Chinese AI Firms Face Losses Until 2030 Amid Compute Crunch
Chinese artificial intelligence companies Z.ai and MiniMax are projected to remain loss-making until 2030, despite substantial revenue growth. This forecast, from Ellie Jiang, head of Asia internet and software research at Macquarie Group, highlights the significant costs associated with competing at the technological forefront.
The primary reason cited was the considerable expense of computing power needed for training and running frontier AI models. Ms Jiang recently told the South China Morning Post that China faces a compute crunch two to three times more acute than the global shortage. This difficulty stems from domestic developers struggling with United States restrictions on Nvidia's most advanced processors.
Speaking at the Macquarie Asia Technology Conference, Ms Jiang stated that Macquarie continued to model losses for Z.ai, also known as Zhipu AI, and MiniMax into 2030. She clarified that their estimate was intentionally conservative. And so, this cautious outlook persists despite expectations of rapid annual recurring revenue expansion.
Annual recurring revenue, or ARR, measures projected monthly subscription income over 12 months. Rapid ARR growth is expected as Chinese AI developers attract more paying users and businesses increase their use of generative AI. Ms Jiang noted Z.ai had planned for ARR of approximately USD 2.4 billion by year-end, while Macquarie's estimate was closer to USD 3 billion.
Z.ai disclosed last week that its ARR reached USD 1.6 billion by the end of Aug. MiniMax Chief Executive Officer Yan Junjie reported last month his firm's ARR hit USD 800 million in Aug. Both companies are presently loss-making. MiniMax's Hong Kong-listed shares closed down 5.59 per cent, while Z.ai took a 10.02 per cent dive, both earlier this week.
These figures fuel scepticism about these AI labs' long-term competitiveness. Jefferies analysts, in a note earlier this week, considered China's large language model industry overcrowded. They favoured full-stack cloud service platforms, such as Alibaba Group Holding and ByteDance, over standalone AI firms. But these cloud giants possess clear advantages.
Cloud providers benefit from superior computing power, extensive data, and strong balance sheets. These allow them to fund AI investment and monetise AI through multiple channels. The increasing gap between revenue growth and profitability demonstrates the substantial capital frontier AI developers must continuously invest in model training, computing infrastructure, and research and development for competitiveness.
Investors have, thus far, tolerated these losses as the market remains in an early adoption phase. Ms Jiang, however, stated that prolonged losses do not signify an inability to monetise AI models. Monetisation is improving, even among Chinese companies developing open-weight models, challenging prior assumptions. So, revenue generation methods are evolving.
Companies can generate revenue via application programming interface usage, enterprise deployments, and commercial licensing arrangements. Increased usage also provides valuable real-world data for refining subsequent model versions. Ms Jiang acknowledged current spending is not fully justified, but noted future justification remains possible.
Z.ai's co-founder, Tang Jie, said last week in Qiushi, the Communist Party's official journal, that computing capacity remains a major company challenge. Expanding application scenarios create a severe shortage of computing power. And HSBC analysts noted last month that MiniMax must increase investment to stay competitive, with revenue growth dependent on securing sufficient computing power.
Ms Jiang stated the industry is seeking alternative metrics to ARR. She explained ARR makes sense in the early adoption cycle due to exponential month-over-month and week-over-week growth. But the market is now examining metrics beyond ARR, which annualises recurring sales for fast-growing software and AI firms.
The search is for more predictable and reliable figures to underpin valuations. These could include metrics such as price-to-sales ratios and gross profit. This shift reflects a maturing industry seeking more traditional financial assessment tools.
Chinese AI firms Z.ai and MiniMax are projected to remain unprofitable until 2030, according to Macquarie Group.
High costs for computing power and US restrictions on advanced processors are significant factors in these losses.
Despite projected losses, rapid annual recurring revenue growth is anticipated for both companies.
Analysts favour large cloud service platforms like Alibaba and ByteDance over standalone AI labs due to resource advantages.
The industry is exploring new valuation metrics beyond Annual Recurring Revenue as the market matures.
Source: SCMP
