Three years ago I put a client's Next.js storefront on a "budget cloud" provider I won't name, because they'd already paid for a year upfront and I didn't want to argue. First flash sale event, the site fell over in fourteen minutes. Fourteen. The client lost roughly £28,000 in orders by their estimate. I've never made that mistake again, and that afternoon is why I'm fairly blunt when people ask me where to host their frontend in 2026.
There's no single right answer, obviously. But there are a lot of wrong answers, and this post is mostly about those.
I'm Gautam. I run Seahawk Media out of London. We've built north of 12,000 sites across every vertical you can imagine. I've had enough firsthand disasters and quiet wins with every major platform to have a real opinion, not just a comparison table I nicked from someone else.
Here's the shortlist.
What "Frontend Hosting" Actually Means in 2026
Worth being precise here, because the category has drifted.
In 2020, frontend hosting meant: deploy your static files, get a CDN, go home. Dead simple. Now a "frontend host" might be running your SSR Node processes, handling your edge functions, managing preview environments, and syncing environment variables across 40 branches. That's not a CDN. That's basically a platform-as-a-service that happens to speak Git.
The hosts worth talking about in 2026 are all operating in that wider territory. If you're still thinking about this as "where do I put my HTML files," you can stop reading and just pick Cloudflare Pages on a free plan. Done. But if you're running Next.js, Nuxt, SvelteKit, Remix, Astro, or anything with server components or edge rendering, the decision genuinely matters.
The three things that separate good from bad
- Cold start latency on serverless functions (this will hurt your TTFB more than anything)
- How well the platform's build system understands your framework (not all of them speak Next.js App Router fluently yet)
- The pricing cliff: where does the free tier end and does the paid tier make any sense for a £3,000/month agency client
Vercel: Still the Default, Still Expensive
I'll be honest: Vercel is the easy answer and also the one that caused me the most invoice arguments with clients in 2024.
The product is excellent. There's no real debate about that. Their Next.js integration is obviously unmatched because they built Next.js. Edge functions cold start in under 50ms consistently. Preview deployments work exactly as advertised. The DX (developer experience) is probably the best in the category, full stop.
But here's the thing: the pricing above the Hobby tier is genuinely punishing for agency work. Pro is $20/month per member, which sounds fine until you've got a team of eight and three separate client accounts. And the bandwidth overages on builds for a high-traffic e-commerce client will ruin your margin if you're not watching carefully.
Back in late 2023, Seahawk had a SaaS client running a fairly heavy Remix app on Vercel Pro. Their bill crept from $180/month to $610/month over six months purely because of function invocation counts on their dashboard routes. Nobody noticed until the client's accountant flagged it. That's not Vercel doing anything wrong, technically. But it's the kind of thing you need to model before you commit.
Who it's genuinely right for: Funded startups with proper budgets, Next.js App Router projects where you want framework-native support, teams that treat DX as a genuine business investment.
Who should probably look elsewhere: Cost-sensitive freelancers, agencies managing lots of small-to-mid client sites, anyone on SvelteKit or Astro who doesn't need the Vercel-specific features.
Netlify: The One That Lost Its Way (a Bit)
I say this with some affection because Netlify basically invented this category. They were doing atomic deploys and deploy previews before anyone else took it seriously.
The problem is that Netlify in 2026 feels like a product committee made every decision after 2022. Features got added, the UI got busier, the pricing tiers got confusing, and the core product got slower. Their build times have regressed. I ran informal comparisons on the same Next.js 14 app in late 2024: Netlify was averaging 4m 20s builds versus Vercel at 2m 45s. Not catastrophic. But noticeable.
Where Netlify still shines: forms (their built-in form handling is genuinely convenient for marketing sites), their CMS integrations, and the fact that their free tier is more generous than Vercel's for pure static sites. If you're running a Gatsby or Eleventy site for a local business client, Netlify free is totally solid.
One thing I do like: their Edge Functions are underrated. Deno-based, low cold starts, and the local dev simulation with netlify dev is actually better than Vercel's for certain workflows. I used it heavily on a headless Shopify project for a fashion brand in Birmingham. Worked cleanly.
Who it's right for: Marketing sites, agencies who got locked in early and have existing workflows, anyone leaning on Netlify's CMS or form features.
Who should look elsewhere: Anyone prioritising build speed, teams running SSR-heavy apps, anyone who finds the current pricing structure hard to reason about (it is).
Cloudflare Pages: The Underdog That's Quietly Very Good
If I'm being completely honest, Cloudflare Pages has become my default recommendation for a surprising number of projects in the last 18 months.
The free tier is absurdly good. Unlimited sites, unlimited requests, unlimited bandwidth. Yes, there are limits (500 builds per month on free, for instance), but for the kinds of sites that need a free tier, those limits are fine.
The Workers integration is where it gets interesting. Cloudflare's edge network is the largest in the world by PoP count, which is not marketing fluff, it genuinely affects latency for globally distributed audiences. If you've got a client with 40% of their traffic coming from Southeast Asia, Cloudflare Pages will serve them faster than Vercel by a meaningful margin.
The trade-off: framework support is still catching up. Next.js App Router, specifically React Server Components, has had a rocky history on Pages. It works better than it did a year ago, but there are edge cases, particularly around streaming and certain caching behaviours, that will bite you if you're not careful. I wouldn't deploy a complex Next.js 15 app with heavy RSC usage on Pages without thorough testing first. Astro, SvelteKit, and pure static sites though? Completely fine.
Who it's right for: Cost-conscious agencies, global audiences, Astro/SvelteKit/Qwik projects, anyone already in the Cloudflare ecosystem.
Who should look elsewhere: Teams with complex Next.js RSC setups, anyone who needs enterprise SLAs out of the box.
Render: The Sensible Middle Ground Nobody Talks About Enough
Render doesn't get the coverage it deserves, probably because it doesn't have a VC marketing budget.
What Render does well: it thinks like a full-stack platform while still being genuinely usable for frontend-only work. You can run your Next.js app as a Node service, set up a Postgres database, add a background worker, and manage it all from one dashboard with sensible pricing. That's a different mental model from Vercel or Netlify, but for certain project types it's exactly right.
The free tier for static sites is competitive. Paid plans start at $7/month per service, which is much easier to explain to a small business client than Vercel's per-seat model.
I used Render for the first time properly on a fintech dashboard project in early 2024. Next.js frontend, FastAPI backend, Redis for sessions. Having it all in one place with shared environment variable management was genuinely pleasant. No separate Railway account for the backend, no stitching together billing from three providers.
Cold starts on Render's free tier are the main complaint you'll hear, and it's legitimate. Spinning up from idle can take 30-60 seconds. On a paid plan it's fine. So just don't use the free tier for anything user-facing.
Who it's right for: Full-stack projects, teams wanting a unified platform across frontend and backend, solo developers managing costs carefully.
Who should look elsewhere: Pure static sites (Cloudflare is better and free), anyone who needs the most polished DX available (Vercel is better).
Fly.io: When You Actually Need Control
Fly.io is not a typical frontend host. But I include it here because there's a category of project, usually anything with persistent connections, WebSockets, or heavy server-side state, where Fly.io is simply the right answer and none of the others are.
Fly runs containers. Real containers, close to metal, across 35+ regions. It's not a managed CDN deploy pipeline. You write a Dockerfile and Fly runs it. That means you can run a Remix app with SQLite (using LiteFS), or a SvelteKit app with a persistent socket server, in ways that just aren't possible on the edge-function model.
The learning curve is steeper. The CLI takes getting used to. And their documentation, while improving, can feel like it assumes you already know what you're doing. But if you do know what you're doing, Fly is remarkably powerful.
Who it's right for: WebSocket-heavy apps, anything needing persistent server state, developers comfortable with containers who want global distribution without the managed-platform constraints.
Who should look elsewhere: Beginners, anyone wanting point-and-click deploys, pure static sites.
How to Actually Choose (A Simple Decision Process)
- Is it a pure static site (no SSR, no edge functions, just HTML/CSS/JS)? Use Cloudflare Pages, free tier, done.
- Is it Next.js with App Router and RSC as a core feature? Use Vercel. Pay the money, it's worth it for that specific stack.
- Is it SvelteKit, Astro, Nuxt, or Remix on a budget? Try Cloudflare Pages first. If you hit framework edge cases, move to Render.
- Does your project include backend services and you want one dashboard? Use Render.
- Do you have persistent connections, WebSockets, or need true container control? Use Fly.io.
- Are you a freelancer managing 20+ small client sites and cost is the main variable? Cloudflare Pages for the statics, Render for anything with a server component, Vercel only when the client specifically needs it.
The Pricing Reality Check
Nobody talks about this honestly enough. Here's what I actually spend across Seahawk's active projects right now:
- Vercel Pro: reserved for two clients with Next.js App Router builds where the framework support is non-negotiable. We pass the cost through directly.
- Cloudflare Pages: eleven sites on free, three on the $20/month Pro plan for build limit reasons.
- Render: four projects, all on the $7/month individual service tier. Total: around $28/month for those four.
- Fly.io: two projects with WebSocket requirements, approximately $18/month combined.
Total hosting outlay for the agency's own managed portfolio: under $120/month for north of 20 live sites. The Vercel bills are client-direct.
If you're paying significantly more than this for a comparable number of projects, I'd genuinely audit your setup.
FAQ
Is Vercel worth the price in 2026?
For Next.js App Router projects specifically, yes, probably. The cold start performance, build times, and framework-native features justify the cost if Next.js is a core part of your stack. For anything else, you're paying a premium that doesn't convert into user-facing improvements.
Can Cloudflare Pages handle a high-traffic site?
Yes, comfortably. Cloudflare's network handles billions of requests daily. The free tier's "unlimited bandwidth" claim is real for static assets. The only serious constraint is the 500 builds/month on free and the framework compatibility caveats I mentioned for complex Next.js setups.
What about AWS Amplify or Firebase Hosting?
I didn't include them because, honestly, neither is a strong recommendation for modern frontend work in 2026. Amplify's DX has always been behind the dedicated platforms, and Firebase Hosting is best suited to projects already deep in the Firebase ecosystem. If you're already there, use it. Otherwise, start with one of the five I've listed.
Is Netlify dying?
No, but it's not growing the way it should be. The product still works. The company raised a significant round and continues to ship features. But they've lost the "obviously the best" status they had in 2019-2021, and for most new projects I'd pick something else first.
What's the best option for a freelancer just starting out?
Cloudflare Pages for static sites, Render for anything with server-side logic. Both have free tiers that are genuinely usable, not just trial bait. Start there, move to Vercel only when a specific project demands it.
---
The hosting conversation is one I have at least twice a week with clients, agency peers, and developers in our community Discord. Nobody gets it perfectly right every time (I certainly haven't). But if you use the decision process above rather than defaulting to whatever you used last year, you'll waste a lot less money and have a lot fewer conversations like the one I had with that client after fourteen minutes of downtime.
Pick the right tool. Then go build something worth hosting.