Perplexity vs Anthropic

Two Layers of the Same Stack

The most revealing way to understand Perplexity and Anthropic is through the lens of the agentic economy. Anthropic operates at Layer 1—building the frontier intelligence that powers autonomous agents. Perplexity operates at the discovery layer—the interface between human intent and AI-mediated information retrieval. They are not competitors so much as complementary layers: Perplexity literally uses Claude as one of its backend models, and Perplexity Pro subscribers can select Claude 4.6 as their preferred reasoning engine. This architectural relationship mirrors the broader pattern where application-layer companies depend on foundation model providers, creating both partnership and dependency dynamics throughout the agentic web.

The Search Disruption: Answers vs. Links

Perplexity’s core thesis—that users want answers, not links—has proven correct at scale. With 45 million monthly active users and 170 million global visitors, Perplexity has demonstrated that AI-mediated discovery is not a niche behavior but an emerging default. The introduction of Model Council in February 2026, which lets users compare outputs from multiple LLMs simultaneously, positions Perplexity as a meta-layer above individual model providers. This is strategically significant: rather than being locked to any single model, Perplexity becomes the routing layer that decides which intelligence serves which query. For businesses, this intensifies the challenge of LLM optimization—ensuring your brand is discoverable not just in one model’s training data but across the ensemble of models that Perplexity queries.

The Protocol Play: MCP and Network Effects

Anthropic’s most consequential strategic move may not be Claude itself but the Model Context Protocol. By open-sourcing MCP and then donating it to the Linux Foundation’s Agentic AI Foundation—co-founded with Block, OpenAI, Google, Microsoft, AWS, and Cloudflare—Anthropic transformed a proprietary integration layer into an industry standard. With over 17,000 MCP servers, 97 million monthly SDK downloads, and adoption by every major AI platform including ChatGPT, Gemini, and Copilot, MCP follows Reed’s Law dynamics where value grows exponentially with subgroup formation. This is a classic platform strategy: even if Claude loses model supremacy, Anthropic retains influence through the protocol layer that connects all agents to the world.

Revenue Scale and Business Model Divergence

The revenue gap between Anthropic ($14B run-rate) and Perplexity (~$200M ARR, targeting $656M) reflects their fundamentally different positions in the value chain. Anthropic captures value from every API call across thousands of enterprise customers—the number spending over $100,000 annually grew 7x in the past year. Perplexity’s February 2026 pivot to subscription-first, abandoning AI-integrated advertising, was a bet that user trust matters more than ad revenue for an answer engine. The $42.5 million publisher revenue-sharing pool through the Comet browser represents Perplexity’s attempt to solve the publisher economics problem that its own model creates: if AI delivers answers without clicks, content creators need alternative compensation. This tension is a defining challenge of the agentic web transition.

Agentic Development: Code as the New Frontier

Anthropic’s Claude Code, now responsible for over 4% of GitHub commits and projected to reach 20%+, represents a paradigm where AI agents don’t just assist developers but autonomously write, debug, and ship software. The five-layer agent architecture—MCP for connectivity, Skills for task knowledge, Agent as primary worker, Subagents for parallelism, and Agent Teams for coordination—creates a full stack for autonomous software development. Perplexity’s agentic ambitions are different: its Deep Research feature demonstrates multi-step reasoning that decomposes complex questions, searches multiple sources, and synthesizes results. Both represent different expressions of AI agent capabilities—one pointed at code, the other at knowledge.

The IPO Race and Market Structure

Anthropic’s reported plans for an October 2026 IPO at its $380 billion valuation would make it one of the largest tech IPOs in history, reflecting market confidence that foundation model providers capture the lion’s share of value in the AI economy. Perplexity’s $21 billion valuation, while substantial, illustrates the market’s view that application-layer companies—however innovative—operate with thinner margins and greater substitution risk. The question both companies face is sustainability: Anthropic must maintain model leadership against fierce competition from OpenAI and Google DeepMind, while Perplexity must prove that an answer engine can build durable moats when the underlying models are commoditizing.