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

Chinese AI firms Z.ai and MiniMax could remain unprofitable until 2030, despite revenue surges, according to Macquarie Group's Asia internet and software research head Ellie Jiang. The reason: the astronomical cost of competing at the technological frontier, particularly the expense for computing power needed to train and run advanced AI models.

Jiang explained that China's compute crunch was two to three times more severe than the global shortage, due to US restrictions on Nvidia's most advanced processors. Macquarie is being cautious in its estimate, saying they are still modeling losses through 2030 for both Z.ai, also known as Zhipu AI, and MiniMax. Despite strong expectations for rapid growth in annual recurring revenue (ARR), Jiang has been more conservative in her projections, estimating Z.ai's ARR at around US$3 billion, compared to Z.ai's own forecast of US$2.4 billion.

MiniMax CEO Yan Junjie revealed last month that his firm's ARR had reached US$800 million in August. Both companies are currently operating at a loss, sparking skepticism about their long-term competitiveness. Jefferies analysts have labeled China's large language model industry as "overcrowded" and suggested favoring full-stack cloud service platforms like Alibaba Group Holding and ByteDance over standalone AI labs.

Alibaba, a major player in the market, owns the South China Morning Post. The disparity between revenue growth and profitability underscores the significant investments required for model training, computing infrastructure, and research and development to remain competitive. While investors have been willing to tolerate these losses during the early stage of AI adoption, Jiang cautioned against assuming that prolonged losses indicate an inability to monetize AI models.

Macquarie's conservative outlook contrasts with the anticipated growth in ARR, driven by more paying users and increased adoption of generative AI. Jiang noted that companies are exploring alternative metrics beyond ARR to better gauge valuation, as the industry is in its early stages of adoption.