Zhipu AI and MiniMax show early commercial traction after IPOs — but questions remain

Post-IPO performance

Zhipu AI (智谱AI) and MiniMax (迷你马克斯) reported their first earnings since listing in Hong Kong in early January, and the numbers point to rapid top-line growth even as losses widen. Beijing-based Zhipu posted 724.33 million yuan (US$104.8 million) in 2025 revenue, up 131.9% year-on-year. Shanghai-based MiniMax said revenue hit US$79 million, a 159% jump. Analysts say these figures show early signs of sustainable commercialisation of foundation models — but can that momentum be converted into profit?

Business models: enterprise first, consumer flank

Both start-ups run model-as-a-service businesses that sell access to in-house AI models to institutional clients. Zhipu focuses on enterprise deployments — on-premise or cloud — positioning itself as a pure-play model provider rather than bundling models into broader cloud platforms as tech giants do. MiniMax mixes an enterprise model-serving business with consumer-facing apps such as the video-generation platform Hailuo AI and AI companion Talkie, giving it more diversified revenue streams. Reportedly, investor appetite in Hong Kong has remained strong despite the companies’ continuing losses.

Geopolitics, compute and the path ahead

These results come amid a fraught geopolitical backdrop. US export controls on advanced AI chips and broader Sino‑US tech frictions have complicated access to top-tier hardware and talent, pushing Chinese AI firms to rely more on domestic chipmakers and cloud partners. Analysts caution that recurring revenue from enterprises is promising, but scaling margins will depend on securing cheaper, reliable compute and convincing large clients to adopt locally hosted models at scale.

Verdict

Short term: clear demand, fast growth. Longer term: tougher. Can Zhipu and MiniMax turn high-growth revenues into sustained profitability while navigating export controls, competition from big incumbents and the capital markets’ patience? For Western readers unfamiliar with China’s AI boom, these two listings are an early test of whether native players can commercialise large language and multimodal models in a market shaped both by opportunity and geopolitics.