Zhipu, the Beijing lab behind the GLM models that markets itself globally as Z.ai, just delivered the kind of headline every AI startup dreams of: revenue up roughly 400 percent in the first half of 2026, annualized recurring revenue crossing $1.6 billion, and the open-platform and API business growing an absurd 2,736 percent. Read the press release and you would think China's answer to OpenAI had finally cracked the commercialization code. Read the balance sheet instead, and the story gets considerably less comfortable.
Because the same interim report that celebrated a quintupling top line carried a second number the marketing deck was in no hurry to highlight: a roughly two billion yuan loss for the half, the kind of burn rate that eats exits for breakfast. Revenue quadrupled. The losses, while narrowing, did not go away.
The Growth Story Is Real, Until You Read the Fine Print
Make no mistake, the topline surge is genuine and worth taking seriously. First-half revenue reached 953.9 million yuan, roughly $142 million, against a year-ago base that made the percentage jump look almost too neat. The company has pivoted hard toward a cloud-first, model-as-a-service model where enterprises host GLMs either on premise or in its cloud, and that bet is clearly landing.
For context, the full-year 2025 figure Zhipu reported after its January Hong Kong listing was 724 million yuan, up about 132 percent, with total losses ballooning 59.5 percent to 4.72 billion yuan. In other words, the fourfold first-half jump is real, but it is also chasing a cost structure that has yet to show any sign of bending toward profitability.
This is the uncomfortable arithmetic of China's AI price war:
- First-half revenue of roughly 954 million yuan against an estimated 2 billion yuan of losses, a ratio that means the company burned more than it earned all half.
- An annualized recurring revenue target of $2.4 billion by year-end, up from $1.6 billion today, an ambition that assumes the pricing war magically stops undercutting every margin.
- A market value near 550 billion yuan, roughly $71 billion, assigned to a business that analysts are already quietly marking down.
Read those three lines together and the gap between the valuation narrative and the underlying unit economics becomes hard to ignore. Institutions have started lowering their valuation midpoints, both Hong Kong-listed AI bellwethers dropped more than 5 percent in a single session, and Z.ai shares slipped below the 1,000 Hong Kong dollar mark after six consecutive declines. When the people who priced the IPO decide the number was too kind, the marketing deck starts to look a lot less persuasive.
Selling Tokens Like Phone Credit
Nowhere is the gap between hype and reality more visible than in the commercialization pivot Zhipu announced in the days after the report. The company opened a flagship store on Alibaba's Tmall marketplace to sell AI tokens and coding subscriptions, priced up to 1,078 yuan a month, in what one analyst described as mobile top-up style retail. Subscribe to an LLM the way you top up phone credit.
It is a clever distribution move, one that follows DeepSeek and touches consumers directly in a market of a billion phones. But it is also a tell. When a frontier-lab flagship starts hawking tokens like prepaid data, it is a sign the enterprise cloud story alone is not paying the bills fast enough to justify a 550 billion yuan valuation.
The bigger question is whether any of this flips the ledger within a reasonable horizon. Analysts have openly asked whether Chinese AI stars such as Z.ai and MiniMax could remain stuck in losses through 2030, and the interim report does nothing to retire that worry. Fourfold revenue growth is a headline. A capital-intensive business that still spends more than it earns, in an industry where every competitor is racing to undercut the last price, is the reality.
Zhipu deserves credit for one thing: it is honest about which direction it wants to go, refusing in its own earnings commentary to be boxed in as a low-margin utility for the platform giants. The intent is credible. The cash-flow statement, however, remains the disciplinarian, and right now it is telling a far more cautious story than the growth chart. A 400 percent revenue jump is a great reason to celebrate. It is not, on its own, an answer to the question every investor should be asking: when does the spending finally stop being the story?
Comments