Z.AI, the Chinese AI company formerly known as Zhipu AI, reported first-half 2026 revenue of RMB 954 million, a 399.7% increase year on year, according to its interim results published on Monday. The Hong Kong-listed company, which trades under ticker 02513, also narrowed its losses, though its adjusted net loss still came in at RMB 1.964 billion for the six-month period.
The results cap a breakout stretch for the GLM model maker, whose revenue grew roughly fivefold from the year-ago period. As AI Buzz Wire has reported throughout August, the lab has been one of the most aggressive open-weights publishers among China's frontier contenders, and the new filing shows that strategy beginning to translate into commercial scale.
Cloud and API Revenue Drive the Surge
According to Reuters, which reported the filing, and Chinese market commentary tracked by outlets including Futu and SCMP, the growth was led by accelerating cloud commercialization, with income from the company's open platform and API business rising sharply. The company's model-as-a-service offering, which lets developers tap GLM models through hosted endpoints, has become the primary revenue engine as enterprises in China and abroad adopt the models for coding, agentic workflows, and production applications.
The headline growth number comes with important caveats. Bloomberg reported that Z.AI missed lofty sales estimates, as China's bruising AI price war weighed on monetization. A sequence of aggressive price cuts across Chinese model providers has pushed API pricing to a fraction of US levels, capping how much of the demand surge converts into revenue even as volumes climb.
A Narrower Loss, Still a Deep One
The company's adjusted net loss of RMB 1.964 billion (roughly US$277 million) narrowed compared with prior periods, per the figures cited by Lookonchain and other outlets tracking the disclosure. That remains a substantial loss against RMB 954 million (about US$134 million) of revenue, and it underscores the capital intensity of frontier model development: training runs for GLM-series models, compute procurement, and talent costs all weigh on the P&L while pricing pressure limits gross margins.
Still, the direction of travel matters for investors. A near-quintupling of revenue in six months, paired with a narrowing loss, gives Z.AI a credible growth story at a moment when Chinese AI issuers have found enthusiastic reception in Hong Kong public markets.
From Open Weights to Open Register
The earnings report is the latest chapter in a whirlwind month for the lab. On August 26, Z.ai launched GLM-5.3-Flash under the permissive MIT license and confirmed that the stealth model that had been topping community leaderboards under the codename "Ox Alpha" was its own. Two days later, on August 28, the company published the full weights of GLM-5.3 itself on Hugging Face, shipping 141 FP8 safetensors files with a license aimed squarely at hyperscalers, as we covered at the time.
That open-weights posture has functioned as both distribution and marketing: every download is a potential API conversion, and every benchmark win feeds the enterprise pipeline that the H1 numbers now reflect. The strategy mirrors, in some respects, the playbook Meta used with the Llama family, though Z.ai has paired openness for the broad ecosystem with commercial terms targeted at the largest cloud operators.
What the H1 Numbers Mean for China's AI Race
Z.AI's results arrive as China's model developers split into distinct strategic camps. Some labs continue to chase frontier capability with closed or semi-closed releases; others, Z.ai chief among them, are betting that openness drives adoption that eventually converts into durable cloud revenue. The H1 filing is the strongest quantitative evidence yet that the second bet can pay.
The competitive backdrop remains punishing, however. Bloomberg's framing of a worsening "brawl" among Chinese AI providers reflects successive price cuts and an escalating capability race, with each lab shipping faster, cheaper models at a tempo that erodes pricing power across the market. For Z.AI, sustaining triple-digit growth will depend on whether enterprise workloads — coding agents, industry applications, and sovereign deployments — stick with GLM models as competitors match capabilities.
The company's next milestones will be watched closely: whether the loss keeps narrowing as API volumes scale, whether the hyperscaler-focused license of GLM-5.3 attracts a marquee cloud customer, and whether the momentum in Hong Kong-listed AI names survives a more discerning market. For now, the H1 print gives the open-weights camp its clearest proof point that, in China's AI market, giving models away can still mean getting paid.
The Read for Enterprise Buyers
For enterprises weighing GLM adoption, the filing offers reassurance on two fronts. Continuity is the first: a listed company with a widening revenue base is a more durable counterparty for multi-year deployments than a venture-funded lab burning toward an uncertain raise, a consideration that matters when models are being embedded into production workflows and agent infrastructure. Momentum is the second: the API-led growth curve suggests the ecosystem tooling, fine-tuning services, and hosted capacity around GLM models will keep expanding rather than stagnating.
The counterweights are equally real. The price war that Bloomberg flagged means today's unit economics can deteriorate quickly if competitive pressure intensifies, and an adjusted net loss still running at roughly twice revenue leaves little room for error. Investors and customers alike will now look to the second half for evidence that narrowing losses reflect operating leverage rather than one-off accounting effects — and that the August open-weights surge converts into a durable commercial franchise.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
