GCP has a discount mechanism that AWS doesn't have an equivalent for
Sustained Use Discounts apply automatically when you run Compute Engine instances for more than 25 percent of a month. The longer the instance runs, the bigger the discount. Run it the entire month and you get up to 30 percent off automatically, with no commitment and no upfront payment
This is genuinely good. It means GCP rewards you for running stable workloads without requiring you to predict the future
But there's a better option on top of it for workloads you're confident about
Committed Use Discounts
One year commitment: 37 percent off. Three year commitment: 55 percent off. Applied at the vCPU and memory level rather than specific instance types, which gives you more flexibility than AWS reserved instances if you change machine types
The mistake teams make with CUDs is the same one they make with AWS reserved instances: committing before they've rightsized
If your instances are 40 percent larger than they need to be and you commit to that size for three years, you've locked in a discounted price on waste
The right sequence on GCP:
Rightsize your compute first. Use the Recommender in the GCP console - it analyzes actual utilization and suggests appropriate machine types. Take those recommendations seriously, test them, implement them
Then look at what's left. What are you running that's stable, correctly sized, and not going away? That's your commitment baseline
Commit to that baseline with CUDs. Let the variable workload stay on the automatic Sustained Use Discounts
The combination of rightsizing plus CUDs on GCP typically saves 45 to 60 percent compared to running on-demand unoptimized instances
The savings are there. The order matters
Top comments (0)