appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Chong Ming Lee

Xinmei Shen

Chinese artificial intelligence companies Z.ai and MiniMax could remain loss-making through 2030 even as their revenues surge, underscoring the enormous cost of competing at the technological frontier, according to Macquarie Group’s head of Asia internet and software research Ellie Jiang.
One reason cited was the expense required for the computing power to train and run frontier AI models. Jiang told the South China Morning Post on Tuesday that China’s compute crunch was two to three times more acute than the wider global shortage, as domestic developers struggled with US restrictions on Nvidia’s most advanced processors.
Speaking on the sidelines of the Macquarie Asia Technology Conference 2026, Jiang said that the investment bank was “still modelling loss-making into 2030” for Z.ai, also known as Zhipu AI, and MiniMax.
“We are trying to be a bit more conservative in our estimate,” she added.
Macquarie’s relatively conservative outlook comes despite expectations for rapid growth in annual recurring revenue (ARR) – a key metric that projects current monthly subscription income over a 12-month period – as Chinese AI developers attract more paying users and businesses increase their use of generative AI.
“The flip side of that story is, we also assumed a much higher ARR growth,” said Jiang. The Macquarie analyst said Z.ai had planned for ARR of around US$2.4 billion by the end of the year, while Macquarie’s estimate was closer to US$3 billion.


Last week, Z.ai disclosed during its first-half earnings that its ARR had reached US$1.6 billion by the end of August. By comparison, MiniMax CEO Yan Junjie stated last month that his firm’s ARR hit US$800 million in August. For now, both companies are loss-making.
The results have fueled growing scepticism about the long-term competitiveness of both AI labs. Jefferies analysts said in a note on Monday that they viewed China’s large language model industry as “overcrowded” and favoured full-stack cloud service platforms, such as Alibaba Group Holding and ByteDance over standalone AI labs. Alibaba owns the South China Morning Post.
These cloud giants had advantages in computing power and data, strong balance sheets to fund AI investment and multiple ways to monetise AI, according to Jefferies. MiniMax’s Hong Kong-listed shares closed down 5.59 per cent on Tuesday, while Z.ai took a 10.02 per cent dive.
The widening gap between revenue growth and profitability highlights the large amounts of money that frontier AI developers must continue to pour into model training, computing infrastructure and research and development to remain competitive.
Investors have so far been willing to tolerate those losses while the market remained in an early stage of adoption, Jiang said.
Still, Jiang said prolonged losses should not be mistaken for an inability to monetise AI models. Monetisation was improving, including among Chinese companies developing open-weight models, challenging the assumption that giving users broad access to model weights prevented developers from making money.
People are trying to find a more predictable, more reliable figure [than ARR] to build the valuation on
Ellie Jiang, Macquarie Group
Companies could still generate revenue through application programming interface (API) usage, enterprise deployments and commercial licensing arrangements, while greater usage could also provide valuable real-world data for improving subsequent versions of their models, she added.
“But it’s not to the point that it completely justifies all the spending,” Jiang said. “That’s where we’re at now, but it doesn’t mean that it won’t justify in the future.”
In an interview last week with Qiushi, the official journal of the Communist Party, Z.ai’s co-founder Tang Jie said that computing capacity remained a major challenge for the company, with an “expanding range of application scenarios” creating a severe shortage of computing power.
In an analyst note last month, HSBC analysts wrote that MiniMax must ramp up investment to stay competitive, noting that any revenue upside hinged on securing sufficient computing power.
Jiang said that the industry was trying to find alternative metrics to ARR.
“The fact is we’re so early to the adoption cycle that only ARR makes sense because things are just growing,” the analyst said. “It’s month over month, week over week, we’re seeing this exponential growth.”
But the market is also beginning to look beyond ARR, a metric widely used by rapidly growing software and AI companies to annualise recurring sales.
“People are trying to find a more predictable, more reliable figure to build the valuation on,” said Jiang, adding that it could include metrics like price-to-sales ratios and gross profit.


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

Kimi maker Moonshot AI to explore dual Hong Kong and Shanghai IPOs to boost capital: sources

Chinese AI firms Z.ai and MiniMax could remain loss-making until 2030, Macquarie says

China’s Moonshot and Z.ai bring AI model subscription race to Tmall’s retail shelves

Z.ai

The Zhipu AI logo, for the firm also known as Z.ai, is seen on a smartphone. Photo: Shutterstock

Tmall is expanding into AI subscriptions as developers seek new ways to reach and retain paying users. Photo: Shutterstock

Why less visibility into how OpenAI’s new GPT-6 Astra ‘thinks’ is sparking safety concerns

OpenAI said the written reasoning of its new Astra model was 'harder to monitor' compared with the previous generation released in July. Photo: Reuters

Frontier AI labs are racing to squeeze greater intelligence out of increasingly expensive computing infrastructure. Photo: Shutterstock Images

Hugging Face opted to run Z.ai’s open-weight model, GLM-5.2, to locally to contain the attack. Photo: Shutterstock Images

Tech race flashpoint: Huang, Musk push back on AI guard rails amid heated US-China rivalry

Jensen Huang speaks at the G20 Innovation Ministerial summit. Photo: Reuters

Elon Musk speaks via video conference to delegates at the G20 Innovation Ministerial summit. Photo: Reuters

US export curbs have reshaped China’s tech scene around ‘chokepoints’: report

A sign for Shanghai’s Star Market, a Nasdaq-style tech board, is seen before a listing ceremony at the Shanghai Stock Exchange in 2019. Photo: Reuters

Tencent’s Hy4 model gains in open-source AI rankings after ecosystem-driven training

Tencent said the full commercial iteration of the Hy4 series would launch later this year. Photo: Shutterstock

The Tencent booth at the 2020 China International Fair for Trade in Services in Beijing. Photo: AP

Tech

Tech Trends

Tech War

Big Tech

As both firms mature beyond early growth, an analyst says the industry is looking for a better metric for valuations


As both firms mature beyond early growth, an analyst says the industry is looking for a better metric for valuations.


Chinese AI firms Z.ai and MiniMax could remain loss-making until 2030, Macquarie says

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}As both firms mature beyond early growth, an analyst says the industry is looking for a better metric for valuations

As both firms mature beyond early growth, an analyst says the industry is looking for a better metric for valuations