Perplexity

A mere four-year-old AI company is currently aiming for a valuation of 30 billion US dollars (approx. 2016 billion yuan). On August 23, it was announced that Nvidia is in talks to participate in a new funding round for the AI search company Perplexity. This round could reach several billion US dollars, which would push the company's valuation above 30 billion US dollars. If the transaction is completed, Perplexity's valuation will increase by at least another 50% within approximately one year. At the beginning of 2024, the company's valuation was still only around 500 million US dollars.

By the end of 2024, it had already reached 9 billion US dollars. In June 2025, it further rose to 14 billion US dollars; in the following months, it successively surpassed the 18 and 20 billion US dollar marks. The latest funding round has now raised the target to over 30 billion US dollars. In less than three years, the valuation has increased from 500 million to 30 billion US dollars – that's almost 60 times. Early investors can even expect an excess return of 240 times. What tempting business opportunities lie behind the highly sought-after Perplexity?

Founded in 2022, Perplexity is the AI version of Google. When you enter a question into the search box, you don't get a list of links to click, but rather a directly compiled answer with source citations for each sentence – it reads dozens of websites for you and extracts the result. Many professionals – journalists, researchers, investors – use it for information gathering and initial research. Perplexity AI Search Interface Source: Perplexity Official Website However, Perplexity's real evolution took place this year. It is no longer just a 'search box' but is developing into an AI agent that can directly help you with your work.

The most surprising thing about Perplexity is not the rapid increase in its valuation, but the fact that it actually generates profits. Perplexity's annualized revenue (ARR) has already exceeded 750 million US dollars. Where does the money come from? There are three main revenue streams: The first and most fundamental is consumer subscription. Perplexity Pro costs 20 US dollars per month, the premium version is more expensive. Like Netflix or Spotify, it is billed monthly. The company has entered into pre-installation agreements with major smartphone manufacturers and mobile network operators worldwide – it is pre-installed on Motorola devices, Samsung Galaxy users receive 12 months free, and 360 million users of Indian telecommunications giant Airtel receive the free Pro version.

This model of 'operator subsidy + free trial → conversion' has helped to quickly build a large user base. The second revenue stream is enterprise services, the fastest-growing area this year. Perplexity has launched Computer for Enterprise, essentially an AI work assistant for businesses, costing 200 US dollars per seat per month. It can be connected to internal company systems and helps employees automatically complete complex tasks – for example, for the query 'Summarize all data changes this week and highlight discrepancies,' it directly provides results with source citations from real-time data.

According to media reports, over 100 enterprise customers contacted them to request a trial period within one weekend after the product launch. The third revenue stream is the API business. Perplexity's Sonar API allows developers to integrate Perplexity's 'search function with source citations' into their own products. Customers in finance, healthcare, and enterprise SaaS use this – the search function is provided like a utility service, where usage is billed based on consumption. 9 million US dollars – an efficiency that reaches peak values across the entire enterprise software industry.

Deng Mingsheng, founding partner of Yushi Capital, told Pencil News: Perplexity is growing fast not only because AI search is trending, but primarily because it is gradually evolving from a search engine into an agent that can directly perform tasks. 'A search engine is initially just a tool. Users with a question or need search for information, then judge for themselves whether this information is reliable and comprehensive, and finally arrive at an answer. In this process, the user makes the actual decision themselves,' said Deng Mingsheng. 'An agent follows a different logic: it takes over part of the user's thought process, acts like an expert, performs analyses, evaluations, and even decisions, and directly delivers the result to the user.'

In his view, today's AI search is only an intermediate state – because user trust in large models and agents is not yet fully developed. However, once this trust emerges, it can grow faster than many expect. 'People fundamentally want to be 'lazy'. If a model perfectly fulfills my need three times in a row, why should I then search dozens of websites myself and spend time on my own evaluations?' At the 30 billion US dollar valuation celebration, the winning team is undoubtedly the founding team. CEO Aravind Srinivas was born in 1994 into a simple family in Chennai, India.

His father worked in the financial sector; the family had no technical background. He studied electrical engineering at IIT Madras – as he was not admitted to the computer science faculty at the time and a change of major was unsuccessful, he taught himself programming and machine learning. After completing his bachelor's degree in 2017, he went to UC Berkeley to pursue a doctorate; his supervisor was reinforcement learning expert Pieter Abbeel.