Every cloud bill contains a line item most finance leaders have never questioned: data transfer out, or “egress.” For years it sat at roughly five to nine cents per gigabyte — small enough to ignore on any single invoice, large enough in aggregate to make moving a workload somewhere else economically painful. That was the point. The egress toll was never really about the cost of moving bytes; it was a switching cost, and switching costs are how platforms keep customers who might otherwise leave.
That toll is now being dismantled, and the change is more consequential for finance than for engineering. When the cost of leaving falls, the balance of power in every renewal conversation moves toward the buyer. Understanding why the shift is happening — and how much of it is real versus cosmetic — is now part of the CFO’s job.
Why the toll existed in the first place
Egress was priced as friction, not as cost. The marginal expense of transmitting a gigabyte across a provider’s network is a fraction of a cent. The five-to-nine-cent list price reflected a strategic decision: make ingress free so data flows in easily, and price egress so that same data becomes expensive to extract. Multiply that asymmetry across petabytes of accumulated storage and the effect is a soft cage. A workload can be moved in theory, but the one-time cost of copying its data out — plus the ongoing egress of any architecture that spans two providers — makes the business case for switching collapse before it starts.
Finance felt the effect without seeing the cause. When an engineering team proposed consolidating vendors or negotiating harder at renewal, the hidden counterweight was always the same: the incumbent knew the data was hard to move. That knowledge shaped every discount the provider was willing to offer, because it never had to price against a credible threat of departure.
What actually changed
Competitive pressure came first. Beginning in 2024, the major hyperscalers introduced programs allowing customers who are leaving the platform entirely to transfer their data out without the usual egress charge, typically as a credit granted through a support request. The scope matters: this is free egress on exit, not free egress in general. Day-to-day cross-cloud data transfer inside a running architecture is still metered at list price.
Regulation followed and went further. The European Union’s Data Act established obligations for providers of data-processing services to remove the commercial and technical barriers to switching. In practice this means switching charges are being reduced during a transition window and then withdrawn entirely, alongside requirements to support functional equivalence and portability. Although the Data Act is European law, providers tend to operate global pricing and contract templates, so the effects ripple outward to customers everywhere.
The direction is unambiguous even where the details are messy. Not every provider term is generous, the credit processes can be bureaucratic, and “functional equivalence” is doing a lot of quiet work in the fine print. But the trend line only points one way: the cost of leaving a cloud is falling toward zero, and it is not coming back.
The strategic value is leverage, not the refund
The one-time saving is the least interesting part. If your organization is not actively migrating, the elimination of exit egress fees produces no line-item change this quarter. The temptation is therefore to file this under “interesting but not urgent.” That would be a mistake, because the real value is not a refund — it is a permanent change to your negotiating position.
A credible exit is worth more than a discount. Procurement leverage has always been a function of alternatives. When the incumbent believed migration was economically impractical, its discount ceiling was low. When migration becomes genuinely feasible, every enterprise agreement, every committed-spend renewal, and every custom pricing conversation is repriced in your favor — whether or not you ever intend to move. The most valuable exit is the one you can execute but never need to.
This is a finance asset, so finance should hold it. Engineering teams rarely frame portability as commercial leverage; they frame it as an architecture problem. The instinct to build for portability lives in the technical org, but the reason it is worth paying for lives on the finance side of the house. Someone in finance should be able to answer, in a renewal meeting, exactly what it would cost and how long it would take to leave.
The trap: manufacturing lock-in above the infrastructure layer
The toll is moving up the stack. As raw data-transfer lock-in weakens, the incentive to create stickiness elsewhere strengthens. Proprietary managed services — bespoke databases, serverless runtimes, machine-learning platforms, and integration glue that has no equivalent elsewhere — are the new switching cost. The bytes may leave freely, but the application rewritten around a single vendor’s proprietary primitives cannot.
Portability is now an architectural-spend question. Every decision to adopt a deeply proprietary managed service is, in financial terms, a decision to re-accept lock-in that regulation just removed at the data layer. That is sometimes the right call — proprietary services can be genuinely cheaper to operate than the portable alternative. But it should be a decision, made with eyes open to the leverage being surrendered, not a default that engineering reaches for because it is convenient.
Track your lock-in surface deliberately. A simple internal measure — what share of annual spend flows to services with no cross-cloud equivalent — turns an abstract worry into a number you can watch. If that share is climbing, the freedom the egress change just handed you is quietly leaking back out through the application layer.
What finance should actually do
Quantify your exit today. Ask for a current estimate of the one-time cost and elapsed time to migrate your two or three largest workloads to an alternative provider. You are not commissioning a migration; you are pricing an option. The number itself, refreshed annually, is the deliverable.
Bring the option to the renewal. Committed-spend agreements are negotiated against the provider’s belief about your alternatives. A concrete, costed exit plan changes that belief. It does not need to be brandished aggressively — its mere existence changes the shape of the conversation.
Guard the lock-in surface. Add a lightweight review step for adoption of deeply proprietary managed services above a spend threshold, and report the proprietary-spend share to the same forum that reviews cloud cost. The goal is not to ban proprietary services; it is to make sure the switching leverage you were just handed is not spent without anyone noticing.
Watch the fine print on “exit.” Free exit egress and reduced switching charges come with conditions — eligibility windows, credit-request processes, definitions of what counts as leaving. Read them before you need them. An option you cannot actually exercise when the moment comes is worth nothing at the table.
The egress toll was one of the most effective lock-in mechanisms ever built into an operating expense, precisely because it hid in a line item no one questioned. Its unwinding hands finance a durable piece of leverage. The organizations that benefit will be the ones that recognize the shift for what it is — a negotiating asset — rather than waiting for a refund that, for most of them, will never come.
CostDefender gives finance a read-only, service-level view of where AWS spend concentrates — the raw input the lock-in-surface exercise in this piece depends on — so the cost of staying is grounded in real numbers before the next renewal.