Aravind Srinivas is going straight for Google (GOOGL). The 31-year-old founder and CEO of Perplexity AI, a rapidly rising startup valued at $9 billion, is building what’s dubbed an “answer engine” to compete with the search giant. Founded in 2022, Perplexity’s main product is a search engine that uses A.I. to generate nuanced answers—instead of links—in response to user queries.
In December 2024, Perplexity closed a $500 million funding round led by Institutional Venture Partners. Even by the lofty standards of other A.I. startups, Perplexity’s growth last year was one for the books: The company was valued at $1 billion in April 2024. Just two months later, that figure rose to $3 billion after an investment from Softbank’s Vision Fund 2. And by the end of the year, its valuation had tripled to $9 billion.
Srinivas was born and raised in Chennai, India—the same town that raised his role model turned rival, Google CEO Sundar Pichai, albeit 22 years apart. Srinivas obtained his Bachelor’s and Master’s in electrical engineering from the esteemed Indian Institute of Technology, Madras. He moved to the U.S. in 2017 to pursue his Ph.D. in computer science at University of California, Berkeley.
In 2019, Srinivas interned as a researcher at the Google-owned DeepMind in London. During this internship, the book In The Plex by Steven Levy, a chronicle of the development and evolution of Google, became his Bible, he told Wired in an interview where he offered to recite passages from memory. He then worked at OpenAI as a research scientist. Shortly thereafter, Srinivas landed a job at Google’s headquarters in Mountain View, Ca., where he focused on developing machine learning programs for computer vision.
In 2022, Srinivas and three co-founders—Jonny Ho, Andy Konwinski, and Denis Yarats—joined forces to develop a new approach to online search. They landed on a product that combined a traditional search index with large language models and called it Perplexity. In A.I., “perplexity” is a metric used to evaluate how well a language model predicts text. It measures the model’s “surprise” when encountering new data—thus, the less perplexed it is, the more accurate the model.
Perplexity’s investors including Nvidia, Amazon founder Jeff Bezos, former YouTube CEO Susan Wojcicki, former GitHub CEO Nat Friedman. It has even won investments from individuals working for competitors, such as Google chief scientist Jeff Dean and Meta chief A.I. scientist Yann LeCunn. Nvidia CEO Jensen Huang, when asked about his use of A.I. chatbots, enthusiastically responded that he turns to Perplexity.
Much like the near-incestuous cross-pollination of A.I. investments, Perplexity’s rivalry with Google is not so cut-and-dry. Google created the critical transformer model architecture, otherwise known as the T in ChatGPT, that has lead to the modern neural network behind all A.I.-driven search tools on the market today. That means, like many A.I. companies, Perplexity is competing with Google using the technology that Google invented.
Perplexity uses a combination of OpenAI’s GPT 3.5-turbo and GPT 4-turbo language models. It also uses customized versions of Meta’s open-source model Llama, one from the French startup Mistral AI, and Anthropic’s Claude. Thus, the company’s big bet is on its unique interface, which gives users the experience of a search rather than an A.I. chat model.
In a post on X last month, Srinivas said the company has trained some models that surpass competitors for answer quality “despite being way cheaper and blazing fast!”
Perplexity has been criticized—and in some cases, sued—by news organizations alleging the companies plagiarized published content without attribution. In response, the company launched its Publishers Program, which promises to share revenue from copyrighted content with publishers.
The company is also expanding beyond A.I. applications to hardware. Last week, at the Mobile World Congress in Spain, Srinivas announced that Perplexity was partnering with T-Mobile parent company Deutsche Telekom to build a new “AI Phone.” The new device, to be revealed later this year, will run “Magneta AI,” which gives users access to Perplexity Assistant as well as Google Cloud AI, Picsasrt and a suite of AI tools. It was described as an “app-less” device controlled primarily by voice, which can do tasks like booking flights and making dinner reservations.
“Perplexity is transitioning from just being an answer machine to an action machine,” Srinivas said onstage at the event. “It is going to start doing things for you, not just answering questions.”
