Cross-Cloud Workload Placement Cost: Where Each Workload Should Run

Ask why a given workload runs on the cloud it runs on, and the honest answer is usually inertia: it was created there, the team knew that provider, and nobody revisited the decision. Inertia is a fine default for most workloads — revisiting everything constantly is its own waste — but for large or fast-growing workloads, default placement leaves real money on the table. A disciplined cross-cloud placement model decides where each significant workload should run by weighing the four costs that actually determine the answer.

Across $2.4B+ in reviewed AWS spend and 500+ engagements, revisiting placement on the largest workloads typically surfaces 8–18% of avoidable spend — not by moving everything, but by correcting the few placements where inertia is expensive.

The four costs of placement

Placement is not a unit-price comparison. The right decision weighs four costs together:

• Unit cost. What the workload's compute, storage, and managed services cost per month on each candidate provider, at realistic commitment pricing — not rate card.
• Egress. The recurring cost of data the workload exchanges across provider boundaries given its placement. A workload separated from its data generates egress forever.
• Commitment burn. How placement affects spend that counts toward your discount commitments. Moving spend off AWS can drop you below a tier and cost discount on everything else.
• Switching cost. The one-time cost to move and re-platform, amortized over the workload's expected remaining life.

A placement that wins on unit cost but loses on the other three is the classic mistake — the same dynamic we dissect in our cloud cost arbitrage tactics guide.

The decision model

For each significant workload, score the candidate providers on total cost of placement:

Total placement cost = unit cost + recurring egress + commitment-burn impact + amortized switching cost

The provider with the lowest total — not the lowest unit cost — is the right placement. For most workloads already on AWS with data on AWS, the incumbent wins because egress and switching cost favor staying put. For a minority — stateless, portable, low-data-gravity workloads, or new workloads with no switching cost — an alternative provider can win cleanly. The model's job is to tell those two groups apart instead of treating every workload the same.

Workload archetypes

Stay-put workloads

Data-intensive workloads with large persistent stores, tightly integrated workloads that chatter with many AWS-native services, and stable production workloads deep in committed pricing almost always belong where they are. The egress and switching cost to move them exceeds any plausible unit-price gain, and moving them would erode commitment burn. Inertia is correct here.

Portable workloads

Stateless compute, batch processing, and workloads built on open, portable interfaces can be placed on cost merit because their switching and egress costs are low. These are the workloads where the placement model earns its keep — and, not coincidentally, the workloads that double as multi-cloud leverage because their portability is genuine.

Greenfield workloads

New workloads have no switching cost and no existing data gravity, so they can be placed optimally from day one. The cost of getting greenfield placement wrong compounds: a workload created on the wrong provider accumulates data gravity that makes it a stay-put workload before anyone revisits the decision. Greenfield placement deserves more scrutiny than it usually gets.

Egress is the hinge

In nearly every placement decision, egress is the factor that decides it. A workload's unit cost might be 20% lower elsewhere, but if placing it there means it exchanges data across the boundary continuously, recurring egress erases the gain and then some. Conversely, a workload that can be co-located with its data on the cheaper provider keeps the full saving. Getting placement right is largely about getting compute and data on the same side of the boundary — the discipline at the heart of our multi-cloud egress optimization guide.

Placement and commitment strategy

Placement decisions and commitment decisions are coupled. Every placement that moves spend changes the spend base that supports your discount tiers, so the two must be planned together. A placement that looks optimal in isolation can be value-destroying if it drops you below an EDP tier or creates a commitment shortfall. The right sequence is to set the commitment strategy first — which provider holds the deep baseline — then make placement decisions that respect it, keeping portable workloads available as leverage without undermining the committed core.

The evenhanded conclusion: placement is not about chasing the cheapest unit price for every workload. It is about correcting the few expensive defaults while leaving the many correct defaults alone, and doing it in a way that strengthens rather than fragments your commitment position.

Placement reviews as a cadence

Placement is not a one-time decision; workloads grow, data accumulates, and pricing changes, so a placement that was optimal at creation drifts over time. The discipline is a periodic placement review on the largest workloads — not constant churn, which is its own waste, but a deliberate annual or semi-annual look at whether the top spenders are still where they belong. The review aligns naturally with renewal cycles: the months before a major commitment renewal are exactly when you want fresh placement data, because placement and commitment decisions are made together.

The reviews also catch the slow conversion of portable workloads into stay-put ones. A workload that was genuinely movable at creation accumulates data gravity and integration dependencies that quietly raise its switching cost until, one review, it has crossed into stay-put territory. Knowing when that crossing happens matters, because a workload's value as leverage ends when its switching cost becomes prohibitive — and you would rather know that before a renewal than discover it during one.

The cost of getting greenfield wrong

Because switching cost compounds with data gravity, the most expensive placement mistakes are greenfield ones. A new workload placed on the wrong provider does not stay wrong cheaply — it accumulates data and dependencies that make the original mistake progressively more expensive to correct. The discipline is to give greenfield placement the four-cost analysis up front, when the switching cost is still zero, rather than defaulting to whichever provider the team knows best and paying for that convenience for years.

Placement, leverage, and the negotiating table

The strategic payoff of disciplined placement is not only the direct cost saving but the negotiating position it builds. An organization that can show, by workload, exactly what it costs to run each one on each provider — and that has genuinely portable workloads it could move — negotiates every renewal from evidence and credible optionality. That is the connection between placement discipline and the broader multi-cloud leverage that determines what you ultimately pay: placement is where the leverage is built, one workload at a time.

What to do this quarter

Rank workloads by spend and run the four-cost model on the top tier. Separate stay-put workloads from genuinely portable ones, and act only where inertia is costing real money. Give greenfield placement decisions the scrutiny they deserve, and sequence any moves against your commitment strategy so you do not trade unit-cost savings for lost discount.

If you would like an independent placement review across your estate, Contact Us.