1. The release of Kimi K3 illustrates that the leading position of cutting-edge large-scale models in China can change rapidly in a single iteration, making it difficult for any single vendor to maintain an absolute lead in the long term. Investors should shift their valuation logic for model companies from betting on a "sole long-term winner" to assessing their ability to remain at the forefront for multiple generations. Due to increased competition, JPMorgan Chase lowered its 2030 forward P/E ratio valuation for Zhipu and MiniMax from 30x to 20x.
2. However, K3 only shortened the lead window of GLM-5.2; it did not remove Zhipu from the top tier. In the coming months, Zhipu will also launch GLM-5.3 and a flagship model with over 2 trillion parameters. As long as it remains at the forefront, given the current limited computing power supply and the early stage of AI commercialization in China, Zhipu is still expected to maintain ARR (Annual Recurring Revenue) growth. Therefore, JPMorgan Chase's revenue forecast adjustment is limited.
3. Currently, the combined ARR of major independent large-scale model manufacturers in China is approximately US$2.1 billion, far lower than Anthropic's approximately US$69 billion, indicating that commercialization is still in a very early stage. Zhipu's stock price has corrected by over 50% after the K3 release, and the market has fully priced in the valuation downgrade, but may be overly pessimistic about its commercial prospects.
4. JPMorgan maintains its "Overweight" rating on Zhipu, lowering its target price from HK$2400 to HK$1600; MiniMax maintains its "Neutral" rating due to insufficient evidence of returning to the forefront, lowering its target price from HK$240 to HK$160.
