What is Hosting?

Why you can’t just use your laptop

When you run npm run dev on your computer, you can see your app at localhost:3000. But:

1. Only you can access it - localhost means “this computer”
2. It stops when you close your laptop - no sleep mode allowed
3. Your home network blocks incoming traffic - your router won’t let strangers in
4. It can’t handle load - your laptop will melt if 10,000 people visit

You need a computer that:

• Has a public IP address anyone can reach
• Runs 24/7 without sleeping
• Has fast internet that can handle many connections
• Is built for this job

That’s a server. And hosting is renting access to servers.

The hosting landscape (simplified)

Tier 1: Deploy in 30 seconds (start here)

These platforms take your code and handle everything else:

How it works: Connect your GitHub repo. Push code. They build and deploy automatically. Done.

Tier 2: More control, more complexity

Tier 3: Maximum power, maximum complexity

Most vibecoders never need Tier 3. Start with Tier 1. Graduate when you hit limits.

What “serverless” actually means

You’ll hear “serverless” a lot. It doesn’t mean no servers—it means you don’t manage them.

Traditional hosting: You rent a server that runs 24/7. You pay even when nobody’s using it.

Serverless: Your code only runs when someone makes a request. You pay per execution. When nobody’s visiting, you pay nothing.

Vercel, Netlify, and Cloudflare Workers are serverless. Your functions wake up when needed and go to sleep after.

Serverless trade-offs:

• ✅ Scales automatically
• ✅ Pay only for what you use
• ✅ No server maintenance
• ❌ Cold starts (first request can be slow)
• ❌ Execution time limits (can’t run forever)
• ❌ Some things are harder (websockets, long processes)

The cost reality

Free tier gets you far: A personal project or early startup can run on Vercel’s free tier indefinitely.

Databases cost money: The app itself might be free, but a PostgreSQL database starts at $5-15/month.

AI APIs add up: Your hosting might be $0, but Claude/OpenAI calls are the real expense. See What is a token?

Scale costs scale: Traffic, storage, and compute all increase with success. A viral launch can blow through free tiers in hours.

Typical monthly costs for a small SaaS:

• Hosting (Vercel/Railway): $0-20
• Database: $5-25
• AI API: $10-500 (varies wildly)
• Domain: $1 (yearly cost / 12)
• Total: $15-550/month

How to choose

Just starting? Use Vercel or Netlify.

• Connect GitHub, push code, it deploys
• Free tier is plenty for learning
• Automatic HTTPS, preview deployments, the works

Need a database too? Add Railway or Supabase.

• Railway gives you Postgres alongside your app
• Supabase is Postgres + auth + storage in one

Outgrowing free tiers? Stay on the same platform, pay.

• Vercel Pro, Railway paid tier, etc.
• Migration is painful—avoid it if you can

Specific requirements? Then research.

• Need GPU? Look at Modal, Replicate, Banana
• Need to run in EU only? Check data residency options
• Need persistent websockets? Fly.io or traditional servers

What hosting actually does

When someone visits myapp.com:

1. DNS translates the domain to an IP address
2. The request reaches your hosting provider’s servers
3. Your code runs (or wakes up if serverless)
4. A response goes back to the visitor’s browser

The hosting provider handles:

• Running your code
• Routing requests to the right place
• SSL certificates (HTTPS)
• Scaling up when traffic spikes
• Keeping things running when hardware fails

That’s a lot of complexity you don’t have to think about.

Further reading

• What is localhost? : Where you test before deploying
• What is a domain name? : The address that points to your hosting
• What is deployment? : The process of getting code to your host
• What is the cloud? : The infrastructure underneath all of this