Payload 3 installs inside a Next.js app, so hosting Payload means hosting a Next.js app that also needs a database, file storage and a secret key. That makes the decision simpler than it looks. This guide compares the main options on those needs; for how we set Payload up for editors, see setting up Payload CMS for a marketing site, and for hosting in general, our hub on website hosting for startups.
Where should you host Payload CMS?
Host Payload CMS where your Next.js site already runs, then choose a managed database and an object store next to it. For most marketing sites that means Vercel plus managed Postgres and Blob storage, a container host such as Railway or Fly running a Node server, or Cloudflare Workers with D1 and R2.
Payload's own deployment docs say it can be deployed anywhere Next.js can, and list Vercel, Netlify, DigitalOcean and AWS among the targets. So the question isn't whether a host supports Payload. It's which host makes the database, uploads, background jobs and upgrades least painful for the people who will maintain the site.
Option | Runtime | Best for | Watch out for |
|---|---|---|---|
Vercel | Serverless functions | Least hosting work for Next.js teams | Upload size limit, jobs need a cron, seat pricing |
Railway or Fly | Long-running container | Teams that want a normal Node server | You manage the image, scaling and health checks |
Cloudflare Workers | Edge runtime with an adapter | Low cost and global caching | Platform limits, patches and more setup |
Your own VPS | Node process you run | Full control or strict data rules | Patching, backups and uptime are all yours |
What does Payload CMS hosting actually need?
Payload CMS hosting needs four things beyond a place to run Next.js: a database, persistent file storage for uploads, a long and secret PAYLOAD_SECRET value, and a way to run background jobs if you use them. Miss any one and the site works in development but breaks in production.
- A database. Payload's official adapters cover Postgres, MongoDB and SQLite. For a marketing site, managed Postgres is the common choice.
- Persistent media storage. Payload's deployment docs warn that uploads on an ephemeral filesystem can disappear when the server restarts. Use a storage adapter: Payload documents official ones for S3, Vercel Blob, Azure, Google Cloud Storage, Uploadthing and R2.
- Secrets in environment variables. PAYLOAD_SECRET and DATABASE_URL belong in the host's settings, never in the repository.
- A job runner, if you queue work. Scheduled publishing and other queued tasks need something to trigger them.
The runtime model matters most. A serverless host spins functions up and down per request, while a container host keeps one Node process alive. Payload works on both, but jobs, uploads and database connections behave differently, as the next sections show.
Can you deploy Payload CMS on Vercel?
Yes. You can deploy Payload CMS on Vercel, and it's the lowest-effort route for a Next.js team, because Payload ships Vercel-specific adapters: a Postgres adapter optimized for Vercel and a Vercel Blob storage adapter. Payload also maintains one-click Vercel templates, so a working admin is a short setup away.
Three details catch teams out. First, Payload's storage adapter docs note that server uploads on Vercel are limited to 4.5MB, and recommend turning on client uploads, which send files straight from the browser to storage, to get past it. Second, Payload's jobs docs say the autoRun option needs a long-running server and won't work on serverless platforms; instead you call Payload's jobs endpoint from Vercel Cron and protect it with a CRON_SECRET. Third, Vercel's paid plans price per seat, which we cover in Vercel vs Netlify vs Cloudflare.
We describe Vercel's behavior from Payload's docs. Vercel's limits and plans change, so check its current documentation and pricing page before you commit. Our Vercel integration page covers how we set Next.js projects up there.
When do Railway, Fly or your own server make sense?
A long-running host such as Railway, Fly or a VPS makes sense when you want Payload to behave like a normal Node server: one process that stays up, runs background jobs on a schedule and holds steady database connections. It's a good fit for sites with scheduled publishing, imports or webhooks that run longer than a serverless request.
Payload's docs include a Docker setup that builds Next.js in standalone output mode and runs it as a Node server, which is the shape these platforms expect. For jobs, the docs recommend Payload's bin script with the --cron flag on a dedicated server, running as a separate process from Next.js. Railway and Fly both run containers and offer database options alongside them; check their current docs for what's managed and what isn't.
Pros
- Background jobs and cron run without workarounds
- No serverless upload or timeout limits to design around
- Predictable runtime that matches local development
- Easy to move between providers with the same Docker image
Cons
- You own the image, health checks and scaling rules
- A single instance is a single point of failure unless you add more
- Static pages still need a CDN in front for global speed
- On a VPS, OS patches and backups are entirely your job
Your own server is the most control and the most work. Choose it when data has to stay in your company's cloud account, and make sure the accounts are in the company's name, as our website ownership checklist explains.
Where should Payload's Postgres database live?
Payload's Postgres database should live in a managed Postgres service in the same region as your app, with automatic backups and point-in-time restore turned on. Distance matters: every admin action and every uncached page query makes a round trip to the database, so a database on another continent slows the admin noticeably.
On serverless hosts, use a provider that handles connection pooling, since many short-lived functions can each open a connection. On a long-running server, a normal connection pool works. Either way, Payload's Postgres docs say schema push is for development only and shouldn't be mixed with migrations, so run migrations as part of each deploy.
Pick the region first
Put the database next to the app's primary region, not next to your office.
Turn on backups
Use the provider's automatic backups and test a restore once before launch.
Run migrations on deploy
Generate migrations when the schema changes and run them before the new version serves traffic.
Back up media separately
Database backups don't include files in object storage, so copy the bucket on its own schedule.
Is Cloudflare Workers a good host for Payload?
Cloudflare Workers is a good host for Payload if you want low running costs and edge caching and have a developer comfortable with platform quirks. It's what we run designify.design on, with a D1 database and R2 for media, but it isn't a one-click deploy.
Payload documents an R2 storage adapter, currently marked beta and aimed at Workers. In our experience the paid Workers plan is needed for the admin, image resizing has to replace sharp, and some packages need patches. The full walkthrough is in hosting Next.js on Cloudflare Workers. If nobody on the team wants to own those details, Vercel or a container host is the calmer choice.
Hypothetical example: picking a host for a 10-page SaaS site
For example, suppose a SaaS startup has a 10-page Next.js site with Payload, two marketers editing and a blog with scheduled posts. The product team already deploys to Vercel.
We'd keep the site on Vercel so there's one platform to manage, add managed Postgres in the same region, switch uploads to Vercel Blob with client uploads on, and trigger scheduled publishing with Vercel Cron.
Now suppose the same company has a strict rule that customer-facing data stays in its own AWS account. Then we'd run the same Next.js and Payload app as a Docker container in that account, with managed Postgres, S3 for media and the jobs runner as a separate process. This is a planning sketch, not a real client.
How do you choose a Payload host for your team?
Choose by who maintains the site, not by benchmark charts. The right host is the one your team can deploy, debug and pay for without thinking about it every week.
Your situation | Host to pick | Why |
|---|---|---|
Already on Vercel, few seats | Vercel | Least new tooling; Payload has Vercel adapters |
Scheduled jobs, imports or long tasks | Railway, Fly or a container host | A long-running process runs jobs natively |
Cost-sensitive, capable developer | Cloudflare Workers | Low running cost and edge caching |
Data must stay in your cloud account | Your own AWS, GCP or Azure | Full control over region and access |
Whatever you pick, keep hosting costs in context: the database, media storage and seats often cost more than compute, as our guide to website hosting cost shows. Add preview deployments so editors and reviewers can check changes before they go live. Our Payload integration page and headless CMS development service describe how we build these setups.
