Is Vercel Burning Your Budget?

How to Actually Reduce Costs in React & Next.js — Without Falling Into Optimization Fanaticism

Vercel is one of the best platforms for deploying React and Next.js applications. Fast deployments, great developer experience, edge infrastructure — it's easy to see why so many teams choose it.

But sooner or later, many teams hit the same moment: the bill starts growing faster than the business.

Let's be clear upfront: In most cases, the problem is not Vercel. The problem is architectural decisions that felt reasonable early on — and became expensive at scale.

In this article, we'll cover:

• why Vercel costs often spike unexpectedly
• what actually affects billing (and what doesn't)
• how to reduce Next.js costs without breaking your architecture
• where optimization turns into unhealthy fanaticism

---

How Vercel Charges You (Simplified)

The main cost drivers are:

1. Serverless / Edge function executions
2. SSR and ISR rendering
3. Middleware invocations
4. Build executions
5. Bandwidth (especially images)
6. Analytics and logging

The most common mistake is focusing on traffic volume, while ignoring rendering strategy.

---

The #1 Cost Killer: Unnecessary SSR

A very common setup

• Next.js App Router
• In the root layout.tsx, you use:
• headers()
• cookies()
• auth()
• draftMode()

Result:

• the entire route tree becomes dynamic
• every request triggers a server execution
• bots and crawlers cost money too

Even for simple marketing pages.

The fix

Strictly separate concerns:

• Marketing & public pages → static or ISR
• App / dashboard → SSR
• Admin / CMS preview → SSR + draftMode

Example:

export const dynamic = 'force-static'

or

export const revalidate = 3600

This single change often cuts costs dramatically.

---

Middleware: Small Code, Big Bills

Middleware runs on every matched request — including:

• bots
• previews
• static pages
• health checks

Common misuse

Using middleware for:

• redirects
• i18n
• analytics
• A/B testing
• marketing logic

Better approach

• No middleware on public routes
• Redirects via vercel.json
• Locale via routing
• Auth only inside the app

"matcher": ["/app/:path*"]

---

ISR Can Work Against You

ISR is powerful, but dangerous when misconfigured.

Typical problems:

• very short revalidate times
• many pages
• frequent invalidation

Red flags

• revalidate: 60 on SEO pages
• cron jobs triggering revalidation
• CMS saving causes rebuild storms

Recommendations

• SEO pages: revalidate: 3600–86400
• Blog / cases: ISR + on-demand revalidation
• Rarely changing pages: full static

---

Image Optimization: Useful, But Not Free

Next/Image is great — but:

• every unique size = a new transformation
• mobile + desktop + retina multiplies usage

Common mistakes

• too many responsive sizes
• dynamic image dimensions
• CMS without constraints

Best practices

• fixed size sets
• aggressive CDN caching
• pre-optimized WebP / AVIF for hero images

---

Builds: The Quiet Cost Center

Every build costs money.

Typical causes

• preview deployments on every commit
• bots triggering deploys
• CMS hooks rebuilding the site

Basic hygiene

• limit preview deployments
• avoid rebuilds on content drafts
• audit deploy hooks

---

Analytics and Overengineering

Analytics is rarely the main issue — but:

• server-side tracking
• verbose logs
• debug modes

can quietly inflate bills.

Rule of thumb

• client-side analytics for marketing
• server-side only when justified
• minimal logs in production

---

When Optimization Becomes Fanaticism

This is critical.

❌ Bad optimization:

• rewriting architecture to save €20/month
• hurting developer experience
• adding complexity without business value
• treating SSR as the enemy

✅ Healthy optimization:

• remove SSR from public pages
• limit middleware
• configure caching properly
• understand what you're paying for

If the product makes money, infrastructure should support growth, not just minimize costs.

---

Cost Reduction Checklist

1. Dynamic APIs in public layouts?
2. Middleware on /?
3. SSR where static would work?
4. Revalidation frequency
5. Build count
6. Image optimization usage
7. Preview deployments
8. CMS draft modes
9. Server-side analytics
10. Bot traffic

---

Final Thoughts

Vercel doesn't "burn money". It charges transparently for architectural choices.

In most cases, you don't need:

• migration
• self-hosting
• Kubernetes

You need:

• clearer rendering boundaries
• less middleware
• smarter caching

Optimization is a tool. Fanaticism is when it starts hurting the product.