The Complete Google Cloud Licensing and Pricing Guide 2026

Google Cloud committed use discounts (CUDs) run 25 to 70 percent off on-demand pricing depending on commitment type and term, sustained use discounts add up to 30 percent automatically without any commitment, and a negotiated Google Cloud Agreement layers a further 8 to 25 percent enterprise discount on top. Network egress, Google Workspace, and Gemini each carry separate commercial mechanics that inflate the bill if left unmanaged. This pillar documents the full Google Cloud commercial surface for 2026 buyers, from the automatic discounts that apply with no contract to the bespoke enterprise agreement that carries the largest single lever.

How Google Cloud pricing actually works

Google Cloud bills on consumption, metered per second for compute and per gigabyte-month for storage, against a public on-demand price list. The on-demand rate is the ceiling, not the price most enterprises pay. Three layers sit below it: automatic sustained use discounts, committed use discounts you opt into, and a negotiated enterprise discount under a Google Cloud Agreement. Understanding which layer applies to which service, and in which order, is the whole game.

The layers stack in a defined order. On-demand pricing is reduced first by any sustained use discount, then a committed use discount applies to the committed portion of spend, and finally the enterprise discount under the Google Cloud Agreement applies across eligible services. Because the reductions compound rather than add, a 30 percent sustained use discount and a 20 percent enterprise discount produce a 44 percent effective reduction, not 50 percent. Buyers who add the percentages overestimate their savings and under-negotiate the enterprise discount as a result.

A further wrinkle is that not every service participates in every layer. Sustained use applies mainly to Compute Engine. Committed use discounts cover compute and a defined set of services. The enterprise discount under a Google Cloud Agreement covers most eligible services but typically excludes or only partly covers third-party Marketplace purchases and certain network charges. Mapping each major service to the discount layers it qualifies for is the first step in any serious Google Cloud cost model.

Sustained use discounts

Sustained use discounts (SUDs) apply automatically to Compute Engine and a few related services when an instance runs for a significant share of the billing month. Google applies them with no contract and no commitment: the longer an eligible resource runs in a month, the larger the automatic discount, reaching up to 30 percent for resources that run the full month on general-purpose machine types. The discount is calculated incrementally across usage tiers within the month, so a workload that runs continuously reaches the maximum automatically.

SUDs matter for two reasons. First, they are free, so any workload that runs continuously is already discounted before you commit to anything. Second, they change the committed use discount math, because a CUD must beat the sustained use discount you would have received anyway to deliver net value. A buyer who commits without modeling the SUD baseline can pay for a discount they were already getting, converting a free reduction into a paid one with no incremental benefit.

Committed use discounts: spend-based vs resource-based

Google Cloud offers two distinct committed use discount models, and confusing them is the most common Google Cloud commercial error. Resource-based CUDs commit to a specific quantity of vCPU and memory in a region for one or three years, in exchange for a deep discount, but they bind you to a machine family and region. Spend-based CUDs commit to a dollar amount of hourly spend on a service category, such as Compute or Cloud SQL, with more flexibility but a smaller discount.

The flexible CUD, introduced to compete with the portability of AWS Savings Plans, applies a spend commitment across multiple eligible services and regions. It carries a lower headline discount than a resource-based commitment but removes the stranded-commitment risk that catches re-architecting estates. The choice is not academic: a resource-based commitment stranded on an abandoned machine family delivers zero value while still billing, so the flexible model is often the safer economic choice even at a lower rate. We compare the two cloud models directly in GCP CUD vs AWS Savings Plan.

CUD terms and discount bands

Committed use discounts come in one-year and three-year terms. The three-year term roughly doubles the discount of the one-year for the same commitment. The bands below reflect typical 2026 Compute Engine commitments and vary by machine family and region.

The deeper resource-based discounts apply to specialized machine families such as memory-optimized instances, where a three-year commitment can reach 70 percent. General-purpose families sit lower, often in the 55 to 60 percent range at three years. The lesson is to commit the deepest where the workload is most stable and most expensive, and to use flexible or spend-based commitments everywhere else. See committed use discounts for the full mechanics, including how commitments are applied across projects in a billing account.

Commitments also interact with each other and with sustained use. A resource-based CUD on a continuously running instance consumes the usage that would otherwise have earned sustained use, so the incremental benefit of the CUD is the gap between its rate and the sustained use rate, not its full headline discount. Modeling that gap, rather than the headline, is how a buyer decides whether a given commitment is worth making at all.

The Google Cloud Agreement and enterprise discount

Large buyers negotiate a Google Cloud Agreement, the enterprise contract that layers a private discount on top of CUDs and SUDs in exchange for a multi-year spend commitment. The enterprise discount typically runs 8 to 25 percent depending on annual commit and term, and applies across most eligible services rather than to a single resource type. It is the closest Google equivalent to an AWS EDP or an Azure MACC, and like those, the discount, the ramp, the service eligibility, and the exit terms are all bespoke.

The Google Cloud Agreement is where the real negotiation happens. A buyer who accepts the standard CUD and SUD discounts without pursuing an enterprise agreement leaves the largest lever untouched, because the enterprise discount compounds on top of everything else. The commitment is usually structured as a ramping annual spend, so the negotiation covers not just the discount percentage but the ramp shape, what counts toward the commit, and what happens if consumption falls short. See the GCP enterprise agreement guide and our Google Cloud negotiation practice.

The credible alternative is what moves an enterprise discount. Google, like every hyperscaler, prices the enterprise discount partly on competitive pressure, so a buyer with written AWS and Azure proposals for the same workloads negotiates from a stronger position than one who has already committed publicly to Google. The discount is a function of the deal's strategic value to Google as much as its size.

Network egress and the data-transfer trap

Network egress, the charge for moving data out of Google Cloud to the internet or across regions, is the cost line that surprises buyers most. Egress is metered per gigabyte at rates that escalate with volume and destination, and it is frequently excluded or only partly covered by the enterprise discount. A data-intensive workload can accumulate egress charges that rival its compute bill, and because egress is a function of architecture, the cost is often locked in long before anyone reviews it.

Egress is negotiable, but only if raised explicitly. Google will discount committed egress volumes and, under data-transfer and waiver programs, reduce egress for qualifying workloads and multi-cloud architectures. Buyers who model egress before committing avoid the architecture decisions that lock in years of avoidable transfer cost, such as cross-region replication patterns or chatty multi-cloud designs that move data out of Google Cloud on every transaction. See GCP egress negotiation.

Google Workspace enterprise pricing

Google Workspace is licensed per user per month across Business and Enterprise editions, separate from the Google Cloud Platform consumption contract. Enterprise editions are quote-based for large organizations, and the per-seat price is negotiable on volume and term in a way the published Business-tier pricing is not. The two contracts, Workspace and Google Cloud Platform, are often negotiated by different teams, which is itself a source of lost negotiating ground when they could be combined into one deal.

The Workspace commercial traps are seat-count drift, edition over-assignment, and the AI add-on. Organizations routinely license Enterprise Plus for users who need only Enterprise Standard, and add Gemini for Workspace across the full seat count rather than the users who use it. Seat-count drift accumulates as leavers are not deprovisioned and contractors are licensed at full edition. See Google Workspace enterprise pricing for the edition-by-edition breakdown and the right-sizing approach.

Gemini and AI pricing in 2026

Google sells AI two ways in 2026: Gemini for Workspace as a per-seat productivity add-on, and Gemini and Vertex AI model usage on Google Cloud Platform billed per token or per request. The two are priced and contracted separately, and both carry the now-familiar AI cost pattern of a per-seat or per-call rate that multiplies quickly across an enterprise. A per-seat add-on applied to every employee, or a high-volume model integrated into a customer-facing product, can become one of the largest lines on the bill.

The buyer discipline mirrors every other AI add-on: scope the per-seat product to users who use it, cap the rate, and secure price protection, while treating Vertex AI model spend as committable consumption that the enterprise discount and CUDs can cover. Token-based model spend should be forecast and, where stable, committed under the enterprise agreement rather than left at on-demand rates. See Gemini enterprise pricing.

Marketplace and committed spend burndown

Google Cloud Marketplace purchases of third-party software can burn down an enterprise spend commitment, which helps a buyer reach a committed amount but, as with other clouds, frequently at a worse discount than direct service consumption. The enterprise discount often does not apply to Marketplace spend, so a dollar spent in Marketplace counts toward the commit but is paid at full price, where the same dollar on direct services would have been discounted.

Negotiating Marketplace treatment into the Google Cloud Agreement, rather than discovering it at burndown, preserves the value of the commitment. Marketplace also offers a procurement channel for software the enterprise already buys, consolidating billing under the cloud contract, but the value depends entirely on whether the Marketplace price beats the direct contract, which it often does not. See GCP marketplace procurement.

Support and the role-based model

Google Cloud support is sold in tiers, from Standard through Enhanced to Premium, priced as a combination of a base fee and a percentage of monthly spend. Premium support adds a technical account manager, faster response targets, and operational reviews, and is the tier most large enterprises require for production workloads. The support cost sits on top of the discounted service spend, not inside it, so it should be modeled separately in any total-cost view.

Support pricing is part of the enterprise negotiation. The percentage component and the base fee are both subject to discussion in a Google Cloud Agreement, particularly for buyers committing significant multi-year spend. Treating support as a fixed list-price add-on, rather than a negotiable line in the overall deal, leaves money on the table on every renewal.

Common Google Cloud commercial mistakes

Four mistakes account for most Google Cloud overspend. The first is committing to resource-based CUDs on machine families a re-architecting estate later abandons, stranding the commitment. The second is ignoring the sustained use baseline, so a CUD is paid for discount the workload already received free.

The third is failing to model egress before architecture decisions lock it in, leaving years of avoidable transfer cost. The fourth is accepting CUD and SUD discounts without pursuing a Google Cloud Agreement, leaving the enterprise discount, the largest single lever, on the table. A fifth, increasingly common in 2026, is rolling out Gemini for Workspace estate-wide on a pilot that is never scoped down. Each is preventable with a baseline built before the commitment.

Negotiation framework

The buyer-side Google Cloud negotiation runs in four phases. Phase 1: baseline. Build an independent consumption forecast, model the sustained use discount the workload already earns, and identify the services that dominate spend and justify CUD or enterprise-discount focus. Phase 2: competitive engagement. Get written AWS and Azure proposals for the same workloads, as detailed in AWS vs Azure vs GCP pricing. The credible alternative is the only thing that moves an enterprise discount.

Phase 3: Google negotiation. Negotiate the enterprise discount, the commitment ramp, CUD structure, egress treatment, Marketplace burndown, support pricing, and exit terms in parallel rather than sequentially, so each concession can be traded against another. Phase 4: contract close. Verify the final Google Cloud Agreement language against the negotiated terms before signing, with particular attention to service eligibility, egress exclusions, and what happens if the committed spend is not met.

Renewal economics

A Google Cloud Agreement renewal begins 9 to 12 months before term end. Google proposes a continuation with a modest discount adjustment and a fresh ramp. The buyer-side preparation is identical to the original: baseline, competitive alternative, and parallel negotiation. Treating renewal as a routine extension typically costs 5 to 10 percent of achievable discount, because the account team prices a passive renewal more aggressively than a contested one.

The renewal is also the moment to correct anything the original deal got wrong: an egress exclusion that proved expensive, a Marketplace treatment that destroyed value, or a commitment ramp that outran actual consumption. A buyer who tracked these through the term arrives at renewal with the evidence to reset them.

Building a true total-cost view

The single most useful artifact a Google Cloud buyer can build is a total-cost view that separates the discount layers from the cost lines they do and do not touch. Compute, reduced by sustained use, CUDs, and the enterprise discount, sits in one column. Egress, support, Marketplace, and AI add-ons, each with their own discount treatment, sit in others. Without this separation a buyer cannot tell whether a proposed enterprise discount is generous or whether it simply applies to a small slice of a bill dominated by undiscounted egress and Marketplace spend.

The view also exposes the compounding order that catches buyers who add percentages. Because sustained use applies first, then CUDs, then the enterprise discount, the effective compute rate is the product of the three, not their sum. Modeling the effective rate per service, and weighting by where the spend actually concentrates, turns a pile of percentages into a single defensible number the buyer can negotiate against and track across the term.

For the surrounding commercial framework, see committed use discounts, the GCP enterprise agreement, egress negotiation, GCP CUD vs AWS Savings Plan, and the Google Cloud vendor hub. For an engagement on an active Google Cloud negotiation, see cloud contract negotiation or software licensing advisory.

The bottom line on Google Cloud cost

Google Cloud cost is governed less by the published price list than by how well a buyer assembles the three discount layers and contains the cost lines that sit outside them. The compute bill, reduced by sustained use, committed use, and the enterprise discount in that compounding order, is the part most buyers focus on, yet egress, support, Marketplace, and AI add-ons frequently decide whether a deal is genuinely competitive or merely looks it on the compute headline.

The buyer-side discipline is consistent across all of them. Model the effective rate per service rather than trusting headline percentages, net out the free sustained use baseline before valuing a commitment, raise egress and Marketplace treatment explicitly in the enterprise negotiation rather than discovering them at burndown, and scope AI to the users and workloads that genuinely consume it. None of these requires special bargaining power; they require a forecast built before the commitment and a contract read before signing.

A Google Cloud Agreement negotiated this way, with the enterprise discount pursued rather than left on the table and the surrounding cost lines contained, routinely lands 10 to 25 percent below the account team's opening position over the term. The difference is preparation: a buyer who arrives with an independent forecast, a credible competitive alternative, and a clear view of which discount layer touches which service negotiates from evidence, while one who reacts to the proposal negotiates from the vendor's framing.