Что случилось
The South China Morning Post reported that Z.ai’s first-half revenue rose nearly 400% year on year to 953.9 million yuan, while MiniMax reported 283% growth to US$116.6 million. The report also described a large gap in the companies’ August annual recurring-revenue figures, while noting an unverified challenge to MiniMax’s calculation.
The South China Morning Post reported that Beijing-based Z.ai, also known as Zhipu AI, reported first-half revenue of 953.9 million yuan, up nearly 400% year on year. The report said Shanghai-based MiniMax reported first-half revenue of US$116.6 million, representing 283% growth, according to its financial report released the previous week.
The report said Z.ai co-founder and chief scientist Tang Jie stated during an earnings call that Z.ai’s annual recurring revenue had reached US$1.6 billion based on August results. MiniMax founder and chief executive Yan Junjie reportedly put MiniMax’s August ARR at US$800 million.
The South China Morning Post further reported that Yan later acknowledged in a private post-earnings call that MiniMax calculated its ARR by multiplying revenue from a single August week by 52, citing a person familiar with the matter. That account is not independently confirmed in the supplied source text, and the report does not establish whether MiniMax revised or formally defended the figure.
Подробности об источнике: scmp.com ↗
Почему это важно
The reported figures offer an early look at how China’s publicly listed AI companies are converting model demand into revenue after their Hong Kong IPOs. They also illustrate why recurring-revenue claims require careful scrutiny: the South China Morning Post says MiniMax’s figure was based on a one-week extrapolation, according to an unnamed source. The comparison may influence how investors assess commercial momentum and technical competitiveness.
The comparison is significant because both companies entered Hong Kong’s public markets in January while promising to capture growing demand for AI in China and overseas, according to the South China Morning Post. Their reported first-half results suggest investors are receiving materially different signals on near-term revenue growth.
ARR is a projection based on current monthly subscription revenue rather than a directly reported 12-month result. A calculation based on one week may be especially sensitive to temporary usage or billing changes, so the reported figures should not be treated as directly comparable without fuller disclosure.
The source also links investor sentiment to model capabilities and technical benchmarks, but it does not provide scores, methodology, customer numbers or independent financial analysis. The supplied report therefore supports a cautious account of reported commercial momentum, not a definitive ranking of either company.
Интерактивный механизм: как он на самом деле работает
Изучите технологию, лежащую в основе этой разработки, в интерактивном режиме.
Which component of an AI application is the machine-learning model itself?
Что посмотреть дальше
Further company filings, earnings disclosures and independently verifiable subscription data should clarify the revenue and recurring-revenue gap. Watch also for public results, customer adoption figures and evidence of whether the companies’ reported model capabilities translate into sustained commercial demand.
Monitor future audited or regulatory filings for clearer definitions and supporting data behind each company’s ARR figure, including the period used for extrapolation and the treatment of subscriptions, enterprise contracts and usage-based revenue.
Watch for independently reproduced model evaluations. The South China Morning Post says analysts are considering technical benchmarks, but the supplied article does not identify the benchmarks or provide results sufficient to verify claims about relative model performance.
The supplied report does not establish product access conditions, pricing, geographic availability or whether either company’s reported growth is recurring beyond the periods disclosed. Those remain meaningful unknowns for users, customers and investors.