Google Cloud discounts & reseller pricing

GCP has more discount layers than any other cloud on this list, and one of them is automatic. The four layers first, then the channel. Layer 1: Sustained use discounts (SUD) — automatic, no sign-up. Once an instance runs for more than 25% of a billing month, GCP starts reducing the rate proportionally; running the full month saves up to about 30%. You did nothing and it already applies. No other major cloud has an equivalent. Layer 2: Committed use discounts (CUD) — the deepest layer. Commit for one or three years, either to resources or to spend: | Type | What you commit to | 1 year | 3 years | |---|---|---|---| | Resource-based CUD | vCPUs / memory in a chosen region | up to ~37% | up to ~55% (up to 70% on memory-optimised shapes) | | Flex CUD | hourly spend | ~28% | ~46% | Flex CUDs can be spread across machine families and regions, which is materially more flexible. Layer 3: Spot. Reclaimable capacity, documented at up to 91% off, in exchange for termination at any time (roughly 30 seconds' notice). Layer 4: Channel rebate. Google Cloud Partner Advantage has four levels (Member / Partner / Premier, with Diamond added as a top tier in 2026). Higher tiers carry larger rebates, and this layer stacks with the official discounts above. ## ⛔ Where a lot of write-ups get it wrong: SUD and CUD do not stack Google's documentation states plainly that discount types cannot be combined, and that CUD takes priority — resources covered by a CUD no longer receive SUD, and SUD applies only to usage above the committed amount. The correct combination is therefore: cover your predictable baseline with CUDs and let SUD handle the variation above it. If you plan to buy a CUD, do not also model the 30% SUD on top — that is double counting. As a top-tier reseller across five major clouds, our overall billing quote typically runs 30–50% below official direct pricing, tiered by volume: smaller commitments land nearer the 30% end, while stable usage with a clear commitment period can get close to 50% off.

Specs & pricing

Why the official site has no discount but a reseller does

One common misunderstanding first: GCP is not "no discounts" — its discounts all sit on the usage side. Three of the four layers are official; they optimise *how you consume*, not *who you buy from*: - SUD — automatic; run longer, pay less, up to about 30% - CUD — commitment for discount, up to about 55% (Flex about 46%), and non-refundable - Spot — reclaimable capacity, up to about 91%, terminable at any time They restructure the *usage* side of "unit price × usage = bill". A reseller acts on the unit price. Google Cloud Partner Advantage has four levels: | Tier | Roughly | |---|---| | Member | Sign-up only; almost no financial incentive | | Partner | Training and assessment completed; a dedicated discount channel opens | | Premier | Revenue targets and technical certifications; meaningful rebates | | Diamond | Added in 2026 as the top tier, for the highest-performing partners | A top-tier reseller sits at the highest level, receives the rebate, and passes part of it through to end customers. The key difference: GCP has no EDP (as AWS does) and no EA (as Azure does). Its commitment discount *is* the CUD, and the channel side contributes only a rebate. So "channel discount" and "official discount" are two different layers that do stack — which is easier to explain here than on AWS.

How the four layers should actually combine

There is one principle: CUDs cover the baseline, SUD handles the variation, Spot takes the interruptible work. Applied to a typical workload: 1. Work out the baseline first — the part you are certain will run all year (a resident API service, a database). That is what a CUD should cover. 2. Let SUD handle everything above it — it is automatic. Note this part does not stack with the CUD: resources covered by a commitment stop receiving SUD, and SUD applies only above the committed amount. 3. Put interruptible work on Spot — batch jobs, offline ETL, the checkpoint phase of model training. Deepest discount, but you must design for resumption. The most common error is multiplying the three discount rates together. Their scope is mutually exclusive, not additive. Adding them yields a saving that does not exist in practice — and then people budget against that number. ⚠️ One more warning: CUDs are non-refundable. Overcommit and the excess is pure waste — more expensive than on-demand. Commit only what you are certain to consume.

Card-free setup: what actually happens

GCP's setup friction sits in two places: payment and quota. 1. A usable payment method is required, normally an international card; new projects may also need the billing account verified first 2. Some resources have default quota limits (GPU types, per-region quotas) that need an increase request 3. Billing account verification sometimes routes to manual review, on an unpredictable timeline A channel handles it this way: the project is opened through a partner that has already passed Google Cloud verification, and the bill is settled centrally — so you provide no personal card and no identity documents. You deploy and manage resources in the Cloud Console yourself and pay us on the agreed terms. Nothing in that chain "bypasses" verification; we carry the verification and settlement responsibility for the setup step.

Getting set up: four steps

1. Tell us the use case and rough volume — workload type (analytics, AI, containers, site), main regions, approximate monthly spend, and whether the baseline is predictable.
2. Get the plan and the rate — discount level, settlement cycle, and how much CUD to commit to (this step determines how much you actually save).
3. Open the project, or connect the one you already have — including card-free payment configuration. Existing projects keep their resources and IAM settings.
4. Top up on your own and settle per plan — 24/7 top-ups credited instantly; renegotiate when your volume changes tier.

When you should not use a reseller

This is not politeness. Recommending a plan that does not pay off costs far more in refunds and reputation than the order is worth. On GCP there is a specific version of this: we will tell you not to commit when the volume is not there, because an overcommit is non-refundable and we do not want you paying for it.

• You already have a corporate account and can buy direct — with an existing billing account and dedicated ops, direct purchase is smoother.
• Your spend is small and occasional — the margin will not cover the extra communication and settlement overhead.
• Compliance requires a direct contract — some industries require Google itself as the contracting party, or a billing entity in a specific jurisdiction.
• You are only testing an idea — free credits or pay-as-you-go are the right entry point for short-term validation.
• Your baseline is entirely unpredictable — CUDs are non-refundable, so committing on unpredictable usage loses money. On-demand plus SUD is the steadier choice.

The three routes side by side

Read the same table from another angle: on GCP, what we replace is not just overseas payment but the arithmetic of "how much CUD should I commit to". Official discounts and channel rebates are separate layers that stack — but making them work starts with getting the baseline right.