Two Commitment Models
Comparing Azure vs GCP committed use discounts means comparing two structurally similar but differently-branded systems. Azure offers Reservations — a commitment to a specific VM family and region in exchange for the deepest discount — and Savings Plans, a commitment to a dollar-per-hour spend level that applies across any VM family, any region, and several compute services for a slightly lower but portable discount. Google offers resource-based CUDs, which lock to a specific machine series and region, and spend-based flexible CUDs, which apply across families and regions at the billing-account level.
The underlying logic is identical on both clouds: you trade flexibility for depth. The resource-locked instrument always pays more; the portable instrument always pays less but survives a change of plan. Where the platforms diverge is in the headline rates and in Google's automatic discounts — and those differences are exactly where the money is. For the wider context, our cloud licensing guide sets out how these commitments sit inside an enterprise agreement.
Side-by-Side: Discount Rates
The table below sets the instruments against each other at their published rates. Treat these as ceilings on standard pricing — the enterprise agreement discount you negotiate sits underneath them and compounds the saving.
| Instrument | Commitment | 1-Year | 3-Year |
|---|---|---|---|
| Azure Reservations | VM family + region | Up to ~72% | Up to 72% |
| Azure Savings Plans | $/hour spend, portable | Up to ~65% | Up to 65% |
| GCP Resource-Based CUD | Machine series + region | 37% | 55% |
| GCP Spend-Based Flexible CUD | Any family/region (billing) | 28% | 46% |
| GCP Sustained Use Discount | None (automatic) | Up to 30% | Up to 30% |
Azure's headline reservation discount runs deeper at the top end, up to 72%, but it demands the tightest commitment — a fixed VM family and region for the term. GCP's resource-based CUD tops out around 55% at three years, with the flexible CUD trading roughly nine points of discount for portability across families and regions. When a VM matches both an Azure Reservation and a Savings Plan, Azure applies the Reservation first because it carries the deeper rate. The same depth-versus-flexibility trade-off plays out in our sibling comparisons of Azure MACC vs CSP commitment and GCP CUD vs flexible CUD.
The deepest published rate is not the best outcome — the best outcome is the deepest rate you can actually consume. A 72% reservation on capacity you stop using is more expensive than a 46% flexible CUD you fully absorb. Coverage utilisation, not headline discount, is the number to optimise.
Sustained Use: GCP's No-Commit Discount
The single biggest structural difference is what happens with no commitment at all. Google applies sustained use discounts (SUDs) automatically, scaling up to about 30% once an eligible resource runs beyond 25% of the month — no contract, no purchase, no risk of stranded commit. Azure has no automatic equivalent: every discount on Azure requires you to actively buy a Reservation or a Savings Plan first.
For variable or unpredictable workloads, that difference is material. On GCP, simply running steady-state capacity earns a discount without locking you in; on Azure, the same workload pays full rate until you commit. The practical implication is that GCP rewards usage while Azure rewards commitment — so the platform that suits you depends on how confident you are in your forecast. For workloads you can forecast, both clouds reward you for committing; for those you cannot, GCP's automatic floor is a genuine edge. Our committed-use cost optimisation guide goes deeper on absorbing commitment without waste.
Layering Commitments Without Stranding Spend
The mistake we see most often is treating the choice as binary — all reservations or all flexible — when the right answer is almost always a layered portfolio. Identify the genuinely stable baseline of your estate and cover it with the deep, resource-locked instrument (Azure Reservations or GCP resource-based CUDs). Cover the predictable-but-evolving layer above it with the portable instrument (Azure Savings Plans or GCP spend-based flexible CUDs). Leave the genuinely variable top uncommitted — on GCP it still earns sustained use discounts automatically; on Azure it sets your next commitment-sizing exercise.
This layering is what keeps coverage utilisation high while protecting against stranded commit, and it is the same discipline behind a strong AWS portfolio in our AWS Savings Plans vs Reserved comparison. Run the same exercise across vendors and the single-versus-multi-cloud question sharpens — the licensing logic that pulls workloads onto one vendor's platform, examined in our Oracle OCI vs AWS database licensing comparison, applies here too. The Cloud Contract Framework sets out the full method.
Which to Choose
Choose Azure's model if your estate is stable and forecastable, you want the deepest available discount, and your workloads sit predictably within fixed VM families and regions — the up-to-72% reservation rate is the highest ceiling on the table. Pair it with Savings Plans for the evolving layer so you are not over-committed on capacity you might re-architect.
Choose GCP's model if your usage is variable or hard to forecast, you value automatic sustained use discounts that need no commitment, and you want flexible CUDs that travel across machine families and regions without renegotiation. The discount ceiling is lower, but the risk of stranded commitment is lower too — and for many real estates that nets out ahead.
Whichever platform you commit to, optimise for consumed discount rather than headline rate, layer deep and flexible instruments deliberately, and negotiate the underlying enterprise agreement on top — the commitment discount and the EA discount compound. To model your own coverage and pressure-test a hyperscaler proposal, request a confidential briefing, or explore the Microsoft and Google Cloud vendor hubs.