The AI Application Spending Report (a16zenterprise.substack.com)

a16z and Mercury analyzed payment data from 200,000+ Mercury customers (June–Aug 2025) to rank the top 50 “AI-native” application-layer companies by where startups actually spend money. Unlike infrastructure metrics (cloud, GPUs, models), this spend-based list reveals which AI products are being adopted in real workflows: general LLM assistants (OpenAI #1, Anthropic #2), workspace LLM integrations (Notion #10, Manus #33), meeting notetakers (Fyxer #7, Otter #41) and a large creative tools cohort (Freepik #4, ElevenLabs #5, Canva, Midjourney, Descript). The report finds 60% horizontal apps (company-wide productivity) vs 40% vertical (role-specific), with creative tools and “vibe coding” notably moving from specialist to broadly used — Replit (#3) leads vibe coding in revenue among Mercury customers, generating ~15x the revenue of Lovable. Technically and strategically, the data shows two important shifts: consumer-first AI products are rapidly moving prosumer and enterprise (≈70% of listed firms can be adopted individually and scaled to teams), and vertical AI splits into augmentors (12 of 17 verticals) and emergent agentic end-to-end “AI employees” (5 companies, e.g., Crosby Legal, Cognition, 11x). Methodology caveats: spend excludes cloud/GPU vendors and non-Mercury transactions; Google’s spend mixes Cloud and Gemini. For builders and investors, this is a real-time signal of where startups are buying AI today — and which product categories are most likely to scale or replace human workflows.