Cloud bills are rarely surprising because of the sticker price, they’re surprising because of what is not obvious upfront: egress fees, minimum commitments, and which discount actually applies to your workload. Here is how Google Cloud and AWS pricing actually compares.
On-Demand and Committed Pricing
Google Cloud bills per second across all VM types with no minimum. AWS bills per second too, with a 60-second minimum, and that applies across Linux, Windows, and SQL Server instances alike, AWS extended per-second billing to Windows and SQL Server back in 2021, so there’s no separate hourly-minimum penalty for running Windows on EC2 anymore. Beyond on-demand, both providers reward commitment differently:
- Google Cloud sustained-use discounts apply automatically once a VM runs more than 25% of a billing month, no commitment required, but this only benefits older machine families: up to around 30% for N1 and the memory-optimized M1/M2 series, up to about 20% for N2/N2D/C2. Newer families, E2, N4, C3, and C4, are not eligible for sustained-use discounts at all, which is easy to miss if you’re comparing against an older Google Cloud pricing guide.
- Google Cloud committed-use discounts reduce compute costs by roughly 55% for resource-based one- or three-year commitments (up to around 70% for memory-optimized machine types), committing to specific resource quantities rather than a dollar amount. As of January 21, 2026, Google has been migrating eligible billing accounts from the older credit-based, spend-based CUD model to a newer direct-discount model, worth confirming which model your account is actually running under before budgeting off an older guide (including this one).
- AWS Savings Plans commit to a dollar-per-hour spend, applying automatically across EC2, Fargate, and Lambda regardless of instance family, up to about 66% off.
- AWS Reserved Instances lock in a specific instance family for one or three years, up to 72% off for Standard RIs (66% for the more flexible Convertible RIs).
The structural difference matters more than the headline percentages: AWS treats portability as the default and prices the commitment as the exception, Google Cloud does the reverse, deeper discounts for accepting a resource lock-in, a free automatic discount for flexibility, though that automatic discount is worth less than it used to be if your workload runs on one of the newer, ineligible machine families.
Spot and Preemptible Instances
For interruptible workloads, Google Cloud’s preemptible VMs discount up to roughly 80% off on-demand, and AWS Spot Instances up to roughly 90% off, both can be reclaimed with short notice, so they fit batch jobs and fault-tolerant workloads better than anything user-facing.
Storage and Network
Both charge storage by the gigabyte-month, with cheaper tiers for infrequently accessed data (Google Cloud Nearline/Coldline, AWS S3 Infrequent Access/Glacier) that add retrieval fees in exchange for the lower base rate. Network egress is where costs most often surprise people, both charge per gigabyte for data leaving the network to the internet, with AWS adding volume-based tiers as usage grows. The specific per-GB numbers on both platforms drift over time, treat any number you see (including in this guide) as a reference point, not a live quote, and check each provider’s current pricing page before budgeting.
Getting the Best Rate
The highest-leverage move on either platform is matching workload shape to discount type: steady, predictable workloads belong on committed-use or reserved pricing; spiky or fault-tolerant workloads belong on spot/preemptible; anything still being sized should stay on-demand until the pattern is clear. On Google Cloud specifically, that now means checking your machine family against the current sustained-use eligibility list before assuming the automatic discount applies at all. For workload-specific comparisons, see our Google Cloud vs AWS for machine learning guide.