3.4x revenue in two years. And it's from hosting apps that AI built.
What's going on?
Vercel CEO Guillermo Rauch publicly signaled IPO readiness at the HumanX conference in San Francisco last week. "Vercel is very much a work-in-public company," he said, adding that "the company's ready and getting more ready for it every day".
This isn't just talk. The numbers back it up. Vercel's annual recurring revenue (ARR) surged 240% from $100M in early 2024 to a $340M run rate by February 2026. That's 86% year-over-year growth on a GAAP basis.
The key driver? AI agents. According to Rauch, 30% of all apps running on Vercel's platform were created by AI agents, not human developers. Forbes reporting reveals an even more telling stat — roughly 70% of agent-driven deployments come from Claude Code.
Vercel started as a web deployment and hosting platform. It also builds and maintains Next.js, the dominant React framework. But as AI coding took off, all that AI-generated code needed somewhere to go. As Rauch put it: "All of that software needs to go somewhere, and we think it's going to be Vercel".
What's actually different?
Here's the key shift — we're moving from "developers deploy" to "agents deploy". Rauch made this vivid at HumanX: "When I started this company, only tens of millions of people could deploy. Now we're seeing that everybody in the world can create an app".
| Traditional web hosting | Agentic era infrastructure | |
|---|---|---|
| Who creates apps | Developers | Developers + AI agents + non-devs |
| Deploy frequency | A few times per week | Dozens to hundreds per hour |
| App lifespan | Long-running | Micro apps (ephemeral to short-lived) |
| Key customers | Enterprise dev teams | Solo founders, non-devs, AI agents |
| Growth engine | Sales/marketing | LLM ecosystem integration (Claude Code, etc.) |
The Forbes data makes this concrete. Claude Code users represent just 1% of Vercel's customer base, but they generate nearly 15% of all deployments. Agent-driven deployments jumped from ~5% in June 2025 to 21% by February 2026.
This isn't accidental. Next.js is so deeply embedded in LLM training data that when AI models like Claude generate code, they naturally default to Next.js and suggest deploying to Vercel. As Accel partner Dan Levine put it: "LLMs seem to love Vercel, and we love them back".
Agents are very prolific at deploying.
— Guillermo Rauch, Vercel CEO (HumanX 2026)
A decade of groundwork
What makes Vercel's growth even more impressive is that this wasn't built overnight. SaaStr's analysis shows this is a classic "10 years of preparation, 1 year of explosion" story.
- 2015: Started as ZEIT
Rauch arrived from Argentina with a vision to make web deployment as simple as typing a single command. Early investor interest was lukewarm. - 2016: Open-sourced Next.js
This framework became the ultimate Trojan horse. It now sees over 500 million weekly downloads, powering frontends for Grok, Claude, and Cursor. - 2019–2021: Slow growth
Revenue went $1M (2019) → $5M (2020) → $21M (2021). Solid, but not AI-level explosive. - 2023–2024: AI SDK + v0 launch
AI SDK (3M+ weekly downloads) and vibe coding tool v0 (3.5M users) marked the pivot to AI infrastructure. - Sep 2025: Series F at $9.3B
Co-led by Accel and GIC with $300M investment. Valuation tripled in 15 months.
Rauch's personal story is compelling too. Growing up in Lanús, south of Buenos Aires, he taught himself to code on a PC his father bought when he was 7. He had to learn English first because there were no programming materials in Spanish. By 17, Facebook tried to hire him but backed off when they learned his age.
The infrastructure war has started
Vercel isn't the only one seeing this opportunity. On the same day (April 13), Cloudflare announced its Agent Cloud expansion. Cloudflare CEO Matthew Prince declared: "The way people build software is fundamentally changing. We are entering a world where agents are the ones writing and executing code".
| Player | Strategy | Strength |
|---|---|---|
| Vercel | Next.js + v0 + AI SDK ecosystem | LLM-native deployment, dev community |
| Cloudflare | Agent Cloud + Workers + Sandboxes | Global edge network, security |
| AWS | Lambda + Bedrock | Enterprise scale, service breadth |
| Supabase | Open-source backend + Edge Functions | Developer-friendly, $5.1B valuation |
| Netlify / Render | Deploy automation | Simplicity, existing customer base |
Vercel's moat is interesting. Rauch told Forbes that "the last thing that you want to vibe code and reinvent from scratch is the foundational stuff that's going to run your software". Infrastructure, like payments (Stripe), is harder for AI to replace.
But Accel's Levine was candid: "It's easier to build a Vercel competitor than it was before." The key is making a service so superior that competitors don't bother: "Why would you want to do that?".
Business insight
Vercel's growth formula is elegantly simple: (1) Dominate the developer ecosystem with open source → (2) Get embedded in LLM training data naturally → (3) Become the default deployment target for AI-generated code. This "LLM-native distribution" model is a genuinely new growth strategy for the AI era.
Getting started with AI agent infrastructure
- Prototype with v0
Try Vercel's vibe coding tool at v0.dev. Generate web apps from natural language prompts. Free tier available. - Build agents with AI SDK
Vercel AI SDK (ai-sdk.dev) connects 60+ AI models through a unified interface. Start with npm install ai. - Set up Claude Code + Vercel workflow
Create Next.js projects in Claude Code and deploy with vercel deploy. 70% of agent deployments take this path. - Compare competing platforms
Evaluate Cloudflare Workers (edge computing), Supabase (all-in-one backend), and Netlify (simplicity) based on your use case.
