AWS, Azure, GCP, one workload, three prices.

AWS, Azure, and Google Cloud will price the same enterprise workload 15 to 30 percent apart, and the spread is set by how you shape the commit and the competition, not by list prices.

How do AWS, Azure, and Google Cloud commit programs differ?

AWS discounts through the Enterprise Discount Program, Microsoft through the Azure consumption commitment, and Google through committed use and spend based agreements. All three trade a multiyear spend promise for a percentage off, but the mechanics differ in ways that matter at exit.

The AWS pricing model applies an EDP discount across nearly all services. The Microsoft Azure consumption commitment counts eligible Azure and marketplace spend toward the commit. Google Cloud pricing layers committed use discounts per resource family under an enterprise spend agreement.

What discount should each provider concede at your commit level?

At 5 to 20 million dollars of annual commit, effective discounts of 12 to 25 percent are the realistic band across all three providers, before service specific programs. Below 5 million the bands compress. Above 50 million everything is bespoke.

• Entry commits, 1 to 5 million a year: 5 to 12 percent effective, with Google often most aggressive to win the logo.
• Mid commits, 5 to 20 million: 12 to 25 percent effective, where competition moves the number most.
• Large commits, 20 million plus: custom economics, private rate cards, and service specific floors.

Why effective rate beats headline discount

The headline percentage hides service mix. A 20 percent EDP on compute heavy spend can cost more than a 15 percent deal with deeper database and egress concessions. Model the effective rate against your actual service mix before comparing offers.

How do you run a credible three way competition?

A credible competition prices a real migration path for 20 to 30 percent of the estate, not a hypothetical full exit. Providers discount against feasibility. Portable workloads, containerized services, and analytics layers are believable. The core ERP database is not.

Sequence the competition correctly

• Baseline first: rebaseline usage and strip waste before sizing any commit; committed waste is permanent.
• Scope the portable tier: identify the 20 to 30 percent that genuinely moves, with engineering sign off.
• Run parallel proposals: request structured offers against the same workload definition and ramp.
• Trade at the end: bring exit terms and unit floors into the final round, when the discount is already set.

Where the common advice on cloud commits is wrong

The standard FinOps advice is to maximize the commit to maximize the discount. We disagree. In roughly a third of the 15 to 20 commit negotiations we advised in 2024 to 2025, overcommitment cost more than the discount returned, through shortfall penalties or forced spend on services that would otherwise have been optimized away. The buyer side move is to commit at 70 to 80 percent of honest forecast and let upside flow at the discounted rate. The discount on spend you should never have incurred is not a saving.

Providers do not price against your spreadsheet. They price against the probability that the workload actually moves.

How do marketplace purchases change the commit math?

Marketplace spend is the quiet flexibility lever in all three programs. Third party software bought through AWS Marketplace retires EDP commit, and eligible purchases count toward a MACC under the Azure pricing framework. Routing planned software renewals through the marketplace can close a commit shortfall without buying compute nobody needs.

Which exit and flexibility terms matter more than the discount?

Four terms decide whether year three belongs to you or to the provider: commit rollover, true down windows, marketplace counting, and renewal rate protection. All four are negotiable at signature and nearly impossible to add later.

• Rollover rights: unspent commit carries into a renewal term instead of expiring as penalty.
• True down window: a defined point to resize the commit after divestiture or optimization.
• Marketplace counting: third party software bought through the marketplace retires commit dollars.
• Renewal protection: discount floors that survive into the next term, not just the current one.

What to do next

The moves below turn this framework into a signed commit on your terms.

A sequence you can run this quarter

1. Rebaseline twelve months of usage per service family and strip the waste before sizing anything.
2. Define the portable workload tier with engineering and document the migration feasibility.
3. Set your commit at 70 to 80 percent of honest forecast and model the effective rate per provider.
4. Request structured proposals from at least two providers against the same workload definition.
5. Negotiate rollover, true down, and marketplace counting before final pricing, not after.
6. Lock renewal discount floors into the order form and diary the rebaseline six months before expiry.

White Paper · AWS

AWS, Azure, GCP. The coordinated competitive framework

Read it free.

Read the white paper

Frequently asked questions

Which is cheaper for enterprises: AWS, Azure, or Google Cloud?

No provider is consistently cheaper; the same workload prices 15 to 30 percent apart depending on service mix and negotiation. Compute heavy estates often favor one provider while data and analytics heavy estates favor another, so the answer is set by your mix and your leverage.

What discount does an AWS EDP give?

AWS EDP discounts typically run from mid single digits at entry commits to 25 percent or more at large multiyear commits. The effective rate depends on service mix, since some services carry deeper programmatic discounts than the EDP itself.

What is the difference between a MACC and an EDP?

Both trade committed spend for benefits, but the Azure MACC is a consumption commitment that eligible marketplace purchases retire, while the AWS EDP is a discount program applied to spend tiers. The marketplace treatment is often the deciding difference for software heavy estates.

How big should a cloud commit be?

Commit at 70 to 80 percent of your honest usage forecast. Overcommitting converts the discount into shortfall penalties or forced spend, and undercommitting only costs you a slightly lower tier, which is the cheaper mistake.

Can you really play the hyperscalers against each other?

Yes, if the competition is credible. Pricing a real migration of 20 to 30 percent of the estate with engineering sign off moved final offers 8 to 15 points in our 2024 to 2025 engagements, even where no workload ultimately moved.

What happens to unspent commit at the end of the term?

By default it expires and you pay the shortfall, depending on contract language. Rollover rights negotiated at signature carry unspent commit into the renewal term, which is why they belong on the term sheet from day one.