Chinese AI goes public

China’s AI is moving from training runs to trading floors.

Zhipu AI made a strong market debut in Hong Kong early this morning, becoming the first major Chinese LLM developer to go public. Backed by state-linked funds, Zhipu surged about 15% after its IPO and ended the trading session valued at roughly $7.3 billion. Management said 70% of the proceeds will go toward research and development for its general-purpose AI models.

Zhipu is the largest of China's so-called "AI Tigers," a group of six prominent AI startups, and the first to go public—but it won't be the last. Rival MiniMax is expected to launch its own offering tomorrow, even as US industry leaders such as OpenAI and Anthropic remain privately held and further from an IPO.

The AI playing field

These IPOs serve as an early test of whether public markets believe China’s nascent AI sector can compete globally.

For now, scale still favors US leaders. Zhipu reported $44.7 million in revenue in 2024, versus OpenAI’s roughly $13 billion in annualized revenue last year. Valuations aren’t comparable, either: Zhipu is dwarfed by OpenAI’s reported $500 billion private valuation. They’re not on the same playing field—yet.

Still, size isn’t everything: Despite being placed on the US Commerce Department’s blacklist over suspected ties to China’s military and facing limited access to cutting-edge chips, Zhipu has continued to improve its LLM capabilities, highlighting China’s ability to build its independent AI capabilities without relying on best-in-class US hardware.

China gets its chips

That hardware constraint may be starting to loosen.

In December, the Trump administration approved Nvidia’s H200 chip sales to China in exchange for 25% of the revenue. Now, China is expected to approve some commercial H200 imports as soon as this quarter, with Alibaba Group Holding and ByteDance each expressing interest in ordering over 200,000 units apiece.

For Nvidia, renewed access to China would reopen a major growth market after months of zero market share. For Zhipu, access to more powerful chips could accelerate training and close part of the performance gap, turning its IPO from a symbolic milestone into a potential catalyst for faster model iteration.

The competitive balance won’t shift overnight, but easing compute constraints could make China’s AI race far more credible.—SY