Aravind Srinivas Predicts Shift in AI Industry, Perplexity CEO Says Multi-Billion Dollar Data Centres To Become Less Relevant As On-Device AI Takes Over

San Francisco, January 5: Aravind Srinivas, the CEO of AI search startup Perplexity, has stated that the current industry obsession with building multi-billion dollar data centres may soon become a secondary concern. Speaking on the future of artificial intelligence, Srinivas suggested that while massive computing power is currently a priority, the long-term relevance of these physical hubs will diminish as AI technology matures and becomes more efficient.

His remarks come at a time when tech giants such as Microsoft, Google, and Meta are committing hundreds of billions of USD towards infrastructure to support Large Language Models (LLMs). Srinivas argues that the focus will eventually shift from sheer hardware scale to the intelligence and efficiency of AI software itself. Layoffs 2026: Goldman Sachs Predicts Fresh Wave of AI-Driven Job Cuts Amid Global Tech Adoption; Firms To Reduce Headcounts.

AI Model Efficiency: Reducing Reliance on Massive Data Centres

Aravind Srinivas highlighted that the AI industry is currently in a phase where performance is often tied directly to the amount of hardware used. However, he believes this trend is unsustainable. As AI models become more sophisticated, they will require less energy and fewer chips to perform complex tasks.

This transition mirrors previous cycles in computing, where initial massive mainframes eventually gave way to more efficient, distributed systems. For Perplexity, which positions itself as an "answer engine," the priority remains accuracy and speed of information retrieval rather than owning large-scale physical infrastructure.

Data Access Challenges: The Biggest Threat to AI Growth

While hardware costs remain a significant hurdle, Aravind Srinivas identified a different major threat to the AI industry: access to high-quality data. As publishers and platforms increasingly block AI crawlers and demand licensing fees, the pool of free public data used to train AI models is shrinking.

Perplexity has faced similar challenges, navigating copyright concerns and implementing "Publishers' Programmes" to share revenue with content creators. Srinivas noted that the ability to access real-time, verified information is more critical to a company’s survival than the number of GPUs it owns.

AI Capital Expenditure Trends and Industry Outlook

The CEO’s comments contrast with current market realities. Reports indicate that global capital expenditure for AI data centres could exceed 1 trillion USD by 2027. Companies are competing to secure Nvidia’s latest chips and access to power grids to meet demand.

Perplexity CEO Aravind Srinivas said that if device‑level AI can run locally on personal chips, users will “own” their intelligence, reducing the need to repeat tasks and potentially upending the data centre model. He argued it no longer makes sense to spend hundreds of billions or even USD 500 billion to USD 5 trillion on centralised data centres, calling it a “USD 10 trillion question” for the industry. 'More To Come Soon': Perplexity CEO Aravind Srinivas Says Company Plans To Expand in India Beyond Launching Local Fund.

Future of AI Search: Perplexity’s Search-First Approach

As the AI landscape evolves into 2026, debate continues over whether "scaling laws"—the idea that more data and more compute always produce better AI—will remain valid. Srinivas concluded that the ultimate winner in the AI race will be the company providing the most utility to users with minimal friction. Perplexity continues to focus on its "search-first" strategy, prioritising direct answers over traditional links.

(The above story first appeared on LatestLY on Jan 05, 2026 09:08 AM IST. For more news and updates on politics, world, sports, entertainment and lifestyle, log on to our website latestly.com).