Vercel

Vercel is the cloud platform built by the team behind Next.js. It turns a git push into a globally distributed deployment with previews, instant rollbacks, and edge functions, then layers AI-native primitives like v0, the AI Gateway, and AI SDK on top. Adobe, Sonos, The Washington Post, and most modern Next.js teams run production traffic on it.

What Vercel does:

• •Push-to-deploy CI/CD connect a GitHub, GitLab, or Bitbucket repo and every commit gets a unique preview URL, with production deploys gated behind branch rules and instant rollback to any prior deploy.
• •Global edge network requests serve from the closest of dozens of regions worldwide, with automatic image optimisation, smart cache headers, and zero-config CDN.
• •Vercel CLI a full command-line tool (vercel deploy, vercel env, vercel logs) for scripting deploys, managing env vars across environments, pulling production secrets, and tailing live logs.
• •v0 AI app builder generate full Next.js components from a prompt, iterate in a browser canvas, then push directly to a Vercel project without leaving the v0 workspace.
• •AI Gateway and AI SDK one API surface that routes prompts to GPT, Claude, Gemini, Llama, and others, with built-in caching, retries, observability, and the open-source AI SDK for streaming UIs.
• •Vercel Functions and Edge Middleware serverless TypeScript functions plus middleware that runs on the edge for auth, A/B tests, geo personalisation, and rewrites.
• •Cache Components and ISR the new Cache Components model in Next.js 16 plus revalidateTag and revalidatePath give precise cache control without manual purge scripts.
• •Vercel Storage stack managed Postgres, KV, Blob storage, and Edge Config behind a single dashboard, billed by usage and pre-wired to Functions.
• •Preview environments per pull request every PR gets its own URL with isolated env vars, so reviewers and product see real changes before merge.
• •Native MCP support and Vercel Agent recent additions that let AI agents read deploys, query logs, and ship fixes through the same Model Context Protocol surface as other modern AI tools.

Pricing:

• •Hobby $0: free forever for personal projects, with 100 GB bandwidth, preview deploys, and a soft fair-use cap on serverless executions.
• •Pro from $20/mo per user: 1 TB bandwidth, advanced analytics, unlimited team members in trial mode, plus metered usage for functions, bandwidth, image optimisation, and storage.
• •Enterprise custom pricing: SSO, SAML, audit logs, dedicated support, isolated build infrastructure, and contractual SLAs for large teams.

Limitations:

• •Usage-based bills can spike without warning a viral post or a malicious traffic spike can push bandwidth and function execution costs into hundreds or thousands of dollars overnight, since Vercel bills for all traffic including bot and DDoS hits unless you configure firewall rules.
• •Hobby plan is not a production tier the free plan caps bandwidth at 100 GB and serverless execution at 60 seconds, so streaming AI responses or long-running tasks will hit 504 errors and force a Pro upgrade.
• •Pricing model is hard to forecast bills span bandwidth, function execution, image optimisation, storage, edge middleware, and AI Gateway tokens on separate meters, so accurately predicting a monthly bill requires tracking five or six usage dimensions at once.