China’s AI Tigers: Who Are They, Are These Companies A Threat to Open AI & Anthropic And Should You Invest?

If you thought the US had the AI race all to itself, think again – China’s so‑called AI Tigers are on the march! What began as a loose cluster of six high‑growth labs has evolved into a broader ecosystem that now includes DeepSeek, arguably the most globally disruptive AI company of 2025–2026.

The Tigers emerged during China’s “War of a Hundred Models”, a period in which dozens of labs attempted to close the capability gap with US leaders. What is striking today is how quickly that gap has narrowed. Benchmarks that once showed a 17–31 point deficit now show a difference of only 2.7 points. China’s open‑weight models have become fixtures in global developer workflows. Its consumer AI apps have reached scale faster than any Western equivalents. And its capital inflows have reached levels that rival Silicon Valley’s most aggressive years.

We thought it was time to take a look at them. Let’s look at tigers individually, then consider China’s broader structural advantages, and ponder whether Australian investors should consider exposure to these companies now that several are listed.

Meet the AI Tigers of China!

1. Zhipu AI

Zhipu AI is the most academically rooted of the Tigers. It emerged from Tsinghua University’s long‑running GLM research program and built its reputation on the GLM‑4 and GLM‑5 series. GLM‑5.2 is widely regarded as China’s most globally respected open‑weight model. It has been praised by senior figures at Vercel, Meta and DeepMind for its reasoning efficiency and its ability to match the performance of closed‑weight US models at materially lower training cost.

Zhipu’s commercial trajectory has been equally notable. It listed on the Hong Kong Stock Exchange and became one of the fastest‑appreciating AI stocks in Asia, rising more than 570% from its IPO. Its valuation passed HK$1 trillion (US$128bn), placing it in the same league as Baidu and Tencent’s AI divisions. The company’s strategy is built around enterprise adoption, open‑weight distribution and a federated ecosystem of GLM‑powered applications. It has positioned itself as China’s answer to Anthropic’s Claude Opus series, but with a more open architecture and a more aggressive approach to developer access.

Zhipu’s growth reflects a broader trend: China’s enterprise sector is adopting AI at a pace that outstrips the US in several verticals, particularly manufacturing, logistics and public‑sector automation. Zhipu’s models are embedded in these workflows, giving it a structural advantage that is difficult for US labs to replicate due to geopolitical restrictions.

2. Moonshot AI

Moonshot AI is best known for Kimi, the consumer agentic AI platform that became China’s fastest‑growing AI application. Kimi’s 2M‑token context window was a breakthrough that allowed users to upload entire books, legal documents and codebases without segmentation. This capability made Kimi a national phenomenon and positioned Moonshot as China’s leading agentic‑AI company.

Moonshot’s funding trajectory has been extraordinary. It raised a US$20bn round at a valuation exceeding US$200bn, making it one of the most valuable private AI companies in the world. Kimi’s models have also found international traction. Cursor, the US agentic coding startup, uses Kimi models for long‑context reasoning and code refactoring. This cross‑border adoption is notable because it demonstrates that Chinese AI is no longer confined to domestic markets.

Moonshot’s competitive positioning relative to OpenAI and Anthropic is clear. It focuses on agentic workflows, long‑context reasoning and consumer‑scale distribution. OpenAI and Anthropic have emphasised safety, enterprise integration and closed‑weight control. Moonshot has taken the opposite path: broad access, rapid iteration and aggressive scaling. This divergence is one of the defining differences between Chinese and US AI ecosystems.

3. MiniMax

MiniMax built its reputation on consumer AI applications, particularly Talkie, a multimodal conversational platform that reached tens of millions of users. Its models are optimised for low‑latency inference and high‑volume consumer interactions. MiniMax’s IPO was one of the most dramatic in Asia. For a brief period, its market capitalisation exceeded Baidu’s, a symbolic moment that reflected the shift from legacy internet companies to frontier AI labs.

MiniMax’s international footprint is stronger than most Chinese AI companies. It has revenue streams in Southeast Asia, the Middle East and parts of Europe. Its multimodal models have been adopted by gaming companies, education platforms and customer‑service providers. MiniMax’s strategy is built around consumer AI rather than enterprise AI, which differentiates it from Zhipu and Baichuan.

MiniMax’s competitive relevance to OpenAI and Anthropic lies in its ability to scale consumer AI at low cost. OpenAI’s consumer products (ChatGPT) have high operating expenses due to closed‑weight inference. MiniMax’s architecture is designed for efficiency, giving it a cost advantage that could become structurally important as global AI usage expands.

4. Baichuan AI

Baichuan AI is the Tiger most focused on domain‑specific enterprise models. Its LLMs are tuned for law, medicine, translation, finance and government workflows. Baichuan’s open‑source releases have been widely adopted in China’s enterprise sector, and its commercial strategy is built around vertical integration rather than broad consumer reach.

Baichuan is preparing for a Hong Kong listing, which would make it one of the few enterprise‑focused AI companies available to international investors. Its competitive positioning relative to OpenAI and Anthropic is more specialised. It does not attempt to match GPT‑5.5 or Claude Opus across general benchmarks. Instead, it aims to dominate regulated verticals where domain‑specific accuracy matters more than general reasoning.

This strategy aligns with China’s broader industrial policy. The government has emphasised AI adoption in manufacturing, healthcare, logistics and public administration. Baichuan’s models are designed for these sectors, giving it a structural tailwind that US labs cannot access due to geopolitical constraints.

5. and 6. 01.AI and StepFun

Let’s deal with these tigers together because their strategic positioning is almost identical. 01.AI, founded by Kai‑Fu Lee, is the Tiger most focused on cost‑efficient frontier models. Its Yi series, particularly Yi‑Lightning, emphasises high‑efficiency reasoning and low‑cost training. 01.AI’s thesis is that the future of AI will be shaped not only by capability but by efficiency. This thesis aligns with China’s broader strategy of building AI systems that can scale without relying on US‑made chips.

StepFun is the Tiger most focused on scale. Its Step‑2 and Step‑3.5 models use trillion‑parameter architectures and are designed for high‑volume inference. StepFun’s developer‑tooling ecosystem has become popular among Chinese startups, giving it a distribution advantage that resembles the early days of Hugging Face.

Both companies illustrate China’s strategy of building models that are competitive with US frontier labs, and are materially cheaper to train and deploy. This combination is one of the reasons China has been able to close the capability gap despite spending 23× less private capital than the US in 2025.

Honourable Mention: DeepSeek

Although not formally part of the ‘Tigers’, DeepSeek has become the most globally disruptive Chinese AI company. DeepSeek‑R1 and DeepSeek‑V4 reshaped global assumptions about China’s AI capability. The company is negotiating external funding at a valuation of US$450bn, which would place it among the most valuable private companies in the world.

DeepSeek’s architecture is designed for efficiency. Its training methods allow it to match or exceed the performance of US frontier models at a materially lower cost – more than 30 times cheaper! DeepSeek’s open‑weight releases have become fixtures in global developer ecosystems, particularly in the US open‑source community. This is one of the most significant shifts in global AI dynamics: Chinese models are now powering US startups, reversing the historical flow of technology.

DeepSeek is the clearest example of China’s ability to challenge US AI dominance, even if is not considered an AI Tiger as all of the above companies are.

Why Have China’s AI Tigers Risen? Its All Down To China’s AI Ecosystem!

China’s AI ecosystem differs from the US in several structural ways. These differences explain why China has been able to scale faster, produce more models and close the capability gap despite lower private‑sector investment.

First and foremost, because China has embraced open‑weight distribution as a national strategy. This approach accelerates adoption, increases developer engagement and reduces reliance on closed‑weight inference. US labs have taken the opposite path, emphasising safety, control and proprietary access. The result is a divergence in ecosystem dynamics. China’s open‑weight models spread quickly, while US models remain centralised.

Moreover, China’s AI supply chain is geographically federated. Beijing focuses on research and regulation. Shanghai focuses on enterprise integration. Shenzhen focuses on hardware, chips and manufacturing. This division of labour allows China to scale AI infrastructure faster than the US, which relies on a more fragmented private‑sector ecosystem. Some looking at the supply chain would note that citation quality lags the US. This is for a good reason: because the tigers focus on commercial adoption rather than academic recognition.

Also consider that China’s AI models are cheaper to train and deploy for several reasons. Domestic chips (Huawei Ascend, Kunlunxin) reduce reliance on Nvidia. Data availability is higher due to permissive norms. Government‑backed data‑centre programs, including a US$295bn national infrastructure plan, provide low‑cost compute. These factors combine to create a cost structure that US labs cannot match.

Should Australians Consider Investing in the AI Tigers?

Several Tigers are now listed or preparing to list. Zhipu and MiniMax are available on the HKEX, and Baichuan is expected to follow. This raises a natural question for Australian investors: should they consider exposure to these companies?

The answer depends on investors’ risk appetite, time horizon and portfolio strategy.

You can’t argue there is no market opportunity for the AI tigers. China’s AI market is growing at roughly 28% YoY and has reached ¥1.2 trillion (US$165bn). The Tigers are central to this growth. Zhipu’s share price appreciation has been extraordinary. MiniMax’s consumer footprint is expanding internationally. Baichuan’s enterprise positioning aligns with China’s industrial policy. These companies provide exposure to the fastest‑scaling AI ecosystem outside the US.

But of course, there are unique risks that must be considered, especially geopolitical risks and valuation volatility. Regarding liqudity – it varies for those that are listed although Australian investors can access HKEX listings easily. If and when those private companies list there could well be a ‘run’ for them, although any such companies are still in hyper‑growth phases rather than stable cash‑flow phases. Portfolio construction should reflect this reality.

Bottom line

China’s AI Tigers represent one of the most dynamic AI ecosystems in the world. They are credible competitors to OpenAI and Anthropic. They are scaling faster than any Western equivalents. They offer exposure to a market that is reshaping global AI dynamics. All should know about them, but that doesn’t mean invest in them. All companies carry risks that must be weighed carefully.