SaaS hosting for a real product at 10,000 users costs roughly $190.90 a month, using current pricing for a real managed-services stack: compute, database, object storage, transactional email, and error monitoring. That number moves a lot depending on one decision in particular, and it isn't the one most cost guides focus on.

SaaS hosting: the real line items

Here's the actual stack, current pricing, verified today rather than estimated.

Two of these numbers need a stated assumption to make sense. The email figure assumes roughly 25,000 transactional emails a month for a 10k-user product, password resets, notifications, receipts, adjust up or down for how notification-heavy the actual app is. The compute tier assumes a general-purpose droplet sized for steady production traffic, not a burst-heavy workload that would need more headroom.

Bandwidth isn't in this total on purpose. Egress cost varies enormously depending on what the app actually serves (a text-heavy SaaS dashboard and a media-upload tool have wildly different bandwidth profiles), and this pillar's Egress pricing compared already has real per-GB rates across six providers. Add that on top of the $190.90 based on actual usage, not a guess.

The cheaper alternative: self-hosting the database

The single biggest line item here isn't compute, it's the managed database, and it's also the one with a real, documented cheaper alternative. Docker + Postgres on a single box covers running Postgres on the same box as the app instead of paying for a managed instance, a real setup this pillar already verified in detail, including exactly where it stops holding up.

Dropping the $60.90/month managed database line and running Postgres on the existing app server instead brings the total down to roughly $130.00/month, about a third less. The tradeoff is real, not free: backups, upgrades, and failover all become the operator's job instead of the provider's.

What that actually means in practice: someone has to verify backups are actually restorable, not just running on schedule, apply Postgres minor version upgrades, and watch disk usage before it becomes an outage instead of a warning. None of that is difficult work, but it's real, recurring work a managed database quietly absorbs into its $60.90. For a single-box setup already handling real production traffic, per that article's own findings, this is a genuinely reasonable tradeoff up to a real ceiling, not a shortcut that works at any scale.

What actually changes the number

Not every line item moves the same way as the user count grows. Transactional email is the one that scales most directly, more users generally means more password resets, more notifications, more receipts, so that line item roughly tracks user growth. Compute, storage, and monitoring are close to flat across a meaningful range, the app server doesn't need to triple in size just because the user count did, and 50,000 monitored errors a month covers a wide range of real-world traffic before an upgrade becomes necessary.

Here's the same stack recomputed at two other scales, holding compute, database, storage, and monitoring flat and only adjusting the email volume assumption:

The total moves about 49% between 2,000 and 30,000 users, a 15x growth in user count, almost entirely on the back of one line item. That's the actual shape of SaaS hosting cost at this scale: not a smooth curve upward, but one variable cost riding on top of a mostly fixed base, until real growth eventually forces an upgrade on compute or database tier too.

What this total doesn't include

This is the real infrastructure bill for the five categories above, not the complete cost of running the product. Left out on purpose, each for a different reason:

• CI/CD: GitHub Actions CI/CD for a team with no DevOps person covers this separately, GitHub's free tier covers a real amount of usage before it becomes a paid line item.
• Backup storage: DigitalOcean's managed Postgres includes backups in the $60.90 price; a self-hosted setup needs its own backup destination, typically a small object-storage cost on top of the $5/month Spaces line, not zero.
• DNS, domain, and SSL: Usually single-digit dollars a month or bundled with the domain registrar, small enough not to move the total meaningfully, but not literally $0.
• Staging or preview environments: A second, smaller instance of the same stack. Real teams often run staging on the cheapest possible tier rather than a scaled-down copy of production, which keeps this addition modest rather than doubling the bill.

None of these are hidden costs. They're just outside the scope of what "infrastructure hosting" usually means, and worth naming plainly rather than folding into one number that quietly undercounts the real total.

Frequently asked questions

How much does SaaS hosting actually cost for a small product? For a real managed-services stack at 10,000 users, roughly $190.90/month: compute, managed database, object storage, transactional email, and error monitoring. The number scales mostly with email volume, not user count directly.

Is self-hosting the database actually cheaper? Yes, by a real, checkable amount, roughly $61/month less in this stack, about a third of the total. The savings are real but come with real recurring work: backup verification, version upgrades, and disk monitoring the provider otherwise handles.

What's the biggest hidden cost in a SaaS hosting estimate like this? Bandwidth/egress, deliberately left out of this total because it varies too much by app type to put one real number on here; see this pillar's dedicated egress cost comparison for real per-GB rates instead of a guess.

Does infrastructure cost scale linearly with user count? Not in this range. The scaling table above shows the total moving about 49% between 2,000 and 30,000 users, a 15x change in user count, because compute, database, storage, and monitoring stay on the same tier well past 10,000 users. Costs eventually step up in jumps when a tier's real capacity limit is hit, not on a smooth per-user curve.

The actual verdict

The number that actually matters here isn't the total, it's which one line item you're willing to own yourself. A managed stack at $190.90/month buys back the time and risk of running a database; a self-hosted one at $130.00/month buys back about a third of the bill in exchange for real operational responsibility. Neither answer is universally right. Pick based on whether the team has the appetite to be its own on-call for the database, not based on which number looks smaller on a spreadsheet.