Learn & Understand

Cloud Pricing Models: On-Demand, Reserved, and Spot Compute Explained

In a hurry? Skip straight to the numbers.

Open the Google Cloud Pricing Calculator →

The companion calculator prices Google Cloud compute using an effective hourly rate, noting that sustained-use and committed-use discounts can shift that rate depending on how continuously a resource runs. Those discounts are part of a rich landscape of cloud pricing models, different ways of buying the same computing power at very different prices depending on your commitment and flexibility. Understanding on-demand, reserved, and spot pricing, why committing to usage earns discounts, and how providers sell spare capacity cheaply turns a cost estimate into an appreciation of the economics behind how cloud compute is priced.

On-Demand: Maximum Flexibility, Highest Price

The default way to buy cloud compute is on-demand: you pay a standard rate for resources as you use them, with no commitment, and you can start and stop them freely at any time. This offers maximum flexibility, you provision what you need when you need it and pay only for the time used, with no obligation, which is ideal for unpredictable or short-term workloads. But this flexibility comes at the highest price per unit of compute, because you are paying for the freedom to use resources whenever you want without committing to anything. On-demand is the baseline rate against which the discounted models are measured, and it is what you pay when you make no commitment and want full flexibility. Understanding on-demand pricing establishes the reference point: it is the most flexible and the most expensive way to buy compute, suited to variable or uncertain needs where the ability to scale up and down freely is worth the premium. The other pricing models all offer discounts in exchange for giving up some of this flexibility, which is the fundamental tradeoff in cloud pricing.

Reserved and Committed Use: Discounts for Commitment

Providers offer substantial discounts to customers willing to commit to using resources over a period, trading flexibility for a lower price.

Trading commitment for discount
ModelTradeoff
On-demandFull flexibility, highest price
Reserved / committedCommit to usage over time, get a discount
Sustained useAutomatic discount for running most of the period

By committing to use a certain amount of compute over a term, you receive a significantly lower rate than on-demand, because the provider gains predictable, guaranteed usage in return and rewards that commitment with a discount. Some providers apply sustained-use discounts automatically when a resource runs for a large portion of the billing period, lowering the effective rate the longer it runs, which is exactly what the calculator notes about entering an effective rate after discount. The logic is straightforward: providers value predictable, committed demand because it lets them plan capacity, so they share the benefit through lower prices for customers who commit. This is why steady, predictable workloads should use committed or reserved pricing, they run continuously anyway, so committing costs nothing extra in flexibility while saving substantially. Understanding commitment-based discounts reveals a key lever for controlling cloud cost: for workloads that run predictably, committing to that usage in advance earns a meaningful discount over on-demand rates. The effective rate the calculator uses should reflect whatever discounts apply, since the same resource can cost quite different amounts depending on the commitment made.

Spot and Preemptible: Cheap but Interruptible

The deepest discounts come from a distinctive model: buying spare capacity cheaply on the condition that it can be reclaimed at any time. Providers have unused capacity at any given moment, and rather than let it sit idle, they sell it at steep discounts as interruptible instances, sometimes called spot or preemptible, which can be taken away with little notice when the provider needs the capacity back. In exchange for accepting this interruptibility, you pay a fraction of the on-demand price. This model suits workloads that can tolerate interruption, batch processing, fault-tolerant computations, tasks that can be paused and resumed, where the risk of being interrupted is acceptable in return for the large savings. It is unsuitable for workloads that must run continuously without interruption. Understanding spot and preemptible pricing reveals how providers monetize spare capacity and how customers with flexible, interruptible workloads can access compute very cheaply: the discount is compensation for accepting that the resource can be reclaimed. This creates a kind of market for spare compute, where interruptible capacity is sold at whatever discount reflects its availability. For the right workloads, this is the cheapest way to compute, dramatically below on-demand, at the cost of guaranteed availability.

Matching the Pricing Model to the Workload

The practical art of cloud cost management is matching the pricing model to the nature of the workload, since the same compute can cost wildly different amounts depending on the model chosen. Unpredictable, short-term, or variable workloads that need flexibility suit on-demand, paying the premium for the freedom to scale as needed. Steady, predictable, always-running workloads suit reserved or committed pricing, capturing the discount for usage they would incur anyway. Interruptible, flexible workloads that can tolerate being paused suit spot or preemptible pricing, reaping the deepest discounts in exchange for accepting interruption. Many organizations use a mix, committed pricing for their baseline steady load, on-demand for variable peaks, and spot for interruptible batch work, optimizing cost across the whole portfolio. This is why the calculator asks for an effective rate: the true cost depends on which pricing model applies, and choosing the right model for each workload is one of the most powerful ways to reduce cloud spending. Understanding how to match pricing models to workloads turns cloud cost from a fixed given into an optimization: the same computing power is available at very different prices, and aligning the pricing model with the workload's flexibility and predictability captures large savings. The pricing model is a choice, not a given, and choosing well is central to cloud economics.

Pricing Compute With the Right Model

Use the calculator to estimate compute cost with an effective hourly rate, and understand the pricing models behind that rate: on-demand offers full flexibility at the highest price, reserved and committed pricing discount steady workloads in exchange for commitment, and spot or preemptible pricing sells spare capacity cheaply to interruptible workloads. Matching the model to the workload's predictability and flexibility captures large savings. The calculation uses your effective rate; understanding cloud pricing models is what lets you choose the model that makes the same compute cost far less.

Ready to Put This Into Practice?

Now that you understand how it works, plug in your own numbers and get an instant, accurate result.

Use the Google Cloud Pricing Calculator Now →