Back in late 2023, I moved a client's e-commerce storefront off Vercel after a single month's bill came in at $340. Mid-traffic SaaS. Nothing extraordinary. The founder rang me on a Friday afternoon sounding genuinely alarmed, and honestly, I didn't have a great defence. I'd defaulted to Vercel because it was what I knew, what felt fast to set up, and what the Next.js docs basically assumed you'd use. That migration to Cloudflare Pages took me one afternoon and cut the monthly cost to about $22. I've been paying closer attention ever since.
So here's where I actually stand in 2026 after running both platforms across client work at Seahawk: Cloudflare Pages has matured enormously, Vercel hasn't stood still either, and the answer to "which one should I use" is less about brand loyalty and more about what you're actually building.
The Pricing Reality (And Where Vercel Still Stings)
Vercel's free tier is genuinely good for side projects. Hobby plan, unlimited deployments, preview URLs, decent build minutes. The moment you go Pro ($20/month per seat), you're fine for a solo developer. But agencies? The per-seat model stacks up fast. Six developers on a team, that's $120/month before you've served a single request. Then you add bandwidth, Edge Function invocations, and Image Optimisation, and the bill starts to look a lot like a mobile phone contract from 2009 where you never quite understood what you were paying for.
Cloudflare Pages sits on a fundamentally different economic model. The free tier includes unlimited sites, unlimited requests, and 500 builds per month. The Pro plan is $20/month flat, not per seat. For Seahawk, that distinction is significant because we manage builds for multiple clients from a single account. I ran the numbers on a typical month in early 2025: 14 active projects, around 2.1 million edge requests combined. On Cloudflare, we paid $20. On Vercel's equivalent usage tier, the estimate came out to roughly $180-$220 depending on which add-ons were active.
That said, Vercel's pricing does buy you something real. More on that in a bit.
The Bandwidth and Egress Trap
This is the thing nobody warns you about until it's too late. Vercel charges for bandwidth on the Pro plan after 1TB. That sounds fine. Most projects never get there. But Edge Functions that return large responses, or projects doing server-side image processing, can chew through bandwidth in ways that feel disproportionate.
Cloudflare's entire network runs on unmetered egress within their platform. Cloudflare's pricing page makes this explicit. Workers and Pages Functions requests are priced at $0.30 per million after the free 100k daily requests. That's transparent. Predictable. I can sleep.
Edge Network Coverage: Cloudflare Wins, But Not by Magic
Cloudflare has over 300 data centres globally as of 2026. Vercel runs on a smaller set of regions, backed largely by AWS infrastructure. For most UK and European projects, the difference in cold-start latency at the edge is real but modest. For a SaaS product serving users in Lagos, Mumbai, or São Paulo simultaneously? Cloudflare's coverage gap narrows to nearly nothing.
I worked on a multilingual SaaS project in 2024, serving users across Southeast Asia. We'd initially launched on Vercel and the p99 latency from Singapore was hovering around 380ms. Not terrible. We moved the API middleware to Cloudflare Workers and got that down to 95ms. Same code, broadly. Different edge node density.
Workers vs Edge Functions: The Runtime Divergence
Here's where things get genuinely technical and genuinely important. Vercel Edge Functions run on the V8 isolate model (same philosophy as Cloudflare Workers) but they're tied more tightly to Vercel's own infrastructure and the Next.js middleware system. Cloudflare Workers have been at this longer. The Workers runtime supports Web Crypto, Streams, KV, R2, Durable Objects, and a suite of bindings that have no direct equivalent on Vercel.
If you're building something that needs persistent state at the edge (think: presence indicators, rate limiting with shared state, real-time collaborative features), Cloudflare Durable Objects are genuinely useful and there's nothing on Vercel that matches them today.
That said, if you're deploying a Next.js 14+ app with App Router and you want zero-configuration ISR, Edge streaming, and middleware that "just works" with your RSC setup, Vercel is still the smoother path. That's not opinion. That's just what I see every time I try to replicate Vercel's Next.js integration on Cloudflare Pages using the @cloudflare/next-on-pages adapter. It works. It's just not frictionless.
Developer Experience: The Honest Version
I want to be careful here because DX is subjective and also where the most tribal takes live. So let me just describe what I actually observe.
Vercel's dashboard is beautiful. Preview deployments are instant. The GitHub integration is the best I've used anywhere. When you push a branch, the preview URL is live in under 45 seconds on a warm cache, and the deployment comment on your PR shows you Core Web Vitals, build duration, and a direct link. That feedback loop is fast and it feels good. That matters, especially when you're onboarding a non-technical client who wants to click a link and see their changes.
Cloudflare Pages has improved its dashboard considerably since the 2022-era days when it felt like an afterthought bolted onto Workers. The build system is faster now, preview deployments work well, and the Pages + Workers integration (via wrangler.toml) is genuinely nice. But the mental model of Cloudflare's product surface is broader and therefore harder. You've got Workers, Pages, R2, KV, D1, Durable Objects, Pages Functions, and they all have slightly different configuration surfaces. The documentation at developers.cloudflare.com is thorough but sprawling.
Honestly, I've seen junior developers at Seahawk take to Vercel in a day. Cloudflare Pages takes about a week before someone stops second-guessing whether their function is a Worker or a Pages Function and which config file governs what.
Local Development and Wrangler
Cloudflare's wrangler dev CLI is good. It runs a local simulation of the Workers runtime and supports most bindings. But "simulation" is the operative word. I've shipped code that worked locally and behaved differently on the actual edge network because of subtle differences in how KV behaves in preview versus production. Small bug in 2025, caught in staging, but it cost me half a day.
Vercel's vercel dev command is more limited in some edge cases but it's almost always a faithful replica of production for standard Next.js projects. That predictability matters.
Framework Support Beyond Next.js
This is an underrated conversation. Vercel is, functionally, a Next.js hosting company that also supports other things. SvelteKit, Nuxt, Astro, Remix, SolidStart all work, and they work well. But Next.js gets features first, gets better defaults, and is clearly the primary use case.
Cloudflare Pages has no framework allegiance. Astro on Pages is a joy. SvelteKit adapter for Cloudflare is maintained and stable. I've been running a handful of Astro projects on Pages since mid-2024 and the build times are faster than equivalent Vercel deployments for the same sites (roughly 30-40% faster on average, based on my own build logs).
If your team is moving away from Next.js or building with multiple frameworks across projects, Cloudflare Pages is more agnostic and that agnosticism pays off over time.
Where I'd Pick Each One in 2026
Let me be direct. Here's the decision logic I actually use:
Pick Vercel if:
- Your project is Next.js-first and you're using App Router deeply (RSC, streaming, ISR, all of it)
- Your clients need to see preview deployments and comment on them without any technical friction
- Your team is small (1-3 developers) and the per-seat cost is negligible
- You're building something where first-class Next.js support is worth a premium
Pick Cloudflare Pages if:
- You're an agency or solo operator running more than four or five live projects
- Cost predictability matters more than the smoothest Next.js DX
- You need Workers bindings (KV, R2, Durable Objects, D1) as part of your architecture
- Your user base is globally distributed and edge latency in non-US regions matters
- You're not tied to Next.js and want framework flexibility
For Seahawk's client work specifically, we've landed on a split: Cloudflare Pages for marketing sites, content-heavy builds, and anything Astro-based. Vercel for complex Next.js SaaS products where the client's dev team is already in that ecosystem.
The One Thing Nobody Talks About: Support
Vercel Pro gets you email support. Enterprise gets more. It's responsive and the team knows Next.js deeply. If something breaks in your deployment pipeline at 2am on a Thursday before a client launch, having someone reply who actually understands what ISR revalidation is doing under the hood is worth something.
Cloudflare's support model for Pages specifically has historically been... patchy. Their Discord community is active and the Workers team is accessible, but Pages-specific support tickets can take longer to resolve than you'd want. This improved in 2025 but I'd still give Vercel the edge here (pun intended, sorry).
---
FAQ
Is Cloudflare Pages genuinely production-ready in 2026?
Yes, full stop. The "Cloudflare Pages is still beta" perception is outdated. It's running serious production workloads. The Pages platform status has been consistently strong. The main caveats are around the Next.js adapter (which is community-maintained, not first-party) and the learning curve around Cloudflare's broader product surface.
Can I use Cloudflare Pages with Next.js App Router?
You can, via the @cloudflare/next-on-pages adapter. It supports a significant subset of App Router features including Server Components, dynamic routes, and middleware. But some features that rely on Vercel-specific infrastructure (like advanced ISR with on-demand revalidation via the Vercel API) either don't work or require workarounds. If you're building a complex Next.js app, test your specific feature requirements before committing.
Does Vercel's free tier work for client sites?
Technically yes, but Vercel's Terms of Service for the Hobby plan restrict commercial use. If you're building for a paying client, you should be on Pro. This is not a grey area they ignore forever. Use Pro for client work.
Which platform is better for SEO performance?
Neither platform gives you SEO by default. But edge delivery and Core Web Vitals are related, and both platforms can produce excellent LCP and CLS scores with good implementation. Vercel's Image Optimisation API (via next/image) is more mature and easier to configure. Cloudflare's Image Resizing feature (available on Pro plans) requires more manual setup but is less expensive at scale.
What about Netlify? Is it still relevant?
Netlify is still around and still used, particularly for JAMstack teams with deep Netlify-specific workflows. But in terms of mindshare, adoption, and platform investment in 2026, it's trailing both Cloudflare and Vercel noticeably. I haven't recommended Netlify for a new project since about 2023.
---
The platform choice in 2026 is genuinely less dramatic than the Twitter threads make it sound. Both work. Both are reliable. The real decision is about cost model alignment with your business and framework alignment with your team. Make that call deliberately rather than defaulting to whatever the framework docs show in their "deploy" button.