VM instance pricing

This page describes the cost of running a Compute Engine VM instance with any of the following machine types, as well as other VM instance-related pricing. To see the pricing for other Google Cloud products, see the Google Cloud pricing list.

Compute Engine charges for usage based on the following price sheet. A bill is sent out at the end of each billing cycle, providing a sum of Google Cloud charges. Prices on this page are listed in U.S. dollars (USD).

For Compute Engine, disk size, machine type memory, and network usage are calculated in JEDEC binary gigabytes (GB), or IEC gibibytes (GiB), where 1 GiB is 2³⁰ bytes. Similarly, 1 TiB is 2⁴⁰ bytes, or 1024 JEDEC GBs.

If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply.

You can also find pricing information with the following options:

See the estimated costs of your instances and Compute Engine resources when you create them in the Google Cloud console.
Estimate your total project costs with the Google Cloud Pricing Calculator.
View and download prices from the Pricing Table in the Google Cloud console.
View more information about costs and usage in Cloud Billing reports.
Use the Cloud Billing Catalog API for programmatic access to SKU information.

Billing model

The following billing model applies to all vCPUs, GPUs, and memory resources. The billing model also applies to several premium images that you run on Compute Engine instances.

All vCPUs, GPUs, and GB of memory are charged a minimum of 1 minute. For example, if you run your virtual machine for 30 seconds, you will be billed for 1 minute of usage.
After 1 minute, instances are charged in 1 second increments.

Instance uptime

Instance uptime is measured as the number of seconds between when you start an instance and when you stop an instance, the latter being when the instance state is TERMINATED. In some cases, your instance can suffer from a failure and be marked as TERMINATED by the system; in these cases, you will not be charged for usage after the instance reaches the TERMINATED state. If an instance is idle, but still has a state of RUNNING, it will be charged for instance uptime. The easiest way to determine the status of an instance is to use gcloud compute with the gcloud compute instances list command or to visit the Google Cloud console.

In the case of reservations, instance uptime is measured as the number of seconds between when you create a reservation and when you delete that reservation. Reserved resources are billed at standard rates, whether they are started or not.

Note that Compute Engine bills for a minimum of 1 minute of usage, so if you use an instance for 30 seconds of uptime, you are billed for 1 minute. After 1 minute, your instance is billed on a per-second basis. For more information, see the billing model.

Resource-based pricing

Each vCPU and each GB of memory on Compute Engine is billed separately rather than as part of a single machine type. You still create instances using predefined machine types, but your bill reports them as individual vCPUs and memory used per hour. If you change the number of threads per core, you are billed for the number of vCPUs defined by a VM's machine type, not the number of threads used by the VM.

The pricing tables in the machine family and machine type sections on this page describe prices for machine types based on vCPU and memory resources, but also include the calculated cost for each machine type. You can also use the Google Cloud Pricing Calculator to better understand prices for different configurations.

Discounts

vCPU and memory usage for each machine type use the on-demand price unless that usage qualifies for a discount. vCPU and memory usage for each machine type can receive one of the following discounts:

Spot prices: automatic discounts for all Spot VMs (and preemptible VMs), which are between 60-91% off for most machine types and GPUs and also provide smaller discounts for some other resources. Spot prices change over time, but the relevant discounts are always in the 60-91% range.

Note: Spot prices can change up to once every 30 days and don't appear in most pricing tables for Compute Engine. For the latest prices, see the Spot VMs pricing page.
Committed use discounts (CUDs): up to a 70% discount for memory-optimized machine types and up to a 55% discount for all other machine types.
Sustained use discounts (SUDs): automatic discount of up to 30% on resources that are used for more than 25% of a month and are not receiving any other discounts.

Discount types cannot be combined. For more information, see Order of discount application.

Spot prices

Spot prices provide 60-91% discounts for most machine types and GPUs as well as smaller discounts for some other resources used by the VM. Spot prices are automatically applied to all Spot VMs and preemptible VMs. However, Compute Engine can preempt Spot VMs and preemptible VMs at any time, so they are only recommended for fault-tolerant applications that can handle VM preemption.

Spot prices can change up to once every 30 days and don't appear in most pricing tables for Compute Engine. For the latest prices, see the Spot VMs pricing page.

For more information, see the documentation for Spot VMs.

Committed use discounts

Compute Engine offers resources at deeply discounted prices in return for purchasing committed use contracts (also known as commitments). When you purchase a commitment, you commit either to a minimum amount of resource usage or to a minimum spend amount for a specified term of one or three years.

Depending on your resource usage requirements, you can purchase 1-year or 3-year commitments and receive CUDs for Compute Engine resources in either of the following ways:

Resource-based committed use discounts: You receive these CUDs when you purchase a resource-based commitment and commit to use a minimum level of Compute Engine resources in a particular region.
Compute flexible committed use discounts: You avail Compute flexible CUDs when you purchase a spend-based Compute flexible commitment and commit to a minimum amount of hourly spend.

Resource-based CUDs

Compute Engine offers a flat committed use discount percentage on its VMs across all regions. When you purchase vCPUs, memory, or both on a 1-year commitment, you get the resources at a discount of 37% over the on-demand prices. When you purchase your resources on a 3-year commitment, the discount increases to 70% over the on-demand prices for memory-optimized machine types and to 55% over the on-demand prices for all other machine types.

Resource-based CUDs are available for the following machine types in each machine family, with the exception of Memory Optimized Upgrade Premium SKUs:

General purpose:
- All general purpose machine types except for N1 shared-core machine types
- All N1, N2, N2D, N4, C3, C3D, C4, and C4A sole-tenant nodes
Memory-optimized:
- All M1, M2, M3 machine types
- All M1, M2, M3 sole-tenant node types
Compute-optimized:
- All compute-optimized (C2, C2D, and H3) machine types
- All C2 and H3 sole-tenant node types
Storage-optimized:
- All Z3 machine types
Accelerator-optimized:
- All A2, A3, and G2 machine types
- All G2 sole-tenant node types

For a more detailed breakdown, see Commitment types.

Combine reservations with committed use discounts

A commitment provides a 1- or 3-year discounted price agreement, but it does not reserve capacity in a specific zone. A reservation ensures that capacity is held in a specific zone even if the reserved VMs are not running. To get zonal resources at discounted prices and also ensure that capacity is reserved for them, you must do both of the following: purchase commitments and create reservations for those zonal resources.

You can also attach reservations to your resource-based commitments to ensure that Compute Engine reserves capacity for your committed resources. When you commit to GPU or Local SSD resources, you must also reserve those resources and attach those reservations to your commitment.

For more information, see Combine reservations with committed use discounts.

Purchase resource-based commitments

To learn how to purchase resource-based commitments, see the following:

Pricing for resource-based commitments

After you purchase a resource-based commitment, you're billed monthly for your commitment and must pay your monthly commitment fee even if you don't use all of your committed resources. Your commitment fee is the sum of the discounted prices of all your committed resources. Compute Engine calculates the discounted price of each resource by using its prevailing on-demand price on the day your commitment becomes active. Your monthly commitment fee and the discounted prices for your resources stay the same until the end of your commitment term, even if the on-demand prices change.

If you use your commitments to run custom machine types, then Compute Engine charges a 5% premium over the commitment prices. Compute Engine charges this premium for the portion and duration of your commitment that you run these custom machine type VMs.

Compute flexible CUDs

Compute flexible CUDs are spend-based CUDs that you receive when you commit to a minimum amount of hourly spend on one or more of the following services:

Compute Engine
Google Kubernetes Engine
Cloud Run

For Compute Engine, you can receive flexible CUDs for your vCPU and memory usage in any of the projects within your Cloud Billing account, across any region, and belonging to any eligible machine types.

You purchase flexible commitments for your Cloud Billing account and commit to a minimum hourly spend amount across these products for a 1-year or 3-year term duration. Specifically, you commit to spend on eligible resources or services that are worth a specified minimum amount of on-demand price, per hour, throughout the commitment's term. Depending on your commitment's term, you receive the following CUDs on that minimum hourly spend amount:

A 28% discount over your committed hourly spend amount for a 1-year commitment
A 46% discount over your committed hourly spend amount for a 3-year commitment

Your commitment becomes active within the first hour of its purchase. This discounted committed spend amount becomes your hourly commitment fee. In return, you receive hourly credits on your Cloud Billing account that are worth your total committed spend amount. Google Cloud uses these credits to offset your hourly spend on usage that is eligible for flexible CUDs. At the end of each month, Google Cloud calculates your total commitment fee for that month and bills you that amount.

Your hourly commitment fee remains your minimum hourly expenditure throughout the commitment term and you have to pay it even if you don't use resources whose on-demand prices total up to your committed hourly spend. Your commitment fee remains the same even if the on-demand prices for your resources change during your commitment term.

For Compute Engine, only memory and vCPUs of the following machine series are eligible for flexible CUDs. For every listed machine series, all available machine types and sole-tenant node types are eligible.

General purpose: C3, C3D, C4, C4A, E2, N1, N2, N2D, and N4 machine series
Compute-optimized: C2 and C2D machine series
Storage-optimized: Z3 machine series

For more information, see the following documents:

To learn more about flexible CUDs, see Compute flexible CUDs.
To learn how to purchase spend-based commitments, see Purchasing spend-based commitments.
To view the full list of Compute Engine SKUs that are eligible for flexible CUDs, see Compute flexible CUDs eligible SKUs.
To learn how to view, analyze, and manage your spend-based commitments, see Analyze the effectiveness of flexible commitments.

Sustained use discounts

Compute Engine offers sustained use discounts (SUDs) on resources that are used for more than 25% of a billing month and are not receiving any other discounts. Whenever you use an applicable resource for more than a fourth of a billing month, you automatically receive a discount for every incremental hour that you continue to use that resource. The discount increases incrementally with usage and you can get up to a 30% net discount off of the resource cost for virtual machine (VM) instances that run the entire month.

Compute Engine automatically calculates and applies SUDs to resource usage within a Cloud Billing account, so there is no action required on your part to enable these discounts.

You receive SUDs in the form of monthly credits. At the end of every month, for every eligible Compute Engine resource that you use, you receive applicable SUDs as credits based on the duration of time for which you used that resource. Any credits that you receive for your SUDs have no cash value. Compute Engine uses these credits to offset your monthly resource usage costs. You can't store or use these credits beyond the month in which you receive them.

To learn more about sustained use discounts, see the SUDs documentation.

You can view all your earned credits during a particular month in the cost table report for your Cloud Billing account on the Google Cloud console. To learn more about credits and how to view them, see View and analyze your credits.

Order of discount application

You can purchase both resource-based and flexible commitments to cover Compute Engine resources for projects within your Cloud Billing account. However, the discount types that Google Cloud offers on your Compute Engine resources are mutually exclusive and can't be combined. At any given point, a resource is eligible for only one kind of discount. If you're receiving a specific type of discount for a portion of your resource usage, then that portion of usage doesn't qualify for any other type of discount.

You can use your resource-based commitments to cover your predictable and stable resource usage that is specific to a project, region, and a machine series. You can use flexible commitments to cover any resource usage that isn't specific to any one machine series, project, or region. If you purchase both resource-based commitments and flexible commitments for your Compute Engine resources, then Google Cloud optimizes the use of your commitments, on an hourly basis, in the following way:

Google Cloud first utilizes your resource-based commitments and applies all the resulting resource-based CUDs to any eligible hourly usage.
Google Cloud then utilizes your spend-based commitments and applies all the resulting flexible CUDs to any remaining eligible on-demand usage.

After utilizing all of your commitments, Google Cloud uses the on-demand rates to charge any additional hourly usage. This overage hourly usage might be eligible for any applicable SUDs.

Additionally, spot prices apply to all Spot VMs (and preemptible VMs), so Spot VMs (and preemptible VMs) cannot receive CUDs or SUDs.

The pricing tables in the machine family and machine type sections on this page compare the on-demand prices for machine types in each machine family with the discounted prices for each of the following:

Spot VMs (and preemptible VMs)
1-year resource-based commitments
3-year resource-based commitments
1-year flexible CUD consumption rates
3-year flexible CUD consumption rates

Viewing usage

The Google Cloud console provides a transaction history for each of your projects. This history describes your current balance and estimated resource usage for that particular project.

To view a project's transaction history, go to the estimated billing invoice page.

General-purpose machine type family

General-purpose machine-types offer predefined and custom machine types in each region. Predefined machine types have a preset number of vCPUs and amount of memory, but are billed using the resource-based pricing model. Custom machine types are billed according to the resource-based pricing.

Not all machine types are available in all zones all the time. To ensure that a machine type is available when you need it, you can preemptively reserve the machine type in a certain zone. For information about reserving predefined machine types in a specific zone, see Reservations of Compute Engine zonal resources.

C4 machine types

C4 standard machine types

C4 high-memory machine types

C4 high-CPU machine types

C4A machine types

If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply.
Spot prices are dynamic and can change up to once every 30 days, but provide discounts of 60-91% off of the corresponding on-demand price for most machine types and GPUs. Spot prices also provide smaller discounts for local SSDs and A3 machine types. For more information, see the Spot VMs documentation.

C4A standard machine types

C4A high-memory machine types

C4A high-CPU machine types

N4 machine types

N4 standard machine types

The following table shows the calculated costs for standard predefined machine types in the N4 machine family. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

Standard machine types have 4 GB of memory per vCPU.

N4 high-memory machine types

The following table shows the calculated cost for the N4 high-memory predefined machine types. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

High-memory machine types have 8 GB of memory per vCPU. High-memory instances are ideal for tasks that require more memory relative to virtual CPUs.

N4 high-CPU machine types

The following table shows the calculated cost for N4 high-CPU predefined machine types. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

High-CPU machine types have one vCPU for every 1 GB of memory. High-CPU machine types are ideal for tasks that require moderate memory configurations for the needed vCPU count.

N4 custom vCPUs and memory

Custom machine types let you set a specific number of vCPUs and GB of memory for your instances to match the needs of your workload. Custom machine types save you the cost of running on a larger and more expensive machine type if your application does not require all of the resources provided by that machine type.

Read Creating a VM instance with a custom machine type to learn how to use these machine types.

The on-demand prices for custom machine types include a 5% premium over the on-demand prices for standard machine types. For an accurate estimate of your billing with custom machine types, use the Google Cloud Pricing Calculator.

If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply.
Spot prices, which typically offer the largest discounts—up to 91% off of the corresponding on-demand price—are listed separately on the Spot VMs pricing page.
The CUDs prices for Custom Machine Types reflect a 5% premium over predefined shapes. The premium applies only for the duration and amount of CMTs used and covered by the CUDs. More details please refer to the documentation for Resource-based committed use discounts.

C3 machine types

C3 standard machine types

C3 high-memory machine types

C3 high-CPU machine types

C3D machine types

C3D standard machine types

C3D high-memory machine types

C3D high-CPU machine types

For N1, N2, and C2 machine types, Compute Engine provides automatic sustained use discounts for all of the predefined vCPU and memory resources that you use in a region. Sustained use discounts for predefined machine types are calculated separately from custom, memory-optimized, compute-optimized, and shared-core machine types. Depending on the machine type, sustained use discounts differ between N1 and N2 machine types:

N1 machine types can receive a sustained use discount up to 30%.
N2 machine types can receive a sustained use discount up to 20%.

E2 machine types do not offer sustained use discounts but provide larger savings directly through the on-demand and committed-use prices. E2 machine types provide consistently predictable pricing without the requirement to run a VM for a specific portion of the month.

For more information, see Sustained use discounts.

E2 machine types

E2 machine types are also available as shared-core VMs. For pricing information about E2 shared-core machine types, see E2 shared-core machine types.

E2 standard machine types

The following table shows the calculated cost for standard predefined machine types in the E2 machine family. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

Standard machine types have 4 GB of memory per vCPU.

E2 high-memory machine types

The following table shows the calculated cost for the E2 high-memory predefined machine types. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

High-memory machine types have 8 GB of memory per vCPU. High-memory instances are ideal for tasks that require more memory relative to virtual CPUs.

E2 high-CPU machine types

The following table shows the calculated cost for E2 high-CPU predefined machine types. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

High-CPU machine types have one vCPU for every 1 GB of memory. High-CPU machine types are ideal for tasks that require moderate memory configurations for the needed vCPU count.

E2 custom vCPUs and memory

Read Creating a VM instance with a custom machine type to learn how to use these machine types.

Not all machine types are guaranteed to be available in all zones all the time. To ensure that a machine type is available when you need it, you can preemptively reserve the machine type in a certain zone. For information about reserving predefined machine types in a specific zone, see Reservations of Compute Engine zonal resources.

N2 machine types

N2 standard machine types

The following table shows the calculated costs for standard predefined machine types in the N2 machine family. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

Standard machine types have 4 GB of memory per vCPU.

N2 high-memory machine types

The following table shows the calculated cost for the N2 high-memory predefined machine types. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

High-memory machine types have 8 GB of memory per vCPU. High-memory instances are ideal for tasks that require more memory relative to virtual CPUs.

N2 high-CPU machine types

The following table shows the calculated cost for N2 high-CPU predefined machine types. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

High-CPU machine types have one vCPU for every 1 GB of memory. High-CPU machine types are ideal for tasks that require moderate memory configurations for the needed vCPU count.

N2 custom vCPUs and memory

Read Creating a VM instance with a custom machine type to learn how to use these machine types.

Sustained use discounts for custom machine types are calculated separately from predefined machine types, memory-optimized types, and shared-core machine types.

N2 extended custom memory

For custom machine types, any memory up to and including 8 GB of memory per vCPU is charged at the standard custom vCPU and memory pricing rate. Any memory above 8 GB per vCPU is charged according to the following extended memory prices. To learn how to create instances with custom machine types and extended memory, see Adding extended memory to a machine type.

N2D machine types

N2D standard machine types

The following table shows the calculated costs for standard predefined machine types in the N2D machine family. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

Standard machine types have 4 GB of memory per vCPU.

N2D high-memory machine types

The following table shows the calculated cost for the N2D high-memory predefined machine types. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

High-memory machine types have 8 GB of memory per vCPU. High-memory instances are ideal for tasks that require more memory relative to virtual CPUs.

N2D high-CPU machine types

The following table shows the calculated cost for the N2D high-cpu predefined machine types. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

High-cpu machine types have 1 GB of memory per vCPU. High-cpu instances are ideal for tasks that require more virtual CPUs relative to memory.

N2D custom vCPUs and memory

Read Creating a VM instance with a custom machine type to learn how to use these machine types.

Sustained use discounts for custom machine types are calculated separately from predefined machine types, memory-optimized types, and shared-core machine types.

N2D extended custom memory

Tau T2D machine types

If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply.

Tau T2D standard machine types

The following table shows the calculated costs for standard predefined machine types in the Tau T2D machine family. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type

Standard machine types have 4 GB of memory per vCPU.

If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply.

Tau T2A machine types

If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply.

Tau T2A standard machine types

The following table shows the calculated costs for standard predefined machine types in the Tau T2A machine family. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

Standard machine types have 4 GB of memory per vCPU.

If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply.

N1 machine types

N1 standard machine types

The following table shows the calculated cost for standard predefined machine types in the N1 machine family. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

Standard machine types have 3.75 GB of memory per vCPU.

N1 high-memory machine types

The following table shows the calculated cost for the N1 high-memory predefined machine types. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

High-memory machine types have 6.5 GB of memory per vCPU. High-memory instances are ideal for tasks that require more memory relative to virtual CPUs.

N1 high-CPU machine types

The following table shows the calculated cost for N1 high-CPU predefined machine types. The vCPUs and memory from each of these machine types are billed by their individual predefined vCPU and memory prices, but these tables provide the cost that you can expect using a specific machine type.

High-CPU machine types have one vCPU for every 0.90 GB of memory. High-CPU machine types are ideal for tasks that require moderate memory configurations for the needed vCPU count.

N1 custom vCPUs and memory

For example, instead of using an n1-standard-8 machine type when you need a machine type with 6 vCPUs, you can create an instance with a custom machine type that has 6 vCPUs and 22.5 GB of memory. Creating a custom machine type can save you up to 40% compared to selecting a larger machine type. Custom machine types are billed according to the number of vCPUs and the amount of memory used.

Read the Creating instances with custom machine types to learn how to use these machine types.

Sustained use discounts for custom machine types are calculated separately from predefined machine types, memory-optimized types, and shared-core machine types.

N1 extended custom memory

For custom machine types, any memory up to and including 6.5 GB of memory per vCPU is charged at the standard custom vCPU and memory pricing rate. Any memory above the 6.5 GB per vCPU is charged according to the extended memory prices that are described in detail below. See the Extended Memory page to learn how to create instances with custom machine types and extended memory.

Compute-optimized machine type family

Compute-optimized machine types are ideal for compute-intensive workloads. These machine types offer the highest performance per core on Compute Engine.

H3 machine types

H3 VMs are powered by the 4th generation Intel Xeon Scalable processors (code-named Sapphire Rapids), DDR5 memory, and Google's custom Intel Infrastructure Processing Engine (IPU). The following table describes the pricing per vCPU and GB of memory for H3 machine types.

H3 Standard machine types

The following table shows the calculated cost for h3-standard-88 machine types, which is the H3 predefined machine type. The vCPUs and memory from each of these machine types are billed by their individual compute-optimized vCPUs and memory prices but these tables provide the cost that you can expect using a specific machine type.

C2 machine types

C2 machine types offer Intel Scalable Processors (Cascade Lake) and up to 3.8Ghz sustained all-core-turbo. Currently, C2 machine types are only available in certain regions and zones. The following table describes the pricing per vCPU and GB of memory for C2 machine types.

C2 standard machine types

The following table shows the calculated cost for c2-standard machine types, which are C2 predefined machine types. The vCPUs and memory from each of these machine types are billed by their individual compute-optimized vCPUs and memory prices but these tables provide the cost that you can expect using a specific machine type.

Spot prices, which typically offer the largest discounts—up to 91% off of the corresponding on-demand price—are listed separately on the Spot VMs pricing page.

C2D vCPUs and memory

C2D Standard machine types

C2D Highmem machine types

C2D Highcpu machine types

Memory-optimized machine type family

Memory-optimized machine types are ideal for tasks that require intensive use of memory with higher memory to vCPU ratios than the general-purpose n2-highmem machine types. Memory-optimized machine types are available in certain regions only. To learn more about memory-optimized machine types, see Memory-optimized machine family.

M3 machine types

M3 machine types offer 3rd Generation Intel Scalable Processors (Ice Lake). Currently, M3 machine types are only available in certain regions and zones. The following table describes the pricing per vCPU and GB of memory for M3 machine types.

The following table shows the calculated cost for m3-megamem and m3-ultramem machine types. The vCPUs and memory from each of these machine types are billed by their individual M3 vCPUs and memory prices but these tables provide the cost that you can expect using a specific machine type.

These machine types are only available in select zones.

Committed Use Discounts apply to memory-optimized machine types only if you buy the commitment type specifically for M3 machine types.
If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply. Spot prices, which typically offer the largest discounts—up to 91% off of the corresponding on-demand price—are listed separately on the Spot VMs pricing page.

M2 machine types

The following table shows the calculated cost for M2 machine types. To use these machine types, you must request quota using one of the following options:

Request access to evaluation quota so that you can test the performance of these machine types. Any VMs you create with these machine types count against the evaluation quota and are billed using the evaluative prices listed below. Evaluation quota persists only for a limited amount of time on your project.
Purchase a 1 year or 3 year commitment for sustained usage. Commitments are not billed incrementally. Commitments bill you a monthly fee for the duration of your commitment term even if you do not use any of the committed resources.

These machine types are only available in select zones.

Spot prices, which typically offer the largest discounts—up to 91% off of the corresponding on-demand price—are listed separately on the Spot VMs pricing page.

M1 machine types

m1-ultramem and m1-megamem machine types have greater than 14 GB of memory per vCPU. The following describes the pricing per vCPU and GB of memory of these machine types.

The following table shows the calculated cost for m1-megamem and m1-ultramem machine types. The vCPUs and memory from each of these machine types are billed by their individual memory-optimized vCPUs and memory prices but these tables provide the cost that you can expect using a specific machine type.

These machine types are only available in select zones.

Committed Use Discounts apply to memory-optimized machine types only if you buy the commitment type specifically for memory-optimized machine types.
If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply.
Spot prices, which typically offer the largest discounts—up to 91% off of the corresponding on-demand price—are listed separately on the Spot VMs pricing page.

Storage-optimized machine type family

The storage-optimized machine family is suitable for workloads that are low in core usage and high in storage density. For example, the Z3 machine series is useful for scale-out analytics workloads, flash-optimized databases, and other database workloads.

Z3 machine types

Z3 VMs are powered by the 4th generation Intel Xeon Scalable processors (code-named Sapphire Rapids), DDR5 memory, Google's custom Intel Infrastructure Processing Engine (IPU), and the latest generation of Local SSD. The following table describes the pricing per vCPU, GB of memory, and GiB of Local SSD for Z3 machine types.

Z3 highmem machine types

The following table shows the calculated cost for z3-highmem-88 and z3-highmem-176 machine types, which are the two Z3 predefined machine types. The vCPUs, memory, and Local SSD storage from each of these machine types are billed by their individual storage-optimized vCPU, memory, and Local SSD disk prices, but these tables provide the total cost that you can expect using a specific machine type.

Accelerator-optimized machine type family

Accelerator-optimized VMs are optimized for massively parallelized CUDA compute workloads, such as machine learning (ML) and high performance computing (HPC). The accelerator-optimized machine family is available in A3 High machine types, A2 Standard and Ultra machine types, and the G2 machine type.

A3 VMs are preconfigured in one shape with eight of NVIDIA H100 GPUs, A2 VMs are preconfigured with a set number of NVIDIA A100 GPUs, and G2 VMs come with NVIDIA L4 GPUs.

The accelerator-optimized machine types are billed for their attached GPUs, predefined vCPU, memory, and bundled Local SSD storage (if applicable).

The total cost of running each accelerator-optimized machine type is shown in the following tables. For VMs that have custom memory options available, the default memory value is used in the calculation.

All accelerator-optimized machine types are eligible for resource-based committed use discounts. These machine types are not eligible for sustained use discounts or flexible committed use discounts.

A3 High machine types

The following table shows the total calculated cost that you can expect for predefined A3 High machine types. This total cost includes the cost for the GPUs, vCPU, memory, and Local SSD storage.

A2 Ultra machine types (total cost)

The following table shows the total calculated cost that you can expect for these predefined A2 Ultra machine types. This total cost includes the cost for the GPUs, vCPU, memory, and Local SSD storage.

**For committed use discounts pricing on the A2 Ultra machine series, connect with your sales account team.
If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply. Spot prices, which typically offer the largest discounts—up to 91% off of the corresponding on-demand price—are listed separately on the Spot VMs pricing page.

A2 Standard machine types (total cost)

The following table shows the total calculated cost that you can expect for these predefined A2 machine types. This total cost includes the cost for the GPUs, vCPU, and memory.

G2 machine types (total cost)

The following table shows the total calculated cost that you can expect for these predefined G2 machine types. This total cost includes the cost for the GPUs, vCPU, and default memory.

The on-demand prices for custom machine types include a 5% premium over the on-demand prices for standard machine types. If you are specifying custom memory, use the Google Cloud Pricing Calculator to calculate the total cost.

Shared-core machine types

Unlike predefined machine types and custom machine types, shared-core machine types are not billed on their individual resources. Each machine type has a defined price for both vCPUs and memory.

Shared-core machine types offer bursting capabilities that allow instances to use additional physical CPU for short periods of time. Bursting doesn't incur any additional charges. You are charged the listed on-demand price for your VMs.

E2 shared-core machine types

When you use E2 shared-core machine types, your VM runs two vCPUs simultaneously shared on one physical core for a specific fraction of time in the following ways, depending on the machine type:

e2-micro sustains 2 vCPUs, each for 12.5% of CPU time, totaling 25% CPU time and effectively consuming 0.25 cores.
e2-small sustains 2 vCPUs, each at 25% of CPU time, totaling 50% CPU time and effectively consuming 0.50 cores.
e2-medium sustains 2 vCPUs, each at 50% of CPU time, totaling 100% CPU time and effectively consuming 1.0 cores.

Each vCPU can burst up to 100% of CPU time, for short periods, before returning to its prescribed CPU time limitation.

E2 shared-core VMs are eligible for discounts from resource-based CUDs and spot VMs. These VMs are not eligible for flexible CUDs and SUDs.

When you purchase a commitment for vCPUs, you need to commit to a minimum of 1 vCPU. e2-micro and e2-small machine types effectively consume less than 1 vCPU. As a result, when you purchase a commitment for either of these machine types, ensure that your commitment meets the following minimum VM requirements:

At least 4 VMs if you want to commit to e2-micro resources.
At least 2 VMs if you want to commit to e2-small resources.

For more information, see E2 shared core machine types and CPU bursting.

N1 shared-core machine types

N1 shared-core VMs are available as f1-micro and g1-small machine types. VMs of both machine types run a single vCPU for a specific fraction of time in the following manner:

f1-micro sustains a single vCPU for up to 20% of CPU time.
g1-small sustains a single vCPU for up to 50% of CPU time.

Each vCPU can burst up to 100% of CPU time, for short periods, before returning to its prescribed CPU time limitations.

Discounts for N1 shared-core machine types apply in the following way:

N1 shared-core VMs are eligible for discounts from SUDs and spot VMs. SUDs for these VMs are calculated separately from N1 predefined and custom VMs.
N1 shared-core VMs are not eligible for resource-based or flexible CUDs.

Learn more about N1 shared-core machine types.

Tier_1 higher bandwidth network pricing

You can configure your N2, N2D, C2, C2D, C3, C3D, C4, C4A, Z3 and M3 machine types to use per VM Tier_1 networking performance. The following machine series require a minimum number of vCPUs to use this feature:

C2 and C3D: at least 30 vCPUs
C2D, C4A, and N2: at least 32 vCPUs
C3: at least 44 vCPUs
C4 and N2D: at least 48 vCPUs
M3: at least 64 vCPUs

Pricing is also dependent upon the regions and zones where the VM is located.

N2, N2D, C2, M3

If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply.
^*Spot prices are dynamic and can change up to once every 30 days. For more information about Spot VM pricing, see the Spot VMs documentation.

C2D

C3

The cost of using Tier_1 networking with C3 is described below.

C3D

The cost of using Tier_1 networking with C3D is described in the following table.

C4

The cost of using Tier_1 networking with C4 is described in the following table.

C4A

The cost of using Tier_1 networking with C4A is described in the following table.

Z3

The cost of using Tier_1 networking with Z3 is described in the following table.

Simulated maintenance event pricing

Starting February 10, 2020, there is no cost to running simulated maintenance events.

Prior to this date, the following charges apply:

Simulated maintenance on instances configured for live migration incur costs for each of the following instance resources:
- Price per vCPU on the instance, where f1-micro and g1-small are each equivalent to 1 vCPU: $0.040
- Price per GB of memory: $0.010
- Price per GB of Local SSD space: $0.001
Simulated maintenance on Spot VMs (and preemptible VMs): Free
Simulated maintenance on instances configured to stop and restart: Free

Suspended VM instances

When you suspend an instance, Compute Engine preserves the memory and device state. While you are not charged for the VM instance as if it were running, suspended instances still incur charges for the following:

Suspended instance memory and device state.
Suspended Local SSD data & metadata.
Any Persistent Disk usage.
Any static IPs attached to the VM instance.

What's next

Read the Compute Engine documentation.
Get started with Compute Engine.
Learn about Compute Engine solutions and use cases.
Learn more about Compute Engine machine types.
Refer to the Pricing Overview documentation.
Review Disk and image pricing, Networking pricing, Sole-tenant node or GPU pricing.
Try the Pricing Calculator.

Request a custom quote

With Google Cloud's pay-as-you-go pricing, you only pay for the services you use. Connect with our sales team to get a custom quote for your organization.

Contact sales