In this guide
Most hosting comparisons are written for hobby projects: which one builds fastest, which has the nicest dashboard, which supports a framework best. Those questions stop mattering the day someone pays you. Then three new questions take over:
- Is this plan allowed for a business?
- What happens when I hit a limit: a pause, an error, or a bill?
- What will I pay at 100 customers, and when I add a second person?
This guide compares Netlify, Vercel and Cloudflare Pages on exactly those questions, using each provider's own published pricing and terms as of November 2026. We don't compare build speeds or framework features; for a small product, all three are fast enough.
The short version
- Netlify counts everything in credits. Its free plan has hard limits that can't create costs, so there are no surprise bills, but sites pause when credits run out. Paid plans start at $9; Pro at $20 includes unlimited members.
- Vercel's free Hobby plan is for personal, non-commercial use. A product you charge for needs Pro at $20 per developer seat.
- Cloudflare Pages is the cheapest at small scale, with server-side code billed as Workers from $5 a month, but with tighter per-request limits.
Side by side
| Netlify | Vercel | Cloudflare Pages | |
|---|---|---|---|
| Free plan | 300 credits a month, 1 concurrent build[1] | Hobby: 100 GB fast data transfer, 1M function invocations, 1M CDN requests[2] | 500 builds a month; Workers: 100,000 requests a day[5][6] |
| Free plan for a paid product? | Pricing page states no commercial restriction | No: personal or non-commercial use only[4] | Limits page states no commercial restriction |
| At the free limit | Sites pause until the next billing cycle, all projects on the account[1] | Hobby can't buy more usage[2] | Server-side requests past the daily cap fail |
| First paid step | Personal $9, 1,000 credits | Pro $20 per developer seat, $20 usage credit included | Workers Paid from $5, 10M requests included |
| Team pricing | Pro $20 with unlimited members | $20 per developer seat | Account-level plan |
| How usage is counted | Credits for deploys, bandwidth, compute, requests | Included credit, then pay as you go | Requests and CPU time |
Netlify: predictable, with a hard ceiling
Netlify now counts all usage in credits. Each production deploy costs 15 credits, bandwidth costs 20 credits per GB, compute costs 10 credits per GB-hour, and web requests cost 2 credits per 10,000[1]. The plans:
| Plan | Monthly price | Credits | Members |
|---|---|---|---|
| Free | $0 | 300 | 1 |
| Personal | $9 | 1,000 | 1 |
| Pro | $20 | 3,000 to 20,000, by tier | Unlimited |
The free plan's limits are hard and can't incur costs, so you can't wake up to an invoice. The trade-off is what happens at the ceiling: a site that reaches its monthly limit pauses until the next billing cycle, and if one project goes over, every project on the account pauses[1]. Auto-recharge exists but is off by default.
What a month costs in credits
Hypothetical example
A small subscription product with a front end, a few server functions and modest traffic, during a normal month and during launch month:
| Usage | Normal month | Launch month |
|---|---|---|
| Production deploys | 8 × 15 = 120 | 30 × 15 = 450 |
| Bandwidth | 5 GB × 20 = 100 | 15 GB × 20 = 300 |
| Requests | 300,000 = 60 | 1,000,000 = 200 |
| Compute | 1 GB-hour = 10 | 3 GB-hours = 30 |
| Total credits | 290 | 980 |
| Fits on | Free, barely | Personal, barely; Pro comfortably |
The interesting line is deploys. In a normal month they're a third of usage; in launch month, when you ship fixes several times a day, they're nearly half. If you deploy often, that's the number to watch.
Good fit when
- You'd rather have a fixed price and a hard ceiling than a variable bill.
- You're two or more people: Pro's unlimited members keeps the price at $20 as the team grows.
- Your product is a front end plus server functions, with the database hosted elsewhere.
Watch out for
- Deploy-heavy weeks. Batch small fixes where you can.
- Large downloads or media. Bandwidth is the most expensive credit line; serve big files from a storage service.
- Keeping several projects on one free account: one going over pauses them all.
Vercel: generous limits, strict terms
Vercel's Hobby plan is technically generous: 100 GB of fast data transfer, a million function invocations and a million CDN requests a month, among other allowances[2]. But Vercel's terms say Hobby may only be used for personal or non-commercial purposes[4], and its documentation repeats that Hobby is restricted to non-commercial, personal use[3]. A product that takes payments is commercial. So for this guide's reader, Hobby isn't really an option once you charge.
Pro costs $20 a month per developer seat and includes $20 of usage credit, with pay-as-you-go usage beyond that, flat-rate CDN pricing with traffic spike protection, and spend management to cap your bill[2]. Past the included credit, function invocations cost $0.60 per million and active CPU starts at $0.128 per hour[7].
Good fit when
- You build with Next.js, which Vercel maintains, and want its features to work without configuration.
- You're a single developer, so the per-seat price stays at $20.
- You'd rather pay a little more than have your site pause.
Watch out for
- Per-seat pricing: two developers is $40 a month before usage.
- Usage beyond the included credit. Set a spend limit on day one; a runaway loop in a function is how small products get large bills.
Cloudflare Pages: cheapest at small scale
Cloudflare Pages hosts static files on its free plan with 500 builds a month, one build at a time, and up to 20,000 files per site[5]. Server-side code runs as Workers. The free Workers plan allows 100,000 requests a day with 10 ms of CPU time per request; the paid plan costs at least $5 a month and includes 10 million requests, then $0.30 per extra million[6].
Its limits page doesn't mention bandwidth caps or commercial-use restrictions[5]. For a small product with light server code, it's hard to beat on price.
Good fit when
- Cost is the priority and your server code is light: short API calls, form handling, simple logic.
- You're comfortable with Cloudflare's runtime, which differs from a standard Node.js server.
Watch out for
- The 10 ms CPU limit on the free plan. Image processing, PDF generation and some libraries exceed it.
- The daily request cap on free: past 100,000 a day, server-side requests fail rather than pause politely.
- Compatibility: some npm packages that assume a full Node.js environment need workarounds.
What you'd pay at 100 customers
Hypothetical example
A typical small subscription product: front end, a handful of server functions, a few GB of traffic a month, database and email hosted elsewhere.
| Setup | Monthly hosting | Notes |
|---|---|---|
| Netlify, one person | $9 | Personal; Pro at $20 if launch weeks need more credits |
| Netlify, two people | $20 | Pro, unlimited members |
| Vercel, one developer | $20 | Pro; usage usually inside the $20 credit at this size |
| Vercel, two developers | $40 | Two Pro seats |
| Cloudflare Pages + Workers | $0–$5 | Paid Workers plan once the free daily cap is a risk |
The spread between cheapest and most expensive is about $35 a month. At 100 customers paying $20, card fees are around $100 a month on their own (see card fees). Hosting is rarely the line that decides whether a small product is profitable. Picking a host you understand, with limits you know, matters more than saving $15.
Free is enough vs paid earns its place
Stay on a free plan when
- You haven't charged anyone yet
- Your plan's terms allow commercial use and you're far from its limits
- A pause for a few days wouldn't hurt anyone
A paid plan earns its place when
- You take your first payment on a plan that excludes commercial use
- Launch weeks push you near the monthly allowance
- Two or more people need to deploy
- A pause in front of paying customers would cost more than the plan
How to choose in five minutes
- Are you charging money soon? Rule out any free plan whose terms exclude commercial use.
- How many people deploy? One: per-seat pricing is fine. Two or more: flat team pricing saves money.
- What framework? Next.js and you want zero configuration: Vercel. Anything static-first: all three work.
- Which risk would you rather manage? Hard limits mean no surprise bills but possible pauses. Pay-as-you-go means no pauses but a variable bill. Set alerts either way.
- Is every dollar critical, and is your server code light? Cloudflare.
Choosing a host for the wrong reason
- Launching a paid product on a non-commercial plan because "it's only a few customers". The terms don't scale with your revenue.
- Not setting a spend limit on usage-based plans before the first real traffic.
- Serving large files from your host instead of a storage service built for it.
- Upgrading during an outage instead of a week before launch.
- Choosing on price alone. The $15 difference is smaller than one hour of debugging an unfamiliar runtime.
Questions
Can I move hosts later?
Yes. For a front end with a few server functions, moving is usually a day's work: connect the repository to the new host, move environment variables, update DNS. It's harder if you rely on a host's proprietary features, so keep those to a minimum early on.
Does my database need to be on the same platform?
No. Most small products use a hosted database from a separate provider. Keep them in nearby regions to keep queries fast.
What about Render, Railway or Fly.io?
They're good options, especially for long-running servers and background workers, which these three handle less naturally. This guide focuses on front-end-first hosting; we may cover server-first hosts separately.
Is the cheapest option good enough for a real product?
Often, yes. The question is whether its limits fit your product. Check the per-request CPU limit and compatibility with your libraries before committing.
Write down three numbers for your product: deploys per week, GB served per month, and how many people need to deploy. With those, this page's tables will tell you which plan you need. Then check the rest of your stack in Free plans, mapped.
What builders are saying
Real public threads on this topic, so you can hear it from people who hit these limits themselves. Summaries are ours; follow the link for the full discussion.
Sources
Checked 14 November 2026. Numbers in brackets in the text point to these sources.
