- The CEO of Perplexity AI warns that the data center-centric AI model could be disrupted if intelligence is packaged and run directly on personal devices.
- According to him, when AI can process and reason directly on the chips of phones, tablets, or laptops, the need to build massive data centers will drop sharply.
- This poses a vital question for the industry: Is it still logical to spend from $500 billion to several trillion dollars building global data center infrastructure?
- Local AI has the advantage of learning from each user’s repetitive behavior, gradually automating tasks and becoming “personal intelligence” owned by the user.
- Apple is considered to have a significant advantage thanks to M1 chips and energy-efficient chip lines suitable for on-device AI.
- Qualcomm and device manufacturers like Apple, Samsung, Lenovo, and HP could also benefit from distributing AI-integrated devices.
- However, technical barriers persist as no AI model is yet small or stable enough to run fully and efficiently on local chips.
- Early deployment is predicted to start on MacBooks or iPads before becoming common on smartphones.
- On-device AI also opens great potential for robotics and labor automation, raising concerns about human replacement.
- In the context of the US economy’s increasing dependence on AI, the collapse of the centralized data center model could become a “single point of failure” for the entire economy.
Conclusion: The CEO of Perplexity AI warns that the data center-centric AI model could be disrupted if intelligence is packaged and run directly on personal devices. He raises a $10 trillion question regarding current infrastructure investment strategies. If local AI becomes feasible, Apple, Qualcomm, and device makers could benefit greatly, while the centralized data center model faces risks. This is not just a tech story, but a risk of an economic bubble and a reshaping of the entire AI value chain.
