Perplexity AI

Perplexity AI

Perplexity AI was founded in 2022 by Aravind Srinivas, Denis Yarats, Johnny Ho, and Andy Konwinski — a team whose prior tenures at Google AI, OpenAI, and Databricks gave the company immediate technical credibility. Srinivas, who interned at OpenAI and later worked at Google Brain, serves as CEO and has become the company's defining public voice. The wealth origin is venture-backed; the firm is not a family office but operates as a venture-capital-funded startup with significant capital deployment across model training, product engineering, and enterprise sales. The company's core product is an AI-native search engine that synthesizes answers from multiple web sources with inline citations — a direct bypass of the traditional ten-blue-links model. Perplexity monetizes through a $20/month "Perplexity Pro" subscription and an enterprise API that allows developers to embed its search capabilities. Confirmed investors include Institutional Venture Partners, New Enterprise Associates, and Jeff Bezos's personal investment vehicle. The firm also launched an internal advertising integration in late 2024, testing sponsored follow-up questions — a move that drew scrutiny from publishers including Condé Nast, which sent a cease-and-desist letter in October 2024 over alleged content scraping. Geographic focus is global, with material user bases in the US, India, and Europe. Headcount is not publicly fixed but was reported at roughly 55 employees in early 2024 and has since grown. The sole known office is San Francisco. Srinivas has publicly discussed raising a new funding round at an $18 billion valuation as of early 2025 (per Financial Times, March 2025). In December 2024, the company announced a $500 million funding round led by IVP, which brought total fundraising to over $900 million. The company does not operate a separate philanthropic arm but has made its Pro tier freely available to students and researchers through select academic partnerships. Perplexity's structural differentiator is its distribution model: it bypasses the traditional search-index-and-rank architecture entirely by combining retrieval-augmented generation with a large language model from the start. This makes it less a Google competitor and more a vertically integrated answer provider — a posture that has drawn the attention of every major search incumbent and has made copyright risk, not compute cost, its binding constraint.