Cloud price comparisons usually go wrong in the same way: they line up one instance type against another and declare a winner. But nobody runs an instance in isolation. A real workload is compute plus storage plus data transfer plus a managed database plus the commitment you did or did not buy, and the cheapest instance often sits inside the most expensive total. The comparison worth doing prices the entire shape, a web app with Postgres, a Kubernetes cluster, LLM inference on four A100s, across all four clouds together, and ranks the monthly totals with the commitment scenarios alongside.
Start from the workload, not the SKU
Describe what you actually run. A workload template captures the moving parts: the compute class and count, the storage type and volume, the database, the egress profile. From there a comparison can be like for like, because it prices the same requirement on each cloud rather than comparing a product name on one against a product name on another. This is also where placement decisions get made honestly, since the ranking can flip entirely once storage and egress enter the total.
The equivalence mapping problem
Clouds do not sell identical products, which is the quiet reason most comparisons are dishonest. An AWS service has a near cousin on Azure, a different cousin on GCP, and sometimes nothing comparable on OCI, or the reverse. A credible comparison publishes its equivalence mapping openly, this thing is treated as comparable to that thing, and it states the gaps rather than papering over them. When a cloud has no true equivalent for part of your workload, the honest answer is to say so, not to substitute a rough match and pretend the totals are exact. That transparency is what lets you trust the ranking, because you can inspect the assumptions behind it.
Read the total, then the commitment scenarios
Two totals matter, and they can point at different clouds. On demand tells you where the workload lands if you commit to nothing, which is the right number for spiky or short lived work. The commitment scenario tells you where it lands once you buy the discount instrument each cloud offers, Savings Plans and Reserved Instances on AWS, Reservations and the savings plan on Azure, Committed Use Discounts on GCP, and Universal Credits on OCI, at coverage sized to a forecast. A cloud that looks expensive on demand can win once a sensible commitment is applied, and a cloud that looks cheap can lose once egress is counted. The point of seeing both is to place the workload on economics rather than on a headline rate.
A data heavy service looked cheapest on the provider with the lowest compute rate, until egress entered the total. The same workload, priced whole, landed materially cheaper on a provider with pricier compute but far lower transfer costs. The instance comparison would have sent it to the wrong cloud and quietly added to the bill every month.
Comparison is a decision input, not the decision
A ranked total is where placement analysis starts, not where it ends. Data gravity, the cost and risk of moving, contractual commitments already in place, and the operational tax of running one more cloud all bear on the call. A comparison that is honest about its equivalence mapping gives you a defensible starting number; the judgment about lock in and migration cost is yours to make, ideally with someone who has no incentive in the answer. For the wider tradeoffs, see when multicloud saves money and when it costs.
Frequently asked questions
Why not just compare instance prices across clouds?
What is an equivalence mapping?
Should I place workloads on the cheapest cloud?
Price your own workload across four clouds
The comparison tool is free and public, and every figure traces to a dated catalog. Compare a preset or describe your own shape, then tour the platform behind it.
Put a defensible number on your cloud spend.
No provider in the room, no published price list. Tell us your footprint and we will scope the savings against your billing data — we reduce your cloud spend or we reimburse our service fee.
The Cloud Spend Navigator: what changed in cloud pricing, commitments, and FinOps — no vendor spin.