How Much Does It Cost to Host a Web Application?

There is no single monthly price for hosting a web application. The bill is the sum of separate line items, each metered in its own way: compute, data transfer (egress), the database, file storage, backups, email and SMS, monitoring and logs, extra environments, team seats and paid add-ons. Which lines you pay for, and where each free allowance ends, depends mostly on the architecture: a static site on a CDN, a managed platform (PaaS), serverless functions or containers on a large cloud, or a virtual private server (VPS).

Published entry prices are low. On 15 October 2026 Vercel listed its Pro plan as a $20/month platform fee (USD, excluding tax) and DigitalOcean listed its smallest Basic Droplet at $4/month. Those are list prices for a starting plan, not what a production application usually pays in total: the database, extra seats, data transfer beyond the allowance and add-ons sit on top. The reliable way to estimate is to list the lines, attach your own usage to each, and price them in the vendor's calculator for your region.

This guide covers the running bill once the application is live. To estimate the build itself, see how to estimate the cost of a custom web application, whose running-costs step points here.

Read across a row to see how the same line is metered in each architecture. Each cell says what the line is billed by and, where a vendor publishes one, where the included allowance ends. Named vendors are examples to check, not recommendations. All figures are vendor list prices or allowances checked on 15 October 2026 and will change.

The table below records what each vendor's own page said on 15 October 2026. Prices are list prices in US dollars as displayed (Vercel and Cloudflare state USD; check the currency shown for your account), excluding tax unless stated. They are starting points for your own model, not typical totals, and vendors change them often.

Environments. A serious application has production, at least one staging environment, and often a preview build for every change. On a PaaS the previews are usually built in, but persistent staging is another project, and some add-ons are priced per project. On a cloud provider or VPS, a realistic staging environment is a second copy of the stack, and the database is the part people forget to copy. Decide which environments run all month and which can be paused or scaled down.

Egress. Data leaving the platform for the internet is metered on most providers once the allowance is used. Images, video, file exports and chatty APIs drive it. AWS includes 100 GB a month from its Regions across services and 1 TB a month from CloudFront (AWS network FAQs, vendor). Serving large files through a CDN or object storage designed for it usually changes this line more than any other choice.

Managed databases. A managed database is normally billed as an always-on instance plus storage, and higher tiers add replicas, more storage and longer backup retention. It is often the line that grows first once real data arrives. Free database tiers are for trying things: Render's free Postgres expires after 30 days and has no backups (Render docs, vendor).

Backups and restore. Check how long backups are kept, whether point-in-time restore is included, where the copies are stored and what restoring costs in time. A backup you have never restored is an assumption, not a plan. The website maintenance guide covers how to schedule and test restores alongside updates.

Observability. Error tracking, uptime checks, logs and metrics each have their own meter, usually events, GB ingested or retention days. Short retention is cheap but limits investigation after an incident. Decide how many days of logs you need, and keep debug-level logging out of production.

Hypothetical example: a business-to-business booking application, built with Next.js and a PostgreSQL database. The table records usage assumptions only. It deliberately gives no totals, because the result depends on the provider, region and plan you pick. Replace each assumption with your own.

On Vercel Pro, as listed on 15 October 2026, Stage 2 adds two deploying seats at $20/month each and password protection at $20 per protected project per month. Stage 3 adds three more deploying seats; the contractor can use a free viewer seat. The 1.4 TB of transfer goes past the 1 TB included, so the extra 0.4 TB is billed per GB at the listed rate, from $0.15/GB. SAML SSO is a $300/month add-on (Vercel Pro plan docs, vendor).

On AWS, the 100 GB monthly data-transfer allowance is used up during Stage 2, and email is billed per 1,000 messages on SES. On a VPS, the instance price stays flat until the server needs resizing, but the database, backups and monitoring become either separate products or work for your team. In every architecture the database line moves at each stage, which is why it deserves its own estimate rather than a share of "hosting".

Notice what does not change much: the base platform fee. The lines that grow are seats, transfer, the database and add-ons that customers ask for.

Region choice changes latency, which services are available and the price. The same service can cost a different amount in a different region, so price your model in the region you will use, in the provider's own calculator. The table lists regions published by each provider on 15 October 2026 for the markets most of our readers ask about.

Static or Jamstack suits marketing sites and apps whose dynamic parts live in a separate API or backend. It is cheap to serve, but it is not suitable when most pages depend on signed-in data or server rendering. Next.js lists static export as "Limited" feature support, while a Node.js server or Docker container supports all features (Next.js deploying docs).

A managed PaaS suits small teams that want deployments, previews and scaling handled for them. The trade-off is per-seat and per-add-on pricing that grows with the team and with enterprise customer demands. It is a poor fit when you need fine control over networking or must keep everything inside one cloud account for contractual reasons.

Cloud serverless or containers suit teams that need specific regions, managed databases and services from one provider, or traffic that varies widely. Every service has its own meter, so the bill needs someone who reads it. It is not suitable if nobody on the team can own cloud configuration and cost alerts.

A VPS suits steady, predictable workloads and teams willing to run servers. The instance price is flat, but patching, backups, monitoring and scaling are your job. It is a poor fit when uptime matters and nobody is on call.

For the technical trade-offs beyond cost, see Kubernetes vs serverless vs virtual machines.

If you are already live and the bill is rising, the work changes from estimating to finding where the money goes: tagging resources, right-sizing instances, pausing idle environments and committing to capacity where usage is steady. That is covered in cloud cost optimisation; its methods apply outside Australia. Hosting changes should also go into the maintenance plan, so that someone owns updates, restores and cost reviews every month.

Managed platforms (vendor pages): Vercel pricing; Vercel Pro plan docs (last updated 15 September 2026); Netlify pricing; Railway pricing; Render: deploy for free; Cloudflare Workers pricing.

VPS and databases (vendor pages): DigitalOcean Droplet pricing; DigitalOcean Managed Database pricing.

Cloud providers (vendor pages and docs): AWS Free Tier; AWS global network FAQs; AWS Lambda pricing; Amazon SES pricing; AWS Regions; Google Cloud free features (last updated 13 October 2026); Google Cloud regions and zones; Azure free account; List of Azure regions (updated 13 October 2026).

Framework: Next.js: deploying.

Accessed 15 October 2026. All prices are vendor list prices and allowances that change; check the vendor's page and calculator for your region before you commit.

Write down your expected users, data size, monthly data out, emails, environments and region. Put each against the bill-builder rows above, price it on the vendor's page for your region, and record the date. Then compare two architectures on the same assumptions, including the team time a VPS or raw cloud setup needs. If you are choosing a stack as part of a new build, our web application development work starts from these running-cost lines.