Cloudflare bought Vite to destroy Vercel

TL;DR

• 1

  Cloudflare bought a deployment abstraction, not just Vite: Theo argues the real prize is Void Zero's push to make infrastructure derive from code, so a Vite app can go live with commands like void deploy instead of hours of setup.

• 2

  AI changed the bottleneck from writing code to shipping it: he contrasts old projects that took 30 to 40 hours to build and 2 to 3 hours to deploy with today's agent-assisted apps that can be coded in minutes but still burn hours on auth, databases, and cloud config.

• 3

  Cloudflare's infra is strong, its DX has been the problem: Theo says Workers, CDN, and edge compute are powerful, but getting real apps onto Cloudflare has meant Wrangler, YAML, conflicting docs, and platform-specific hacks that Next.js on Vercel usually avoids.

• 4

  Vercel solved framework-to-compute, but stopped short of the data layer: his platform map shows Vercel owning a clean path from React and Next.js down to compute, while relying on partners like Clerk, WorkOS, Convex, PlanetScale, and Neon for auth and databases.

• 5

  Lakebed is Theo's proof that agent-native cloud workflows are possible: he demos npx lakebed new and npx lakebed deploy, then has Cursor build and deploy a live Kanban with chat in seconds, arguing the future is clouds shaped around what agents can do well.

• 6

  The deal makes Cloudflare a much scarier rival than Vercel for the next wave: with Evan You, Vite, Vitest, Rolldown, OXC, a new Cloudflare CLI direction, and a $1 million Vite ecosystem fund, Theo sees Cloudflare finally gaining the developer experience layer it could not build alone.