Perplexity CEO Aravind Srinivas has suggested that the future of artificial intelligence could significantly disrupt the data centre industry as AI capabilities move from centralised servers to local devices. Speaking on a podcast with YouTuber Prakhar Gupta, Srinivas said that powerful on-device AI could reduce the need for large-scale data centres that currently handle most AI inference workloads.
On-Device AI vs Data Centres
According to Srinivas, the “biggest threat to a data centre” would be a scenario where AI intelligence can be efficiently packed onto a chip running directly on consumer devices. In such a model, AI inference would no longer depend on centralised cloud infrastructure, potentially reshaping the economics of the global data centre ecosystem.
He explained that if AI models become capable of running locally on devices such as smartphones, laptops, or wearables, it could reduce the massive investments currently being made in data centre buildouts worldwide. This shift could lead to a more decentralised AI ecosystem, with computation happening closer to the user.
Shift From Chatbots to Agentic AI
Srinivas also highlighted a broader transformation within the AI industry, noting that artificial intelligence is moving beyond traditional chatbots toward agentic operations. These agent-based systems are designed to perform tasks autonomously, making AI more proactive and context-aware rather than purely reactive.
He added that on-device AI has the potential to be deeply personalised, adapting to individual user behaviour and preferences without relying heavily on cloud-based processing.
Human Intelligence vs Artificial Intelligence
Beyond infrastructure and technology, Srinivas discussed the fundamental differences between human and artificial intelligence. He pointed out that the human brain is significantly more energy-efficient than modern data centres when measured per watt. More importantly, he said AI systems lack the biological curiosity that drives humans to ask original questions, challenge assumptions, and explore ideas independently.
Looking ahead, Srinivas suggested that widespread access to personalised AI could democratise capabilities across society, much like smartphones have done over the past decade. He believes such tools could help individuals compete with larger institutions by placing advanced intelligence directly into their hands. He also noted that age is not a barrier to adopting AI – curiosity and openness to learning are far more important.
Overall, Srinivas’ remarks point toward a future where on-device AI, agentic systems, and decentralised intelligence could redefine how artificial intelligence is built, deployed, and experienced, while posing a long-term challenge to the dominance of centralised data centres.
