped, algorithmic.
— Wei ZheIt's not that you go to the US and Europe once you're big — you go there to get big
Many companies tell Wei Zhe ‘I'll go to the US and Europe once I'm big’; he says it's exactly backwards — you go to the US and Europe so you can get big. He gives the swimwear company Cupshe as an example: a husband and wife, 50,000 dollars of inventory, and their first stop was the US. Fifty thousand dollars is sales inventory, actually very little, so there's no minimum threshold. Looking recently at AI smart hardware, they start from a thousand or ten thousand products on the Kickstarter crowdfunding platform, even taking prepayment before placing orders, with no capital — that's also a great way to test the water. He even advises big companies entering the US and European markets to learn from small startups. They always say the needle has to be fine enough to pierce through, and both big and small companies should first nail the four singles: single market, single category, single channel, single brand.
— Wei ZheAfter AI, good companies don't just find demand, they create it
Wei Zhe says that with AI, excellent companies shouldn't go looking for demand — they should lead it and create it. He gives the example of a Chinese company making astronomical telescopes: it used to be unpowered, and they electrified it, adding a chip and an algorithm; the algorithm is actually very simple, with GPS positioning, since the movements of the sky follow rules. He wasn't an astronomy hobbyist before but he'd buy one: at home in Silicon Valley you tap once, it automatically aligns Orion for you, tells you three in the morning is the best time, you set it and go to sleep, and the next day it has taken the shot for you. That lets anyone become a very good astronomy hobbyist. Another example: a cross-border class student making curtains; once powered they didn't sell well in the US because American electricians are expensive, so he put a solar panel on the window and turned electric curtains that needed an electrician into something you can DIY. Revenue went from one or two hundred million to four or five hundred million, and profit is now one or two hundred million.
— Wei ZheIn the AI era, still charging subscriptions means you're behind
Wei Zhe says that in the AI era, still thinking in terms of subscription fees means you're behind again; they're all urging companies toward back-end charging, software charging — what's called tokenized. A monthly subscription fee is the dumbest approach, a bit like advertising that used to charge by time rather than by usage. This time AI has brought out the most ideal charging model in all of human education: paying according to your usage of AI software, services or content. He gives an example: Bessie, if you take two weeks of vacation this month, isn't a monthly subscription wasting two weeks? They advise that AI-transformed hardware should be charged like AI, not go back to the old SaaS model of 9.99 dollars a month. One AI reading platform he looked at charges by how many articles you read, and charges nothing if you don't read; the same for smart bike trainers — after topping up, one dollar per ride, two dollars for two hours.
— Wei ZheAdding AI beats AI adding
Wei Zhe has a formulation: AI-plus or plus-AI. He flips the phrase around — what he favours more is plus-AI, not today's AI-plus. The difference: plus-AI takes hardware humans have already accepted and makes it AI; rather than using AI to create hardware nobody has seen before. It's very hard for humans to accept an entirely new piece of hardware. Take Oura making a ring, because you already accepted rings; something that hangs a camera under your neck is harder for me — I've never accepted that hardware. So they're very bullish on adding AI to already-accepted hardware. One capacitive stylus they invested in is number one globally, because everyone already writes on a Pad; it doesn't add another pen, it just makes your pen AI. He also particularly wants to transform one of the top one or two overseas fitness-equipment companies, replacing the fitness instructor on the screen with an AI-generated one — you want a man, it's a man; you want a Black instructor, it's a Black instructor; install two cameras and it can coach your workout with AI.
— Wei ZheChannels aren't laid out flat — they're a pyramid
Wei Zhe says channels aren't about more channels meaning more roads, with each channel parallel or even competing; excellent brands all have a channel pyramid. Like luxury brands and traditional brands, there must be different channel strategies, pricing strategies, and rules about which channel a new product launches in first. What counts as a light-goods channel, what counts as the main waist channel for volume — the pyramid design differs by category and brand. He gives Anker: making power banks and mecha gear seems a bit low, so to enter the consumer-electronics high ground with Soundcore headphones you have to get into Apple flagship stores; whether it really sells a lot doesn't matter, but it can transform the brand — Apple is the pinnacle of the pyramid. Pop Mart has to open on London's Oxford Street; even if it doesn't make money, how much can it lose — but it has to hold that position. He says some companies treat Temu and AliExpress as parallel channels for selling goods, which definitely won't work: AliExpress is a light-goods channel, somewhere to dump stock when Amazon delists you; you can't sell on Amazon and here too, and consumers get confused.
— Wei ZheR&D is the 1, marketing is the 0
Wei Zhe says he really dislikes throwing money at marketing. Many Chinese brands going overseas, or cross-border e-commerce firms, are very willing to add 2% of revenue as marketing spend, but consumers don't get it. He says: do you dare add 2% of revenue as R&D spend? He has met many bosses at the one or two billion scale; at one billion they say 20 million for R&D, and I don't know what for. He says, see, there's the problem — you don't know what for. Your 2% thrown at marketing makes the platform happy and consumers get nothing; but 2% thrown into R&D, consumers get it. He has always firmly believed a good product speaks for itself, and marketing is just — your R&D is the 1, marketing is the 0; if your thing is good you can amplify it, but if you don't have that 1, never put on a lot of 0s. He gives three signals for judging: first, dare to find a benchmark company, definitely a mid-to-upper brand, and crush it by doubling its performance; second, you sell at the same price as it; third, you sell online and it pushes offline channels to come find you.
— Wei ZheAI doesn't just change products, it rebuilds every role
Wei Zhe says two things about AI matter to companies, and big, medium and small firms all need to do them well: first, combining products with AI; second, how the organisation changes through AI. He sees cross-border e-commerce companies as, of almost every industry he's seen, the one that most should use AI to transform every role. Every role should ask what AI can do for you: product selection, content generation, ad-placement choices — all essentially algorithms; before, it relied on human experience, and experience is definitely no match for algorithmic iteration. Customer service AI-ified: there's no customer service anymore, customer service is just you training your customer-service agent; for ad placement you train an agent; product selection, product display, product image and video generation all AI-ified. He says this is the first and only industry he's seen where the entire company's organisational chain can be transformed by AI. He also says what they oppose most is putting lots of people overseas — with or without AI they oppose it — which they call: no unnecessary people overseas.
— Wei ZheDon't just let AI do the work — let it into the decisions
Wei Zhe puts AI and organisation very simply: many companies have a P series and an M series — P is professional, M is management. Many people are used to letting AI do the P work, doing the job; generating images was originally P-series work. But truly formidable companies don't just let AI do P-series work — they gradually let it do the work of a Manager, Director or VP. He gives a real AI pricing case: the company wanted to sell at 35 euros, AI said you can sell at 55, the company really went to 55 and it really sold, with volume about the same as at 35, but earning 20 euros more. He says AI has no courage, but it's bolder than you — it has compared everything across the whole web for you. He also says that even if AI hallucinates, with so many large models in China and the US, they won't all dream the same dream at the same time, so you must cross-validate.
— Wei ZheIn their own words · checked verbatim
Price-performance ratio means: my performance and quality are the same, you've matched me, so I'll compete you down on price. Price-performance ratio is what I dare to define: once I've set it, I don't cut the price; if a rival catches up to the same performance, I add performance.
性价比 制价比是指 我性能质量一样 你跟我做了一样 我就给你卷价格 价性比是我敢于定义的 我定完以后 我不降价了 有同行跟我追到性能一样 我增加性能
Wei Zhe9:08
Many companies tell me, I'll go to the US and Europe once I'm big. I say it's exactly backwards — you go to the US and Europe so you can get big.
很多公司跟我说 我做大了我去欧美 我说正好反了 你去欧美才能做大
Wei Zhe12:29
You say an unpowered product gets copied by someone in a month. Powered? You also have a lot of European and American market access, so you'll lead by about six months. OK, the chip you carry can add another twelve months. OK, the algorithm you carry might lead for a lifetime.
你说不带电的商品 一个月被别人抄袭了 带电呢 你还有很多欧洲美国的准入 你大约会领先六个月 OK 你带的芯片还能再加十二个月 OK 你带的算法有可能领先一辈子
Wei Zhe23:16
For an independent site or independent app, the craft lies beyond the transaction. It has four major functions, and very few Chinese companies — apart from the ones we coach — pay any attention to the other three.
独立站或者独立APP 功夫在交易以外 它有四大功能 中国很少有企业 除了我们辅导的那些 他们根本不重视 另外三功能
Wei Zhe34:26
I have always firmly believed a good product speaks for itself, and marketing is just — your R&D is the 1, marketing is the 0; if your thing is good you can amplify it, but if you don't have that 1, never put on a lot of 0s.
我一直坚信好产品自己会说话 营销只是 你的研发是1 营销是0 你好东西就能放大 你没有那个1 千万不要放很多0
Wei Zhe37:26
Even if they hallucinate, they won't all dream the same dream at the same time, so you must cross-validate.
他们即使有幻觉 不会同时做同一个梦的 所以你一定要交叉验证
Wei Zhe44:30
Truly formidable companies don't just let AI do P-series work — they gradually let it do the work a Manager, Director or VP should do.
真正厉害的公司 不是只是让AI来做P系列干的活 是你逐渐让他做Manager Director VP该干的活
Wei Zhe48:30
Figures
| Years since Jiyu Capital was founded | 15th anniversary | 0:00 |
| Number of Jiyu Capital IPOs | more than 10 | 0:00 |
| Cupshe's starting inventory outlay | 50,000 dollars | 12:29 |
| Smart-curtain revenue after transformation | from one or two hundred million to four or five hundred million | 22:16 |
| Smart-curtain profit after transformation | one or two hundred million | 22:16 |
| Original price in the AI pricing case | 35 euros | 46:16 |
| Price AI recommended in the pricing case | 55 euros | 46:16 |
| Extra earned in the AI pricing case | 20 euros | 46:16 |
| Marketing spend as a share of revenue | 2% | 36:14 |
| R&D spend as a share of revenue | 2% | 36:14 |
Glossary
- tokenized
- Wei Zhe's term for charging by AI usage rather than by monthly subscription.
- price-performance ratio (价性比)
- Wei Zhe's coinage: daring to set a price and not cut it, and when rivals match your performance, competing on performance instead.
- the four singles (四个单一)
- Single market, single category, single channel, single brand — Wei Zhe's advice for overseas companies to focus first.
- product 3.0 (产品3.0)
- Wei Zhe's overseas category advice: unpowered becomes powered, powered gains a chip, chipped gains an AI algorithm.
How to listen
Founders building or preparing overseas brands, cross-border e-commerce operators, and early-stage investors watching consumer going-global.
The opening small talk and the 15th-anniversary congratulations, about 1 minute, can be skipped.