Two Commitment Models, One Decision
Google Cloud sells two fundamentally different committed use discounts, and the GCP CUD vs Flexible CUD decision turns on how much your compute fleet moves. A resource-based CUD commits you to a minimum quantity of Compute Engine resources — vCPU and memory — in a specific region and machine series, for one or three years. A spend-based flexible CUD commits you to a minimum hourly dollar amount and then applies the discount to any eligible usage, regardless of machine family, region, or whether the workload runs on Compute Engine, GKE, or Cloud Run.
That structural difference is the whole story. Resource-based commitments are a quantity promise tied to a fixed configuration; flexible commitments are a spend promise that travels with your architecture. Before signing either, map your fleet into the stable core that will not move for the term and the volatile remainder that will — because the right answer is almost always a blend of both, not one or the other. The same logic underpins the wider Google Cloud CUD optimisation framework.
Discount Depth: GCP CUD vs Flexible CUD
Resource-based CUDs deliver the deeper discounts. For most machine types you receive up to 55% off on-demand pricing on a three-year commitment, and memory-optimised series (M1, M2, M3, M4) reach as high as 70% off on three-year terms. On general-purpose machines the one-year rate sits around 37% and the three-year rate around 55%. Flexible CUDs trade depth for reach: a flat 28% for one-year and 46% for three-year commitments, applied across any machine family or region in the billing account.
| Dimension | Resource-based CUD | Flexible CUD (spend-based) |
|---|---|---|
| 1-year discount | ~37% (general purpose) | 28% (all eligible compute) |
| 3-year discount | Up to 55%, up to 70% memory-optimised | 46% (all eligible compute) |
| Scope locked to | Region + machine series | Billing account (any region/family) |
| Covers GKE / Cloud Run | No (Compute Engine resources) | Yes (VMs + containers + serverless) |
| Best for | Stable, predictable steady-state fleet | Changing, migrating, multi-service fleet |
On comparable three-year general-purpose usage, a resource-based CUD delivers roughly 37% off on a locked machine type while a flexible CUD delivers 28% off across any compute in the account. That 9-percentage-point gap is the explicit price of portability — and the number you should put against your forecast change risk before deciding.
Portability and the 2026 Multiprice Change
The flexible CUD's advantage is that it survives change. A single flexible commitment now covers your entire eligible VMs-plus-containers-plus-serverless footprint, so when you migrate from an older machine series to a newer one, shift a workload to a cheaper region, or move a service from Compute Engine into GKE Autopilot, the discount follows. A resource-based CUD does not — if you stop running the committed configuration, you keep paying for the commitment while losing the discount on what you actually run. That stranded-commitment risk is the single most common way CUD value leaks away.
From January 2026 Google migrated spend-based CUDs to a multiprice, or direct-discount, model. Previously the discount was structured as a commitment fee plus a credit offset calculated at list price, which made invoices hard to reconcile. Under the new model the discount applies straight to the SKU price. The headline 28% and 46% rates are unchanged — the change is reconciliation clarity, not economics. Separately, from 16 June 2026 CUD sharing is enabled by default on new billing accounts and is being extended to many resource-based CUDs across projects, which reduces the waste from a commitment sitting idle in one project while another pays on demand.
Which Commitment to Choose
Use resource-based CUDs for the predictable core — the production baseline that will run the same machine series in the same region for the full term. That is where the extra 9-plus points of discount are pure margin, and the lock-in costs you nothing because you were never going to move. Use flexible CUDs for everything with change risk: fleets mid-migration, autoscaling container platforms, multi-region footprints, and any workload where the machine family roadmap is uncertain. The deeper discount you forgo is cheap insurance against a stranded commitment.
In practice the strongest position is a layered one. Cover the stable baseline with resource-based CUDs to capture maximum depth, then place a flexible CUD on top of the variable layer to keep the discount portable. This mirrors the same depth-versus-flexibility trade enterprises face on flexible CUDs versus standard commitments, and it parallels how AWS buyers blend deep Savings Plans against Reserved Instances. The full vendor context sits in our Google Cloud vendor hub.
Negotiating the Enterprise Layer
The published CUD rate is the floor for a large buyer, not the ceiling. At enterprise scale, CUDs sit inside a Google Cloud agreement that can add a further negotiated discount on committed spend, plus credits, migration funding, and custom terms. Google discounts meaningfully against a multi-year spend commitment and against credible competitive pressure — the same dynamic we work in any cloud contract negotiation. Treat the 28%/46% and up-to-70% published rates as your starting reference, then negotiate the agreement layer on top.
The headline discount is rarely where deals go wrong; the terms are. Scrutinise the shortfall and true-up clauses that penalise underspending the commitment, the ramp schedule that front-loads obligation before your usage arrives, and how the agreement treats CUDs purchased mid-term. A deep discount wrapped in a punishing shortfall clause can cost more than a smaller discount with room to breathe. For the framework we use to structure these commitments, see the Cloud Contract Framework and the wider cloud contract negotiation guide. If a commitment renewal is approaching, request a confidential briefing before you sign.