Vercel vs Cloudflare Pages: Measured, Not Marketed
Disclosure: stackverified.com itself runs on Vercel — we pay them, and this benchmark partly audits our own hosting choice. Both vendors' pricing pages were fetched live on the audit date; TTFB samples were run from a single US residential vantage point (limitation stated plainly below). No affiliate relationship with either platform.
Methodology: N=8 sequential HTTPS fetches per target, time-to-first-byte measured client-side, cache-status headers recorded, from one US residential connection. Targets: each platform's own flagship properties (vercel.com, nextjs.org, developers.cloudflare.com, pages.cloudflare.com). Single-vantage — your region will differ; the point is the method, and the bill.
🚨 The Executive Verdict
For cached static content, edge speed is a statistical tie — and any review that claims a winner on TTFB alone is selling you something. The real differences are metering, bandwidth caps, and functions.
Our samples (N=8, cache HIT on every request): vercel.com median 108ms, nextjs.org 110ms, developers.cloudflare.com 111ms, pages.cloudflare.com 205ms (single heavier page, max 424ms). Read that honestly: three of four targets sit within noise of each other at ~110ms from our vantage. Static assets on cached edges are fast everywhere in 2026.
1. What We Measured and Pulled (Aug 29, 2026)
| Line item | Vercel | Cloudflare Pages |
|---|---|---|
| Cached TTFB (median, N=8) | 108-110 ms (HIT) | 111-205 ms (HIT) |
| Bandwidth model | Metered: then ~$0.15/GB | Unlimited on all plans |
| Request metering | Edge requests from ~$2/1M | Not billed per-request on Pages |
| Entry paid tier | Pro with $20 usage credit | Free; Workers Paid $5/mo |
| Serverless runtime | Node functions (warm after cold start) | Workers isolates (no traditional cold start) |
2. The Bill Is the Product
Vercel's model is generous at small scale ($20/mo usage credit on Pro) and precise at scale: every edge request, every GB of fast data transfer, every image optimization is itemized. That transparency is genuinely good — you can see exactly what drives cost.
Cloudflare Pages' model is blunter: unlimited bandwidth on every plan, no per-request billing on Pages. A viral traffic spike that would triple a Vercel bill costs Cloudflare nothing extra. The flip side: the runtime is Workers' isolates — fast and cold-start-free, but a different runtime with real constraints (no Node APIs natively, CPU-time limits).
🔮 Decision Rule
Vercel if…
- • You're on Next.js — it's the reference runtime
- • You need full Node functions, image optimization, ISR
- • Traffic is predictable and you want itemized cost control
Cloudflare Pages if…
- • Traffic is spiky or huge (unlimited bandwidth is real money)
- • Your stack fits Workers runtime constraints
- • You want a $0 floor that survives virality
Our own stack sits on Vercel because this site is Next.js and the usage profile is quiet. If StackVerified ever goes viral, this article is the first thing we re-read.
Hosting Costs You Haven't Modeled?
The Free Stack Audit maps your real traffic profile against both rate cards.
Take the Free Audit ←Methodology: N=8 sequential HTTPS fetches per target from one US residential vantage, Aug 29, 2026; TTFB = time to first response byte; every sampled response was a cache HIT. Pricing captured live from both vendors' pages the same day. Single-vantage benchmarks are directional — run your own from your users' regions. Powered by StackVerified Intelligence.