Artificial Intelligence thread

[Randomuser]

Please, Log in or Register to view URLs content!

Chinese AI Chip Maker Biren Technology Plans $623 Million Hong Kong IPO​

The company expects its shares to start trading in Hong Kong on Jan. 2​

Chinese AI chip maker Shanghai Biren Technology seeks to raise up to about US$623 million in a Hong Kong initial public offering, laying the groundwork for what could be another strong year of listings in the city, dominated by artificial-intelligence companies.

Biren Technology, one of China’s “Four Little Dragons” in the graphics-processing-unit sector, aims to raise as much as 4.85 billion Hong Kong dollars, equivalent to US$623.3 million, by issuing 247.7 million shares at between HK$17.00 and HK$19.60 each. The Shanghai-based company expects its shares to start trading in Hong Kong on Jan. 2.

A successful listing could spur more Chinese AI-related companies to go public in a city that is experiencing an IPO revival after a yearslong drought, turbocharging the Hong Kong bull market further.

Over the weekend, Chinese generative AI startup MiniMax Group filed its documents with the city’s exchange operator after passing the Hong Kong Stock Exchange’s listing hearing.

Another large-language-model developer, Knowledge Atlas Technology, better known as Zhipu AI, is also racing toward an IPO in Hong Kong, according to a document submitted Friday.

Neither MiniMax nor Zhipu disclosed a timeline for their offerings or the amount they plan to raise.
All three AI companies are in the early stages of commercialization and haven’t turned a profit due to significant initial investment in research and development.

Chinese AI startups have accelerated their efforts to go public and tap capital markets for funding amid intensifying competition in the AI space. Beijing has also made technological self-sufficiency a top priority for its next five-year plan, and investor enthusiasm for all things AI remains high in China.

Biren’s IPO plan follows the blockbuster debuts of peers

Please, Log in or Register to view URLs content!

Technology and MetaX Integrated Circuits—also among the Little Dragons—in Shanghai. Moore Threads’ shares surged 425% on the first trading day, while MetaX’s jumped almost eightfold.

All this comes as Hong Kong is set to reclaim its global IPO crown this year. Funds raised through IPOs in the first 11 months of 2025 totaled HK$259.4 billion, more than tripling from the previous year, according to HKEX data.

Paul Lau, head of capital markets and professional practice at KPMG China, said he expects another strong year of listings for the city in 2026, with AI-related IPOs poised to accelerate as the technology matures and is adopted more widely across industries.

 

[bsdnf]

Zhipu and Minimax are currently undergoing IPO procedures on the Hong Kong Stock Exchange. To attract attention, they released limited test versions of GLM 4.7 and Minimax-M2.1, both have significant improvements compared to their predecessors.

However, their financial prospectus doesn't look so promising. While user growth is rapid, R&D and advertising costs are high, resulting in substantial losses, similar to OpenAI.

 

[tamsen_ikard]

Zhipu and Minimax are currently undergoing IPO procedures on the Hong Kong Stock Exchange. To attract attention, they released limited test versions of GLM 4.7 and Minimax-M2.1, both have significant improvements compared to their predecessors.

However, their financial prospectus doesn't look so promising. While user growth is rapid, R&D and advertising costs are high, resulting in substantial losses, similar to OpenAI.

All these startups working on general purpose frontier models are going to get destroyed by the the established players like Alibaba, Bytedance. LLM is a mature tech now with lots of back and forth iteration happening and lots of distillation training. Distillation ensures that anyone can train with the top model and gain predictive power of that model very easily.

If even Xiaomi can come up with a frontier level model, you know its very easy now. These startups need to find their own niche now and focus becoming the best in specific niche instead of trying to compete for the general purpose models.

 

[tphuang]

https://twitter.com/i/web/status/2003199255692070945

GLM-4.7 came out just now. Pretty decent bump over 4.6. Not like a monumental improvement, but pretty good for 3 months of progress.

One thing on Zhipu. From what I hear, Tsinghua will buy some H200s when it comes available and that will allow Zhipu to train on a larger cluster.

 

[Eventine]

All these startups working on general purpose frontier models are going to get destroyed by the the established players like Alibaba, Bytedance. LLM is a mature tech now with lots of back and forth iteration happening and lots of distillation training. Distillation ensures that anyone can train with the top model and gain predictive power of that model very easily.

If even Xiaomi can come up with a frontier level model, you know its very easy now. These startups need to find their own niche now and focus becoming the best in specific niche instead of trying to compete for the general purpose models.

There is yet to be a company in China that is as dominant at the entire supply chain of AI, as Google has proven to be in the US, so there’s still room for companies to race against the existing players.

Alibaba does not have the best frontier model in China (unlike Google), and neither does Byte Dance. Instead it’s still Deep Seek, Moon Shot, Z.Ai, etc. so I would say the pure AI labs are still holding their own, similar to Open AI and Anthropic in the West. China also needs the service of these labs (& would benefit as such from subsidizing them) because it cannot afford to lose the R&D race to the US by adopting a strictly profits first approach.

But I agree that ultimately there will be consolidation around platform companies with immense capital & fully integrated supply chains. Huawei, Xiaomi, Alibaba, Byte Dance, and Kuaishou are all candidates for winning the eventual competition and I expect at some time window they will fully acquire the aforementioned pure AI labs as their R&D subsidiaries.

 

[henrik]

China Is Winning the AI Race, Says FTSG’s Webb​

 

[bsdnf]

China Is Winning the AI Race, Says FTSG’s Webb​

To me, this remains just another boring American narrative manipulating stock market sentiment.

The AI race is as ambiguous as AGI. Like where is the finish line? If we don't know, then the race may not even exist from the first place

People just want better products.

 

[mossen]

The gap between Closed Source and Open Source is about 7 months.



It was unchanged during the year. Closed source didn't increase their lead but open source didn't close it either. If trends hold, then we will likely see a Gemini 3-type Open Source model during summer next year. At the very least during autumn. Exciting.

 

[bsdnf]

MiniMax M2.1 benchmark