Macquarie Group warns that Chinese AI startups Z.ai (Zhipu AI) and MiniMax could remain loss-making through 2030, even as their revenues climb sharply, highlighting the brutal economics of frontier artificial intelligence development. The assessment was delivered by Ellie Jiang, Macquarie's Head of Asia Internet and Software Research, on the sidelines of the Macquarie Asia Technology Conference 2026 on Tuesday, September 8, 2026.

Why It Matters

Jiang identified compute costs as the primary drag on profitability. She noted that China's compute crunch is two to three times more acute than the broader global shortage, a direct consequence of US export restrictions that have curtailed domestic developers' access to Nvidia's most advanced processors. Without affordable, high-performance chips, training and running frontier AI models carries an outsized cost burden that eats into even fast-growing revenue streams.

The Macquarie Forecast

Macquarie said it was 'still modelling loss-making into 2030' for both Z.ai and MiniMax. Jiang described the bank's stance as deliberately cautious: 'We are trying to be a bit more conservative in our estimate,' she said. The projection is notable because it runs counter to the broader optimism surrounding annual recurring revenue (ARR) growth, as more paying users and enterprises adopt generative AI tools from Chinese providers.

Revenue Growth vs. Profitability Gap

Despite the gloomy profit outlook, Macquarie expects robust ARR expansion for both companies. Jiang explained the paradox: 'The flip side of that story is, we also assumed a much higher ARR growth.' For Z.ai specifically, the company had internally planned for ARR of around US$2.4 billion by end of 2026, while Macquarie's own estimate sits closer to US$3 billion — suggesting the investment bank is actually more bullish on revenue than the company itself, even as it remains pessimistic on the path to profit.

The Competitive Backdrop

The findings underscore a structural tension running through China's AI sector. Developers are racing to compete at the technological frontier while operating under chip-supply constraints that their US and European counterparts do not face at the same severity. The cost of API inference, model training runs, and data centre buildout continues to balloon, compressing margins even as top-line growth accelerates. Both Z.ai and MiniMax are among China's most closely watched AI unicorns, and their prolonged loss-making trajectory could test investor patience as the global AI investment cycle matures.

What's Next

All eyes will be on whether tighter US chip restrictions intensify further, which would worsen China's compute disadvantage and push break-even timelines even further out. Domestic alternatives to Nvidia hardware are being evaluated, but analysts have consistently noted performance and ecosystem gaps. How quickly Z.ai and MiniMax can convert surging ARR into operating leverage will define the profitability story well into the next decade.