Investment Memo: Perplexity
Prepared for: Investment Committee Stage framing: Hypothetical Series A evaluation Author's note: Perplexity is well past Series A in reality; this memo evaluates the thesis as if we were seeing it at that stage.
---
Thesis
Search is the single largest and most defensible profit pool in software — Google generates ~$175B+ annually from it — and it has not faced a credible architectural challenge in twenty years. LLMs create the first genuine discontinuity: users increasingly want answers, not ten blue links. Perplexity's bet is that a native answer engine, unburdened by an ads business it must protect, can capture the highest-intent slice of search behavior (research, purchasing, professional queries) before Google fully cannibalizes itself.
The fund-returner case is not "beats Google at general search." It's narrower and more plausible: Perplexity becomes the default research tool for knowledge workers — the Bloomberg-terminal-shaped wedge into a market so large that even 2–3% of query share, monetized via subscriptions and eventually transactional/ad revenue, supports a $50B+ outcome. In consumer internet, being the verb for a new behavior ("just Perplexity it") is the whole game.
Product & Wedge
The core product is conversational search with inline citations: query → synthesized answer → sources → follow-ups. The citation layer is the wedge — it solves the trust problem that makes raw ChatGPT unreliable for factual work, and it makes the product feel like research infrastructure rather than a chatbot toy.
The strategy is deliberately model-agnostic: Perplexity orchestrates frontier models (OpenAI, Anthropic, its own fine-tuned Llama-based Sonar models) on top of its own retrieval/ranking stack. The proprietary asset is the search index, ranking, and orchestration layer — not the LLM. That's the right call: don't compete on frontier models you can't fund.
Expansion vectors: Pro subscriptions ($20/mo), enterprise seats, an API, a browser (Comet), shopping/agentic features, and early ad experiments. The consumer product is the top of funnel; enterprise and API are where margin lives.
Market & Competition
TAM is effectively the global search + knowledge-work market — hundreds of billions. But the competitive field is the most brutal in tech:
- Google: 90%+ search share, distribution via Chrome/Android/default deals, and has shipped AI Overviews and Gemini directly into the results page. It will bundle "good enough" AI answers for free at planetary scale.
- OpenAI / ChatGPT: The real existential threat. ChatGPT has search with citations, ~10x+ Perplexity's user base, and the strongest consumer AI brand. Perplexity's core use case is a feature inside ChatGPT.
- Microsoft Bing/Copilot: Distribution through Windows and Office.
- Anthropic, Meta AI, Grok: All converging on search-augmented answers.
Perplexity's differentiation — focus, speed, citation quality, no legacy ad conflict — is real but thin. This is a knife fight against the four best-capitalized companies on earth.
Traction & Business Signal (public information only)
- Usage: Company has publicly cited hundreds of millions of queries per month (reported ~780M monthly queries in 2025) and tens of millions of MAU. Exact DAU/retention: unknown.
- Revenue: Reports place ARR in the ~$100M–150M+ range (2025), primarily subscriptions. Gross margin after inference and licensing costs: unknown, likely thin.
- Fundraising: Serial raises at escalating valuations — reportedly ~$9B (late 2024) to $18–20B (2025), backed by NEA, IVP, Nvidia, Jeff Bezos, SoftBank. Strong investor signal; also signals ferocious burn.
- Retention/cohorts, CAC, enterprise seat counts: unknown — and these are the numbers that matter.
Risks (the three that kill the deal)
1. Feature, not company. If ChatGPT's search reaches parity — and it is close — Perplexity's reason to exist for the median user evaporates. Perplexity has no distribution moat: no OS, no browser install base (Comet is nascent), no default-search deals it can afford. Google pays Apple ~$20B/year for the Safari default; Perplexity cannot play that game. Consumer AI habits are consolidating fast around one or two apps, and Perplexity may not be one of them.
2. Structurally bad unit economics. Every query costs real inference money; Google's costs per query are fractions of a cent. Monetizing via $20/mo subs means the heaviest users are the least profitable. The eventual ad model puts Perplexity in direct margin competition with Google's perfected machine — from a standing start.
3. Legal and supply-chain exposure. Perplexity has been sued or sent cease-and-desists by News Corp/Dow Jones, and accused of scraping violations by Forbes and Wired. Its product is a synthesis layer over other people's content and other people's models. Adverse copyright rulings, or content/model providers cutting access or raising prices, attack the core input costs and legality of the product simultaneously.
Verdict: Invest (at Series A terms, with conviction sized to risk)
At a hypothetical Series A — sub-$150M valuation, pre-consolidation of the market — the asymmetry justifies the bet: elite product velocity, a founder (Aravind Srinivas) with genuine research pedigree and distribution instincts, attacking the largest prize in software. The failure modes are real but the power-law payoff is exactly what a fund exists to underwrite. (At the current reported ~$20B valuation, we would pass — the risk/reward has inverted.)
The one question to answer first: What is 6-month retention among weekly-active free users, and does it hold when ChatGPT search is offered side-by-side? If users stay because of Perplexity specifically — not because they haven't tried the alternative — the wedge is real. If not, we're funding a beautifully executed feature.