Quiet Moves, Loud Signals
In late 2025, two names surfaced on the Hong Kong Stock Exchange hearing list that weren’t yet household names beyond mainland China. Zhipu AI and MiniMax. Both had filed for IPOs. Both had passed their listing hearings in mid-December—Zhipu on the 16th, MiniMax on the 17th. And both were poised to hit public markets before the likes of OpenAI or Anthropic.
It was not a flashy announcement. No global press tours. No massive model demos. Yet the implications were clear to anyone watching the artificial intelligence race: China wasn’t just training AI models. It was ready to capitalise on them.
A Tale of Two Models
Zhipu AI, which has its headquarters in Beijing, claims to be the most technically orientated language model developer in China, having the GLM-4.7 model, released on December 23, 2025, as open-source, designed mostly for reasoning that is agent-willing and software engineering performance. It is used in enterprise, government, and academic applications.
The company posted 2024 revenue of approximately £35 million (RMB 312.4 million / $44.4 million). Its backers include major Chinese tech entities like Alibaba, Tencent, and Ant Group. The choice to release its models as open source is a clear move to accelerate adoption and ecosystem influence.
MiniMax, based in Shanghai, reported $30.5 million in revenue for 2024 and reached $53.44 million in the first nine months of 2025. Its flagship model, M2, is designed for high-speed deployment and affordability, optimised for generative consumer tools.
The company focuses heavily on end-user applications through tools like Talkie AI and Hailuo AI. It also benefits from significant investment by Tencent, Alibaba, Hillhouse Capital, Sequoia China, and MiHoYo.
Where Zhipu is embedding foundational infrastructure, MiniMax is optimising speed-to-market in consumer and enterprise interfaces.
Why Go Public Now?
There is no waiting or stalling for a bigger global brand story. Nor is the wait for OpenAI or Anthropic to set IPO timelines in the West. Both Zhipu and MiniMax are taking a bet when they file in Hong Kong: that the first AI-modelling IPO will define valuation baselines, attract early global money, and generate stronger strategic alliances in the hands of public ownership.
The Hong Kong Stock Exchange is right on the doorstep of mainland support, and there is a solid success record of tech IPOs. It’s a blend of doorstep regional comfort and genuine global investor access.
As of late 2025, OpenAI’s valuation is reported to be around $157 billion following a late 2024 funding round. Anthropic is approaching a $40 billion valuation. Yet both remain private. This opens a window—and Zhipu and MiniMax are stepping through it.
At present, Zhipu and MiniMax are each valued at approximately $4 billion, making them critical benchmarks for public-market appetite toward pure-play LLM companies.
Disruption by Distribution
Zhipu is not just releasing open models. It is turning them into infrastructure. With over 2.7 million paying API users and annual recurring revenue exceeding RMB 100 million for developer tools, the brand is scaling quickly. Its models support applications in municipal governance, education tech, and private sector SaaS platforms.
MiniMax is executing a different plan: rapid consumer deployment. With more than 70% of its 2025 revenue coming from overseas markets, its tools are built to scale across global user bases. Talkie AI, its overseas voice agent product, is already gaining traction in English-speaking markets.
This dual disruption—one institutional and one consumer-led—is not just reshaping China’s AI landscape. It’s redefining who scales AI first and how that scale translates into revenue.
A Global Shift in AI Narratives
The underlying postulation that AI innovations typically find their inception in the San Francisco Bay Area, towards the West Coast of the United States, slowly but rapidly becomes a thing of the past. In the meantime, OpenAI and DeepMind have remained dominant in model benchmarking and media coverage, yet Zhipu and MiniMax are taking steps to parallel progress alongside.
Zhipu launched numerous versions of the GLM series till the end of 2025. GLM-4.7 with ~400 billion parameters actually surprisingly ranks very highly on multilingual and coding benchmarks. The M2 model, quietly introduced by MiniMax, happened to score well on cost efficiency and throughput performance across open-source leaderboards.
Both companies are part of a broader Chinese AI ecosystem that includes Baidu, SenseTime, iFlyTek, and 01.AI—each pursuing different verticals of AI integration. But Zhipu and MiniMax are unique in being the first to test public markets for foundational models.
Competitive Friction Is Building
Zhipu’s model of transparency—offering open-source access to a flagship model—challenges the closed frameworks popularised by OpenAI. Its pricing models, developer access, and ecosystem integrations aim for deep institutional entrenchment over global media presence.
MiniMax, less vocal in its public roadmap, has prioritised commercial application and velocity. It turns research into API endpoints, chat platforms, and subscription tools in short cycles. This capability to monetise quickly and iterate across real user bases stands out in a landscape dominated by long-cycle lab development.
Neither company is trying to become a direct Western clone. They are building from their own context and local infrastructure needs.
From Local Dominance to Global Attention
While both brands are primarily China-focused today, their listing on Hong Kong’s exchange opens the door for international investors. It also puts them on the radar of global partners seeking AI solutions outside US-dominated ecosystems.
There are still questions about how these companies will operate under geopolitical pressures, IP regulations, and cross-border data policies. But those are not questions unique to China. They reflect the broader uncertainties surrounding AI globally.
What Zhipu and MiniMax offer right now is a working case study in market-first AI scaling. One is optimising for adoption by state and enterprise systems. The other is embedded into digital consumer workflows. Both are outpacing their Western peers in turning models into monetised products and publicly accountable organisations.
Watching What Happens Next
More IPOs may follow. More Chinese model labs may go public. But these two brands are the first.
That alone reshapes the AI IPO narrative. And it challenges the belief that only the US can commercialise large language models at scale.
Whether their stocks perform post-listing is a financial question. But the branding moment is already in motion.
The AI race isn’t waiting. It’s already being priced in.