Hourly vs Monthly Server Billing: Which Model Saves You More?

Hourly billing charges you for the hours a server exists. Monthly billing charges you for the calendar month whether it exists or not. The saving is entirely a function of duty cycle: below roughly 70% uptime, hourly wins; above it, the two converge — and on a plan with a monthly cap, hourly can never lose. This guide works through the arithmetic with real prices instead of hypotheticals.

How each model actually bills

Monthly billing is a subscription. You commit to a period, usually with a term discount for committing to more of them, and you pay the full amount whether the machine runs for thirty days or three.

Hourly billing meters existence, not usage. This is the point people get wrong most often: a stopped instance is usually still billed, because the disk, the IP and the reserved capacity all still exist. Only destroying the instance stops the meter. Check any provider's policy on stopped instances before you rely on "just shut it down at night" as a cost strategy.

Capped hourly is the hybrid, and it is what we run. Every plan has an hourly rate and a monthly ceiling; you pay the hourly rate until the total reaches the monthly price, then it stops. The €4/month Ion instance bills at €0.0056/hr; run it for a full 730-hour month and the raw arithmetic gives €4.09, so you are charged €4.00. There is no scenario in which the hourly rate costs you more than the monthly one.

The prices, so the maths is checkable

Plan	Specs	Hourly	Monthly cap	Break-even
Ion	2 GB RAM · 1 vCPU · 40 GB SSD	€0.0056	€4.00	~714 h (98%)
Pulse	4 GB RAM · 2 vCPU · 80 GB SSD	€0.0083	€6.00	~723 h (99%)
Core	8 GB RAM · 2 vCPU · 120 GB SSD	€0.0139	€10.00	~719 h (98%)
Forge	16 GB RAM · 4 vCPU · 160 GB SSD	€0.0222	€16.00	~721 h (99%)
Apex	24 GB RAM · 6 vCPU · 200 GB SSD	€0.0417	€30.00	~719 h (98%)
Titan	32 GB RAM · 8 vCPU · 300 GB SSD	€0.0806	€58.00	~719 h (98%)

A 30-day month is 720 hours; a 31-day month is 744. Because every break-even point sits at or just below a full month, the cap does the work: you are effectively on a monthly plan when you run continuously, and on a pay-per-hour plan whenever you do not.

Four scenarios, costed

1. Always-on production server

A web application that runs every hour of every day on a Core instance. 730 hours × €0.0139 = €10.15, capped at €10.00. Monthly equivalent: €10.00. No difference. Hourly billing costs an always-on workload nothing, which is the point of the cap — you are not paying a premium for optionality you never use.

2. Weekday-only development environment

A Forge instance for a development team, up from 08:00 to 20:00 on weekdays only. That is 12 hours × 22 working days ≈ 264 hours. 264 × €0.0222 = €5.86 against €16.00 monthly. A 63% saving, and no term commitment.

3. CI runners that only exist during builds

Four Apex instances spun up per pipeline run, alive for 25 minutes, forty runs a month. 4 × 0.42 h × 40 = 67 hours. 67 × €0.0417 = €2.79. Four monthly Apex instances would be €120. This is the case where hourly billing stops being a discount and becomes a different way of building infrastructure — see our DevOps and CI/CD page for the pattern.

4. Seasonal capacity

A retailer adds six Titan instances for eleven days over a sales peak. 6 × 264 h × €0.0806 = €127.67 against €348 for six monthly instances. The saving is real, but the bigger benefit is that the capacity did not need to be planned a quarter ahead.

Where hourly billing does not save you money

Being straight about this, because the model is frequently oversold:

• Always-on workloads. Scenario 1 above. The cap means you break even, not ahead. Anyone claiming hourly billing saves money on a server that never stops is selling you something.
• Long-term commitments elsewhere. Providers that offer 20–40% discounts for annual prepayment will beat any hourly rate for a workload you are certain will run for a year. If your load is genuinely fixed and known, a term discount is the cheaper instrument. We do not discount for term, and that is a real trade-off in the other direction.
• Workloads you forget to destroy. The single largest source of waste in hourly billing is instances nobody turned off. A stopped instance still occupies disk and an IP. Tag everything, and review what is running weekly.
• Storage that outlives compute. If you destroy an instance but keep a snapshot or volume, the storage keeps billing. Compute elasticity does not make storage free.

The part that is not about money

Cost is the easy comparison. The more consequential difference is what each model does to how you work.

Monthly billing creates a psychological floor. Because the server is already paid for, it stays up "just in case", accumulates unrelated services, and becomes the pet nobody wants to rebuild. Capacity planning becomes a quarterly negotiation.

Hourly billing with sub-minute provisioning removes the reason to keep anything. If a fresh environment costs eleven cents and appears in fifty seconds, you stop nursing broken ones. Staging environments become disposable. Testing a change on the exact production plan size becomes a normal thing to do rather than a budget request. That behavioural shift is usually worth more than the line-item saving.

What to look for in any hourly billing policy

Providers implement hourly billing very differently. Five questions worth asking before you rely on it:

1. Is there a monthly cap? Without one, an always-on instance costs more than the equivalent monthly plan, sometimes considerably more.
2. Are stopped instances billed? Almost always yes for the disk and IP. Confirm rather than assume.
3. Is there a minimum billing period? Some providers round every instance up to an hour, or to a day, which destroys the economics of short-lived CI runners.
4. Are bandwidth and IPs metered separately? A cheap hourly rate with metered egress can be far more expensive than it looks. Our plans are unmetered with no overage billing, so the hourly rate is the whole price.
5. Is there a minimum term or setup fee? Either one turns hourly billing back into a monthly commitment with extra steps.

Which model should you choose?

Your situation	Better model
Production service running continuously	Either — capped hourly is identical to monthly
Development or staging, business hours only	Hourly, by a wide margin
CI runners and build agents	Hourly, by an order of magnitude
Seasonal or event-driven capacity	Hourly
Load you cannot yet predict	Hourly, until you can
Fixed load, certain for 12+ months, provider offers term discounts	Annual prepay elsewhere may be cheaper
Sustained heavy load at high duty cycle	Compare against bare metal — see VPS vs dedicated

Frequently asked questions

Does an hourly VPS cost more if I run it all month?

Not on a capped plan. Billing stops at the monthly price — €4.00 for Ion, €58.00 for Titan — regardless of how many hours are in the month. On uncapped hourly plans elsewhere, yes, it usually does.

Am I charged if I shut the server down but do not delete it?

A stopped instance still holds its disk, its IP address and its place on a host, so it continues to bill. Destroy the instance to stop the meter, and keep a snapshot if you need the data.

Can I switch between hourly and monthly?

On our plans there is nothing to switch — every plan is hourly with a monthly ceiling, so you get whichever is cheaper without choosing in advance.

How is a partial hour billed?

Ask any provider this directly. Rounding a 3-minute instance up to a full hour changes the CI-runner arithmetic in scenario 3 above by more than an order of magnitude.

Does hourly billing apply to dedicated servers?

No. Physical machines have provisioning and hardware costs that do not amortise over hours, so our dedicated range is monthly, from €209.

Work out your own number

Multiply your expected hours per month by the hourly rate, compare it with the cap, and take the lower one. The VPS cost calculator does it across plans, and our guide to hourly VPS hosting covers the deployment side of the model. If you are still sizing the instance itself, start with choosing the right VPS plan.