Imagine spending billions training a superintelligence capable of rewriting codebases and fixing critical production bugs, only to stick it on a digital shelf next to phone screen protectors and HDMI cables.
That’s exactly what Zhipu AI did. They opened a flagship store on Tmall, China’s Amazon, selling \”Tokens\” for 118 yuan a month. On opening day, search traffic skyrocketed 40x. The result? Zero sales. The internet laughed. The stock dropped 5%, wiping out billions in market cap.
But the analysts chuckling at the \”failed retail experiment\” are missing the entire point. Zhipu doesn’t care about selling monthly passes to consumers. They are putting on a show for Wall Street.
You don’t put a revolutionary intelligence on a shelf next to discounted phone cases to make a quick buck. You do it to stage a multi-billion-dollar magic trick for the capital markets.
Over the last six months, Zhipu pulled off a terrifying financial pivot. They went from relying on slow, high-margin government enterprise projects to making 86.5% of their revenue from cheap, high-volume API calls. Their gross margin tanked to 24.6%. They are bleeding cash, with R&D spending hitting 2.2 times their actual revenue. They needed to prove to investors that their \$1.6 billion ARR (Annual Recurring Revenue) wasn’t just a temporary blip from a new model launch, but a sustainable, retail-ready pipeline.
When you can’t prove your profitability, you prove your distribution. The Tmall store isn’t a shop; it’s a billboard designed to calm down nervous investors.
By putting Tokens on an e-commerce shelf, Zhipu is telling the market: \”Look, our API business is so standardized, we can sell it like phone data plans.\” They are betting that the illusion of retail readiness will mask the reality of their profitability anxiety. And Tmall is playing along, using Zhipu’s \”first-mover\” halo to establish itself as the future home of AI commodities.
But here is the dark, absurd reality of the \”Token Economy\” that no one wants to admit. We are treating AI like a utility—like water, electricity, or mobile data. You buy a 10,000-token package, just like you buy 10 gigabytes of data. But AI isn’t a standard good. One developer might use 10,000 tokens to build a micro-app; another might burn through it debugging a single complex codebase. You are buying an unpredictable brain by the gallon.
We are trying to price a synthetic brain by the fluid ounce, and wondering why the market refuses to drink.
The people buying Tokens aren’t the ones paying for them. A programmer uses the AI, but they pay out of pocket because their boss still thinks AI is a gimmick. And the friction is brutal. You don’t just click \”buy\” on Tmall. You buy, jump to Zhipu’s platform, bind your account, extract an API key, and configure it in your IDE. Every step slaughters conversion rates. Consumers want zero-friction digital goods like Netflix subscriptions. They don’t want to do homework to spend money.
Zhipu is fighting a two-front war. On one side, they are slashing prices with lightweight models to build habit, betting they can drop hardware costs by using domestic chips to escape Nvidia’s pricing monopoly. On the other side, they are pushing up the value chain, betting that one day, AI won’t be sold by the Token, but by the task. If AI can fix a critical production bug, its value isn’t in the 50,000 tokens it consumed, but in the \$10,000 disaster it averted.
Right now, Zhipu’s Tmall store is a ghost town. But it’s a necessary ghost town. It’s a desperate signal to the market that API revenue can be retailized. If Tokens become the new electricity, Zhipu just claimed the power plant. If not, they just wasted time selling a non-standard product to an audience that only wants to watch. The store might have zero sales today, but the audience it was really built for—the capital markets—is definitely buying the story.
FAQ
Q: What is the key takeaway?
A: See the article.