Sustained use discounts apply automatically to eligible Compute Engine usage once a resource runs for more than 25 percent of a billing month, increasing for every additional hour. The maximum is a net discount of up to 30 percent for general purpose N1 instances run the whole month, and up to 20 percent for other eligible families such as N2, N2D, C2, and the M1 and M2 memory optimized machines. There is nothing to buy and nothing to manage. The catches: newer machine series do not earn sustained use discounts at all, the discount does not stack on resources already covered by a committed use discount or other discount, and the cap means it can never replace a commitment for steady workloads.
Treat sustained use discounts as a free floor, then build commitments on top. Here is exactly how they behave.
How do sustained use discounts actually work?
Compute Engine watches how much of the billing month each eligible resource runs. Once usage passes 25 percent of the month, every additional hour earns a discount that grows as usage climbs, so a machine that runs the entire month reaches the maximum rate. The discount is calculated and applied automatically, with no purchase, no term, and no risk.
Crucially, the discount is based on aggregated usage of like resources in a region, not just one instance running continuously. Compute Engine combines eligible usage to maximise the discount, so a fleet that collectively runs a lot of vCPU hours benefits even if individual machines come and go. This is one reason fragmenting a workload across many projects can matter, as covered in projects, folders, and cost structure.
The headline for a buyer: sustained use discounts are the reward for simply running steadily. You earn them whether or not you do anything, which is exactly why they are not a strategy on their own.
Which machine types earn sustained use discounts, and which do not?
This is where teams overestimate the benefit. Sustained use discounts apply to a specific set of families and rates.
| Machine family | Sustained use discount |
|---|---|
| N1 general purpose | Up to about 30 percent |
| N2, N2D general purpose | Up to about 20 percent |
| C2 compute optimized | Up to about 20 percent |
| M1, M2 memory optimized | Up to about 20 percent |
| Newer series such as Tau T2D, N4, C3, C4 | Not eligible, use committed use discounts |
The trap is assuming every machine earns the discount. The newer general purpose and compute series, which many teams adopt for better price performance, do not earn sustained use discounts at all and rely on committed use discounts for their savings. Confirm the current eligibility list on the GCP pricing documentation before you assume a family qualifies (indicative, families change).
Why is a sustained use discount a floor, not a strategy?
Two limits keep sustained use discounts from doing the heavy lifting. First, the cap: even a machine that runs every hour of the month tops out around 20 to 30 percent depending on family. A committed use discount on the same steady workload can reach much deeper, often well beyond that. Second, the discounts do not stack: usage already covered by a committed use discount does not also earn a sustained use discount, and the committed rate applies first.
So for any workload that is genuinely steady, a committed use discount beats riding the sustained use discount, because the commitment goes deeper and the sustained use discount would otherwise be the only thing applying. The sustained use discount earns its keep on the steady usage you have not yet committed and on workloads too variable to commit confidently. It is the safety net beneath your commitment plan, which is why it features in the risk adjusted approach to GCP commitments.
A worked example: leaving discount on the table
A European SaaS company believed its steady GCP fleet was well discounted because sustained use discounts were applying automatically. Most of that fleet was steady enough to commit, and a slice had already moved to a newer machine series that earned no sustained use discount at all. We layered committed use discounts on the hardened baseline, which went far deeper than the sustained use cap, kept the sustained use discount working on the uncommitted remainder, and flagged the newer series machines that were silently earning nothing. Effective discount on the steady fleet improved well beyond what the automatic discount alone delivered. Figures are verified against billing data and anonymised.
How should you use sustained use discounts in practice?
Let them work, then build past them. Confirm which of your machines actually earn the discount, since newer series do not, commit the steady baseline with committed use discounts that go deeper than the sustained use cap, and leave sustained use discounts to cover the variable remainder for free. Watch for fleets that fragment across projects in ways that dilute the aggregation. This sits alongside commitment and rightsizing work in the GCP cost optimization guide.
Frequently asked questions
Get the buyer side GCP discount playbook
We confirm which of your GCP machines actually earn sustained use discounts and layer commitments where they go deeper, with zero provider commissions. It is the same discipline behind the 31 percent median reduction we see in the first 90 days. Our guarantee: we reduce your cloud spend or we reimburse our service fee.
Put a defensible number on your cloud spend.
No provider in the room, no published price list. Tell us your footprint and we will scope the savings against your billing data — we reduce your cloud spend or we reimburse our service fee.
The Cloud Spend Navigator: what changed in cloud pricing, commitments, and FinOps — no vendor spin.