
HONG KONG – Shares of Chinese AI model maker Zhipu surged after JPMorgan Chase & Co. raised the stock’s price target and picked it as a winner against close rival MiniMax.
Zhipu, which trades as Knowledge Atlas Technology, jumped as much as 48 per cent on June 15, one of its biggest surges since a successful initial public offering in Hong Kong to kick off the year.
Zhipu was the first of China’s “AI tigers” – start-ups building large language models to rival OpenAI and Anthropic – to go public. MiniMax’s market debut came a day after Zhipu in January.
JPMorgan analysts kept their overweight rating on Zhipu and raised the target to HK$1,400 from HK$950, touting the firm’s model visibility and what the bank sees as pricing power in a hotly contested arena. They cut their MiniMax rating to neutral. Its shares fell as much as 4.8 per cent.
After trading toe-to-toe for weeks, the two companies have diverged sharply since mid-March, driven by Zhipu’s ability to raise prices following the rollout of its GLM-5 series models. MiniMax has not shown the same capacity in coding and language models, but has instead positioned itself as an all-rounder with video generation and consumer apps under its belt.
The reassessment comes as US peer Anthropic has been ordered by the US government to curtail access to its latest AI models, Mythos and Fable, after warnings about the capabilities of those tools. As concern about maintaining control over increasingly strategic new technology rises in the US, Chinese contenders are standing out for their embrace of the widest possible market.
“Zhipu is being rewarded because investors see a very clear contrast in AI strategy,” said Charu Chanana, chief investment strategist at Saxo Markets. “At a time when Anthropic has been forced to restrict access to its most advanced models for foreign users, Zhipu is moving in the opposite direction by opening its latest model more broadly. That sends a powerful message: Chinese AI firms may try to win through scale, distribution and ecosystem adoption, not just model performance.” BLOOMBERG



