The Core Trade-off: Commitment vs Flexibility
The choice between a cloud Enterprise Agreement and pay-as-you-go pricing is a bet on your own forecast. Pay-as-you-go (PAYG) charges list rates with no commitment, no lock-in, and no minimum — you pay only for what you consume by the second or hour. An Enterprise Agreement, or any commitment-based construct, trades that flexibility for a discount: you promise a baseline of spend or capacity over one or three years, and the provider rewards the certainty with lower unit rates.
The discount is real and large — up to 72% in some AWS constructs — but it is only earned on usage that actually materialises. Commit to capacity you do not consume and the "discount" becomes a premium: you have pre-paid for compute that sits idle. This is why cloud pricing decisions should never start with the discount percentage. They should start with how confident you are in your steady-state demand over the commitment term.
How Each Hyperscaler Prices Commitment
The three major providers use different mechanics to price the same bargain. Azure leans on the Microsoft Azure Consumption Commitment (MACC) and committed-use discounts; AWS uses Savings Plans and Reserved Instances layered under an Enterprise Discount Program (EDP); Google Cloud uses spend-based and resource-based committed use discounts (CUDs). The headline numbers below are the published programmatic rates — the starting point, not the negotiated outcome.
| Provider | Commitment Construct | 1-Year Discount | 3-Year Discount |
|---|---|---|---|
| AWS | Compute Savings Plan | up to ~54% | up to 66% |
| AWS | EC2 Instance Savings Plan / Standard RI | up to ~40% | up to 72% |
| Google Cloud | Spend-based CUD (Compute) | ~25–37% | ~52–55% |
| Google Cloud | Flexible CUD | ~28% | ~46% |
| Azure | MACC + committed-use / reservations | ~10–20% | ~20–40% |
Azure's committed model behaves differently from the other two. A MACC is a dollar commitment — you pledge to consume, for example, $1 million of Azure per year — and the discount comes from a blend of reservations, Azure Hybrid Benefit, and any negotiated private rate rather than a single published percentage. Microsoft now steers most heavy committed Azure spend onto the evergreen Microsoft Customer Agreement rather than the legacy Enterprise Agreement, a shift covered in our enterprise software pricing benchmarks guide.
When Pay-as-You-Go Wins
Pay-as-you-go is the correct choice more often than FinOps dashboards suggest. It wins decisively for short-lived, spiky, or uncertain workloads: a three-month migration project, a seasonal retail burst, a proof-of-concept that may be cancelled, or any environment whose architecture is still changing. Committing a one or three-year reservation to a workload you might re-platform in six months locks you into the wrong instance family at the wrong rate.
PAYG also wins as a deliberate buffer. Even enterprises with deep commitment coverage keep a slice of demand on demand pricing so that genuine growth, acquisitions, or unplanned projects do not trigger over-commitment penalties or force a mid-term renegotiation from a weak position. The same logic appears in the broader subscription debate we cover in pay-per-use vs subscription: flexibility has a measurable option value, and that value rises with demand volatility.
When an Enterprise Agreement Wins
Commitment pricing wins wherever demand is predictable and persistent. Steady-state production estates — the databases, application tiers, and platform services that run every hour of every day for years — are the natural home for three-year commitments, where the 52–72% discount bands compound into the largest line-item savings in the cloud bill.
An Enterprise Agreement also unlocks commercial terms that PAYG never touches: marketplace drawdown against the committed amount, custom private pricing, funding for migration and training, and access to the named account team that negotiates exceptions. For organisations spending more than $1 million a year, the EA or MACC is less about the headline discount and more about the negotiated envelope around it. We compare the programme structures across vendors in volume licensing programs compared.
The most common cloud overspend we see is not paying list price — it is committing 100% of forecast demand to a three-year term, then watching 20% of that capacity sit idle as architectures change. Under-committing is recoverable; over-committing is not.
The Hybrid Coverage Model
The answer is rarely all-commitment or all-PAYG. Mature cloud buyers run a layered coverage model: a base layer of three-year commitments sized to the floor of historical usage, a middle layer of one-year commitments for demand that is stable but evolving, and a top layer of pay-as-you-go for the variable remainder. Most enterprises target 70–85% commitment coverage of steady-state compute and leave 15–30% flexible.
This structure captures the bulk of the discount while preserving the option value of flexibility. It also turns the renewal into a manageable, rolling event rather than a single cliff: with staggered one and three-year commitments expiring on different dates, you are never forced to re-commit your entire estate under deadline pressure — the same trap that erodes leverage in software renewals, as we explain in software pricing models explained.
Negotiating Beyond the Published Tiers
The published Savings Plan, CUD, and reservation rates are the floor. Above roughly $1 million in annual spend, a private pricing agreement layers an additional negotiated discount on top — AWS through the Enterprise Discount Program, Azure through custom discounting attached to the MACC, and Google Cloud through negotiated commitment contracts. These private rates are where the real differentiation sits, and they are never volunteered.
The levers that move them are familiar: a credible multi-cloud alternative, a larger or longer commitment, alignment of the deal to the provider's quarter or fiscal year-end, and a documented benchmark of what comparable enterprises pay. Treat the cloud contract like any other major renewal — start 9 to 12 months out, benchmark before you negotiate, and never accept the first private-pricing offer. To pressure-test your committed cloud spend before you sign, request a confidential briefing, or download our Cloud Contract Framework for the full negotiation checklist.