BlockBeats reported on July 21 that JPMorgan, in its latest report on China’s artificial intelligence sector, said the release of Kimi K3 has changed the valuation framework for China’s frontier model companies, while arguing that the selloff in Zhipu has already gone too far.
The call came as U.S.-listed AI names have seen sharp swings. In that backdrop, the market at one point treated Kimi K3 as a new “DeepSeek moment”: a Chinese model built around low cost, strong performance, and open weights that revived investor concerns over returns on U.S. AI capital spending, model pricing power, and demand for computing capacity.
The Wall Street Journal said Moonshot AI’s new model added to anxiety among chip investors. JPMorgan said K3’s launch materially changed how the market values leading Chinese model developers.
Valuation multiples cut after K3 launch
According to the report, the market began to question the long-term leadership of companies including Zhipu and MiniMax after K3 was released. JPMorgan said that led investors to cut valuation multiples to 20x 2030 estimated P/ARR from around 30x previously.
Still, the bank said Zhipu’s share-price correction of more than 50% has already more than priced in that pressure. JPMorgan maintained an Overweight rating on Zhipu, while lowering its target price to HK$1,600 from HK$2,400. For MiniMax, it kept a Neutral rating and cut the target price to HK$160 from HK$240.
JPMorgan says Zhipu remains in China’s frontier-model group
The core view in the report is that Kimi K3 shortened the lead window of GLM-5.2, but did not remove Zhipu from the ranks of China’s frontier model players. JPMorgan said GLM-5.2 is still among the country’s top production-grade models.
The bank also said Zhipu is expected to test its competitiveness again over the coming months through GLM-5.3 and a 2T+ flagship model. If those new releases return to the frontier in coding, reasoning, and agent tasks, Zhipu would still be able to sustain commercial growth.
Commercialization remains early in China
JPMorgan said the bigger point is that model commercialization in China is still in an early phase. The bank estimated Zhipu’s latest indicative ARR at about $1 billion, DeepSeek at about $500 million, and MiniMax and Kimi at about $300 million each. That puts the combined ARR of China’s leading independent model suppliers at about $2.1 billion.
By comparison, Anthropic’s ARR has reached about $69 billion, according to the report. On JPMorgan’s reading, that means Chinese model companies are still operating from a much smaller revenue base, and their growth runway has not been exhausted by the arrival of Kimi K3.
K3 also signals a pricing shift
The report said Kimi K3 sent another message to the market: stronger domestic models are trying to prove their value at higher prices. JPMorgan noted that K3 API pricing is materially above the previous K2.7 Code model, reflecting stronger capabilities in higher-value tasks such as coding.
If customers are willing to pay for stronger reasoning, longer context windows, and programming capability, the business model for Chinese model companies could move away from low-price competition and toward capability-based pricing.
JPMorgan’s overall tone was not bearish. In its view, Kimi K3 has triggered a valuation reset and amplified worries in the U.S. AI trade around “cheap Chinese models,” but it has also shown that China’s model layer is moving closer to the global frontier. For Zhipu, the next model release cycle will be the key window for repricing.