In an interview with CNBC, Aravind Srinivas, chief executive officer of Perplexity, said that companies that can extract the most economic value from the energy consumed by artificial intelligence will be the long-term winners in the market. He argued that success will depend on maximizing value per watt and per user while balancing accuracy, latency, cost, privacy and the level of intelligence.
Perplexity's focus on agent-based AI
Perplexity is increasingly concentrating on agent-based AI — systems capable of carrying out more complex, longer-running workflows beyond simple question–answer interactions. The company introduced its Perplexity Computer agent in February, designed to handle such extended tasks autonomously.
On Wednesday, Perplexity announced that its Personal Computer product is now available on the Microsoft Windows operating system. The Windows release allows the AI agent to connect directly to applications such as Word and Outlook and to access files stored on the user’s device.
Intense competition and valuations
Perplexity faces stiff competition from major players including OpenAI, Anthropic and Google, all of which are investing heavily in their own AI agents. The most recently reported market valuation for Perplexity was $20 billion, considerably below Anthropic’s roughly $1,000 billion and OpenAI’s about $850 billion.
Anthropic filed a confidential IPO application in the United States this week.
Why this matters
Srinivas’s point is that market leadership may hinge not only on model accuracy or feature set but on how effectively companies can optimize the economic efficiency of AI energy use. That factor becomes particularly relevant for agent-based applications, which may run longer and demand more resources. In practice, better energy efficiency can translate to lower costs, reduced latency or stronger privacy guarantees — all elements that will shape competition and long-term valuations in the AI sector.
This article is not investment advice or a recommendation.
