$EXA - Exa is building a search engine optimized for AI agents that is already better than Google in many ways, serving a massive TAM as agentic search will exceed Google's business by the 2030s. The company has strong product-market fit with 5,000+ business customers and is being recommended by AI agents themselves.
$COGNITION - Cognition's Devin coding agent is using Exa for web search and seeing significant improvements in accuracy and fewer mistakes, demonstrating strong product-market fit for AI coding agents.
$AIMODELS - The trend toward smaller, hyper-intelligent models using retrieval tools will be more efficient than large models, solving the token apocalypse and compute shortage. This represents a fundamental shift in AI architecture that will become very noticeable by end of 2026.
$INFRA - Infrastructure to handle exponentially more searches from AI agents (10x-1000x more than Google) hasn't been built yet, creating a massive opportunity for vector databases and high-throughput infrastructure.
$ALTDATA - There is a massive opportunity in accumulating and organizing the world's unrecorded data, as agents will increasingly need comprehensive information that doesn't currently exist on the web.
Bearish:
$GOOGL - Google's search is optimized for consumer queries but fails at deep research and comprehensive information retrieval that businesses and AI agents need. Their 20 years of click data advantage doesn't matter for agentic search, and they face structural disadvantages in serving the new agentic economy.