Google Cloud's flexible commitments now discount G2 and G4 GPU VMs

Google Cloud made its compute flexible committed use discounts (CUDs) generally available for the G2 and G4 GPU machine series on August 11, 2026. G2 VMs use NVIDIA L4 GPUs and G4 VMs use NVIDIA RTX PRO 6000 GPUs. A flexible CUD is a commitment to an hourly spend rather than to specific machines, and the release note says that for these GPUs it needs no attached reservation.

The discounts

Google’s committed use discounts overview lists these rates for flexible CUDs:

What is covered	1-year term	3-year term
G2: vCPUs, memory and NVIDIA L4 GPUs	21%	43%
G4: vCPUs, memory and RTX PRO 6000 GPUs	16%	42%
Most general-purpose series, for comparison	28%	46%

GPUs are covered only on G2 and G4. No other accelerator-optimized machine series is on the list of eligible series, so this doesn’t apply to Google’s other GPU machine types.

How a flexible commitment works

• You commit to a minimum hourly spend on eligible resources, not to a number of GPUs.
• You pay that hourly amount for the whole term, whether or not you use it.
• The commitment applies to eligible spend across Compute Engine, GKE and Cloud Run.
• You can switch between eligible machine series and regions and keep the discount.
• Local SSD disks on these series are covered too.

A commitment bought in the first 49 minutes of an hour starts at the next hour; one bought later starts the hour after that.

Who it helps

Teams that run steady inference, video or rendering work on L4 or RTX PRO 6000 VMs and want a discount without tying it to one region or one machine type, or to a GPU reservation.

What to do

Commit only to the hourly spend you are sure to use for the full term, because unused commitment is still billed. On G4 the 1-year discount is small at 16%, so weigh it against the 42% 3-year rate and how long you expect to need the GPUs.