Yesterday, three major US AI giants called for a slowdown in development. JPMorgan Chase data shows that long-term funds increased their holdings in Google and Meta that day. The investment bank stated that if the market begins to believe that training cutting-edge AI models needs to slow down, funds may shift from "AI shovels" (referring to companies focused on AI development) to large cloud platforms, cybersecurity, inference, and low-volatility tech stocks:
1. Cybersecurity software such as CrowdStrike, Zscaler, and Palo Alto Networks are seen as clear beneficiaries; companies focused on inference applications, such as Apple and HP, may also benefit relatively.
2. Semiconductor equipment and training network equipment that rely on long-term AI training capacity expansion may face pressure, such as Lam Research, KLA, and Ciena.
3. However, the report emphasizes that this is more of a market scenario reassessment; the selling is mainly for reducing positions rather than panic. Furthermore, AI security, model alignment, and third-party evaluation may themselves require more computing power; therefore, "slowing down training = decreased computing power demand" is not a definitive conclusion. (The above views are from a JPMorgan Chase report dated September 14th.)
