appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Minxiao Chang

Ben Jiang

Chinese artificial intelligence company Z.ai on Monday reported a 400 per cent increase in first-half revenue, driven by explosive growth in its open platform and application programming interface business, which helped narrow total losses despite higher research and development spending.
Revenue for the six months ended June 30 rose to 953.89 million yuan (US$142 million). Full-year sales were expected to expand 514 per cent to 4.45 billion yuan, according to consensus estimates from analysts polled by Bloomberg.
Beyond its first-half results, Z.ai said that its annual recurring revenue (ARR) – a key metric that projects current monthly subscription income over a 12-month period – had reached US$1.6 billion by the end of August. For comparison, Yan Junjie, CEO of rival firm MiniMax, said last week that his firm’s ARR had risen to US$800 million in August.
Z.ai’s total loss for the six months ended June 30 narrowed 12.1 per cent to 2.07 billion yuan, while adjusted net loss increased 12.1 per cent to 1.96 billion yuan.
Research and development expenses rose 33.6 per cent to 2.13 billion yuan as Z.ai – known domestically as Zhipu AI – continued to invest aggressively in computing power and base model performance.
Shares of Hong Kong-listed Z.ai closed up 9.63 per cent at HK$1,195 on Monday ahead of the earnings release. The stock remains down about 60 per cent from its record high of HK$2,980 reached in June, when the company’s market capitalisation briefly approached HK$1 trillion (US$127.5 billion).
Z.ai and Shanghai-based MiniMax, China’s first two publicly traded large language model developers, saw their valuations soar after blockbuster initial public offerings in January, though both now trade below their peaks.
Top-line performance in the first half was primarily driven by cloud-based deployment services, where revenue jumped 2,736 per cent year on year to 825 million yuan from 29 million yuan.
The segment accounted for 86.5 per cent of total revenue, up from 15.2 per cent a year earlier, reflecting Z.ai’s continuing shift away from customised, on-premise deployment towards cloud application programming interfaces and subscription-based agent products.
Revenue from on-premise deployment fell 20.5 per cent to 128.7 million yuan.
Gross profit increased 163.7 per cent to 252 million yuan, yielding a gross profit margin of 26.4 per cent compared with 50 per cent in the same period last year. Cloud inference generally carries lower margins than on-premises deployment because of its heavy computing overhead.
The company has accelerated product releases this year to convert model capabilities into recurring revenue. It launched GLM-5-Turbo in March, introduced GLM-5.1 in April and released GLM-5.2 in June.
Its latest flagship, GLM-5.3, launched in August and scored 60 on benchmarking firm Artificial Analysis’ Intelligence Index – on par with Moonshot AI’s Kimi K3, widely regarded as China’s benchmark open-weight model.
Z.ai said GLM-5.3 improved coding and long-horizon agent capabilities entirely through post-training, without altering the underlying base model.
The company last week also released GLM-5.3-Flash, a lower-cost multimodal model anonymously tested as Ox Alpha on OpenRouter and OpenCode. Z.ai said the test traffic was served on a large cluster of domestically developed AI chips.
The launches come as Z.ai attempts to evolve from a model-as-a-service provider into an agent platform spanning its ZCode coding tool, AutoGLM and consumer-facing products.
Registered users on Z.ai’s GLM model-as-a-service platform surpassed 7.4 million in August, up 144 per cent from the beginning of the year.
Liu Debin, chairman of Z.ai, said the company had begun work on its next-generation model, which would feature a larger parameter size, longer context window and native multimodal capabilities.
Z.ai was actively exploring overseas expansion through partnerships with global cloud service providers, one executive said. These agreements included local deployments of its open-source models and revenue-sharing arrangements.
However, the transition remains constrained by computing supply and high inference costs.
Z.ai previously acknowledged that shortages had affected service quality during peak periods, while arguing that customers continued to accept price increases because of model performance.
Asked whether the company had sufficient computing power to sustain its model training and inference demands, an executive said Z.ai had been expanding compute capacity “at a very healthy pace” since earlier this year, adding that domestic chips would handle an increasing share of its inference workloads.


Chinese AI firm Z.ai faces reputation hit after users spot unauthorised uploads

Top Chinese AI models make 10% of OpenAI, Anthropic revenue despite high valuations: report

China rejects calls for ‘pacing’ on AI development, fearing it would entrench US tech lead

What is AGI? How China’s AI firms are catching up as OpenAI Astra sparks debate

Z.ai

China’s Z.ai revenue jumps 400% as total losses narrow on explosive cloud gains

Z.ai’s full-year sales were expected to expand 514 per cent, according to analysts’ estimates. Photo: Shutterstock

DeepSeek nears pre-IPO funding round as 2027 market debut takes shape: sources

The DeepSeek logo on a wall in Hangzhou, eastern China's Zhejiang province. Photo: Getty Images

Alibaba signals faster AI payoff, margin gains halfway through US$56 billion capex plan

The Alibaba cloud unit’s growth has not yet peaked, according to analysts. Photo: Shutterstock

Sony Music, Warner accuse Anthropic of ‘blatant theft’ in major new lawsuit

Anthropic is facing a fresh lawsuit over its use of music to train its AI models. Photo: Reuters

Nvidia ships first H200 chips to China, but forecasts no data-centre computing revenue

Nvidia CEO Jensen Huang in Tokyo last month. Photo: Reuters

An Nvidia H200 AI GPU. Photo: Handout

Tech

Big Tech

Tech Trends

Tech War

Business

Banking & Finance

China Business

Companies

Markets

First-half revenue from cloud-based deployment services soared over 2,700 per cent year on year


First-half revenue from cloud-based deployment services soared over 2,700 per cent year on year.


China’s Z.ai revenue jumps 400% as total losses narrow on explosive cloud gains

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}First-half revenue from cloud-based deployment services soared over 2,700 per cent year on year

First-half revenue from cloud-based deployment services soared over 2,700 per cent year on year