Zhipu AI clears Hong Kong IPO hearing as losses soar on AI compute

Zhipu AI crossing the Hong Kong IPO hearing threshold is an inflection point for China’s foundation model ecosystem. The filings lay bare the economic reality of frontier model development: H1 2025 revenue of just under RMB 200 million against more than RMB 2.3 billion in losses, with over RMB 1.5 billion spent on R&D and roughly RMB 1.1 billion on compute alone.([companies.caixin.com](https://companies.caixin.com/2025-12-20/102395298.html)) This is China’s closest analogue to an OpenAI-style capital machine, now seeking public-market funding rather than relying solely on state-linked or strategic private capital.

Strategically, the listing would institutionalize China’s “AI six tigers” narrative, giving public investors direct exposure to a general-purpose model company rather than only to downstream application or hardware plays. Zhipu’s disclosed metrics—6.6% share among independent Chinese general-model vendors, 8000+ institutional customers, and tens of millions of devices and model downloads—signal both traction and the brutal cash burn behind staying in the top tier.([growthhk.cn](https://www.growthhk.cn/cgo/model/149835.html)) If public markets reward this model, expect a queue of other model labs in China and beyond to follow.

For the global race to AGI, this is another step toward making frontier-model development a capital-markets story, not just a Big Tech or sovereign-wealth game. It diversifies the funding base for non‑US labs and raises the competitive bar for Western incumbents that had grown used to a short list of serious rivals.