MiniMax once led Zhipu in Hong Kong’s AI stock race. How the tables have turned

When Chinese artificial intelligence developers Zhipu AI and MiniMax made their debuts on the Hong Kong stock exchange in January, investors initially saw one as a better bet than the other. Beijing-based Zhipu closed its first trading day on January 8 with a market capitalisation of HK$57.9 billion (US$7.4 billion). A day later, Shanghai-based MiniMax listed at HK$106.7 billion, almost twice as large. Five months on, however, Zhipu – traded as Knowledge Atlas Technology – has surged ahead. At the market close on Monday, its market cap stood at HK$585.8 billion, nearly 2.7 times higher than MiniMax’s HK$159.3 billion. The reversal, analysts said, reflected looming share unlocks, stronger AI model capabilities and investor preference for enterprise-focused businesses. Analysts also warned the gap could widen further in the coming months, with MiniMax facing a major round of share lock-up expiries and Zhipu benefiting from advanced model capabilities and fresh liquidity from the Stock Connect programme. Both firms, however, are expected to encounter bumpy roads ahead amid intensifying domestic and global competition. In early July, Zhipu will release about 6 per cent of its shares to the market, while MiniMax will see 46 per cent unlocked, according to Ritchie Sun, internet research analyst at HSBC. Lock-ups are designed to stabilise newly listed stocks by preventing major shareholders from selling immediately after an initial public offering. “In our view, this means MiniMax is likely to see more near-term share price volatility [than] Zhipu,” Sun said. Zhipu could also benefit from a “liquidity buffer” after being added to the Stock Connect programme on Monday, opening its shares to mainland investors. MiniMax, because of its weighted voting rights structure, would not qualify until August, according to Sun. Beyond liquidity, investors gravitated towards Zhipu’s technology and a more enterprise-focused business model, analysts said. Earlier this year, the company raised its application programming interface (API) prices by 83 per cent, yet requests jumped 400 per cent in the first quarter, CEO Zhang Peng said in an April earnings call. “Due to stronger tech accumulation, Zhipu now has more pricing power and confidence in pushing model monetisation forward,” said Tilly Zhang, China technology and industrial policy analyst at research firm Gavekal. Zhipu’s latest flagship model, GLM‑5.1, is priced at US$0.90 per one million tokens, blending cache hit, input and output costs. MiniMax’s newest M3 model, by contrast, came in at US$0.22 per one million tokens, according to Artificial Intelligence. MiniMax’s cost-effective approach may not offer a “sufficiently differentiated” competitive advantage relative to other firms, according to Gavekal’s Zhang. This approach left it more hesitant to raise prices and slower to monetise compared with enterprise-focused rivals, she said. Zhipu coding capabilities have also been a draw. Sun said the company’s model capabilities – particularly its coding performance – were “widely acknowledged by investors as a stand-out strength”. But Jefferies analysts said in a note published on Monday that “Zhipu’s edge looks overstated”, with industry contacts “universally” citing Claude as the best global coding model, while Alibaba Group Holding’s Qwen and DeepSeek were the most mentioned Chinese alternatives. Alibaba owns the South China Morning Post. Despite the lower pricing, the research firm judged MiniMax’s M3 more cost‑effective than GLM‑5.1. On the DeepSWE coding benchmark, the MiniMax flagship ranked 10th, two spots ahead of Zhipu’s GLM‑5.1. The two firms’ revenue streams also differ. Zhipu’s business revolves around its model-as-a-service platform, serving institutional clients via API or local deployment. MiniMax, while offering enterprise services, leans heavily on consumer-facing apps, including video generation platform Hailuo AI and AI companion app Talkie, which analysts said carried less bargaining power. Zhipu’s potential was “easier to understand”, said Hong Hao, managing partner and chief investment officer at Lotus Asset Management. Both may face headwinds. Coming listings of leading US AI leaders – including SpaceX, Anthropic and OpenAI – could pressure valuations, with both MiniMax and Zhipu “trading at a premium to US peers”, Sun said. Domestic rivals such as DeepSeek also posed pricing challenges, he added. Zhipu and MiniMax did not immediately respond to requests for comment. Both stocks were added to the Hang Seng Tech Index on Monday, the first pure-play AI firms to enter the city’s benchmark technology gauge. Although their initial weightings were small – 0.53 per cent for Zhipu and 0.36 per cent for MiniMax – Morgan Stanley estimated in April that the pair could eventually make up 5 to 7 per cent of the benchmark combined. The bank also calculated that had the two firms been included at the time of their initial public offerings, the index’s year-to-date return would have been 5 percentage points higher.

MiniMax once led Zhipu in Hong Kong’s AI stock race. How the tables have turned

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Zhipu’s valuation now eclipses MiniMax, but looming lock-up tests may unleash volatility in Hong Kong’s AI stocks, analysts say

Zhipu’s valuation now eclipses MiniMax, but looming lock-up tests may unleash volatility in Hong Kong’s AI stocks, analysts say