Perplexity AI co-founder and CEO Aravind Srinivas says the next major shift in artificial intelligence may not come from data centers, but from users’ personal devices.
In an interview on PGX Ideas, Srinivas describes a future where AI adapts to users locally, watching how tasks are performed and gradually automating them without sending data back to centralized servers.
He says the key breakthrough would be persistent, personalized learning that stays on the device.
“Imagine we crack something like test-time training, where there are some tasks that you repeatedly do on your local system and the AI watches it, and you’re fine with the AI watching it because the AI is living on your computer. It’s not going to the servers, and it adapts to you. And over time, it starts automating a lot of the things you do. That way, you don’t have to repeat it. That’s your intelligence. You own it. It’s your brain.
Then that really disrupts the whole data center industry.”
Srinivas is careful to note that this vision has not yet been realized, emphasizing that today’s models still rely heavily on centralized compute.
“I’m talking about a really good end-to-end agent that just looks at whatever you’re looking at on the computer and actually helps you get things done, but lives locally. That hasn’t yet happened. No one’s actually shipped a model that can be packaged on your computer, a very efficient chip locally, and then it’s very intelligent enough that it can complete tasks reliably. That’s not yet happened.”
When that threshold is crossed, Srinivas says the impact on data centers remains an open question.
“When that happens, I think it’ll be very interesting whether it’ll completely disrupt the data center business or whether it’ll pour some water on it, but not entirely disrupt it.”
He believes the biggest beneficiaries of AI agents living on local devices would be companies with strong chip capabilities and control over hardware distribution.
“Who makes the money? Apple has a massive advantage here because they have a great chip business. M1 chips are great. Their devices are very power-efficient. Qualcomm obviously is going to be relevant. So the chip companies have a lot to gain here.”
He adds that original equipment manufacturers (OEMs) would play a central role as the distribution layer for local AI systems.
“And then the OEMs have a lot to gain here because in the end, they’re the distributors for the actual devices. So Samsung, you know, Apple, of course, Lenovo, HP, they’re all going to be relevant in that business.”
Srinivas is not alone in the view that local AI agents could disrupt the data center story. Last month, Atreides Management chief investment officer Gavin Baker said edge AI is the most plausible and scariest bear case for the AI boom. He also predicts that Apple will win in an environment where AI models are run locally due to its privacy policies.
Disclaimer: Opinions expressed at CapitalAI Daily are not investment advice. Investors should do their own due diligence before making any decisions involving securities, cryptocurrencies, or digital assets. Your transfers and trades are at your own risk, and any losses you may incur are your responsibility. CapitalAI Daily does not recommend the buying or selling of any assets, nor is CapitalAI Daily an investment advisor. See our Editorial Standards and Terms of Use.