Cloud exit strategy: negotiating out of AWS, Azure or GCP.
Exit is rarely about leaving. It is about holding a credible, priced alternative that resets the terms of every renewal. This note breaks down where lock-in actually originates, what an exit truly costs, the exit clauses to secure before you sign, and how to convert exit optionality into discount.
You almost never need to actually leave — you need to make leaving credible and priced. The commitment penalty, not egress, is the largest exit cost, and every meaningful protection is negotiable only before you sign. Buyers who run 20–30% of spend on a genuine second platform earn 3–8 points of additional discount from the incumbent.
01 Key findings
Commitment is an obligation, not a subscription. AWS EDP, Azure MACC and Google Cloud commitments bill the full committed amount regardless of consumption. A $20M three-year deal exited after 12 months still owes roughly $6.7M plus any penalties.
The penalty dwarfs the egress bill. Egress at $0.02–$0.05/GB reaches $50,000–$500,000+ for data-heavy estates — real, but small against a seven- or eight-figure commitment shortfall. Model exit as a commitment problem first.
Architecture, not contract, is the deepest lock-in. Managed databases, serverless and proprietary APIs turn "exit" into "rebuild." Buyers who never planned for portability discover refactoring costs that exceed the migration itself.
Exit protection is a pre-signature good. Data-portability windows, egress waivers, transition support and no-acceleration clauses are freely negotiable before signing and nearly impossible to win after.
Credible alternatives price better than loud ones. Vendors distinguish genuine multi-cloud investment from theatre. A real 20–30% second-provider footprint is worth 3–8 discount points; a bluff is worth nothing.
02 Sources of lock-in
Exit friction is engineered, not accidental. Lock-in accumulates across three layers — commercial, data-gravity and architectural — and each demands a different countermeasure. Diagnose which layer binds you before negotiating.
| Lock-in source | Mechanism | Typical exit drag | Countermeasure |
|---|---|---|---|
| Committed spend | EDP / MACC / commitment bills full amount on early exit | Largest single cost — $5M–$50M+ | No-acceleration & early-termination clauses |
| Data gravity (egress) | $0.02–$0.05/GB to move data out of the network | $50K–$500K+ at 10–50TB scale | Portability window & egress waivers |
| Managed services | Proprietary databases, queues, IAM tied to platform | Weeks–months of re-architecture | Portable primitives; abstraction layers |
| Serverless & APIs | Functions and SDKs with no cross-cloud equivalent | Application refactoring, $250K–$2M+ | Contain proprietary surface at the edges |
03 The exit-cost anatomy
Four cost categories dominate any exit. Three are within your control; one — the commitment shortfall — is fixed the day you sign. First-10TB internet-egress list rates give a feel for the data-movement component, though negotiated waivers matter far more than headline price:
Buyers optimise the wrong number. Egress is visible and easy to model, so teams fixate on it — while the commitment shortfall, often 10–100x larger, is treated as fixed. It is not. Commitment relief, no-acceleration terms and migration credits are all negotiable, and moving even a fraction of the penalty is worth more than eliminating egress entirely. Model exit as a commitment problem with an egress footnote, never the reverse.
04 Exit clauses to secure
The only reliable moment to buy exit protection is before signing a multi-year commitment. Six clauses belong in every enterprise cloud contract; each is negotiable at signature and nearly unwinnable afterwards.
| Exit clause | What to demand | Why it matters |
|---|---|---|
| Data portability window | 90–180 days post-termination at reduced or zero egress | Prevents egress cost from exceeding exit savings |
| Egress waivers | Migration-purpose transfer waived to 50TB–500TB threshold | Removes the data-gravity toll on departure |
| Transition support | 30–60 days of contracted technical assistance | Documents help with extraction, APIs, troubleshooting |
| Early termination rights | Exit on price hikes, SLA misses or roadmap changes | Rare but decisive escape conditions |
| No commitment acceleration | Obligation assessed by time elapsed, not triggered on notice | Stops instant payment of the full remaining commitment |
| Competitive price match | Match a lower Azure or GCP offer received mid-term | Keeps leverage live throughout the term, not just at renewal |
05 Egress-on-exit and data portability
Extraction runs at standard egress pricing unless you have negotiated otherwise — AWS charges around $0.02/GB from US regions, so 1PB costs $20,000+ and 10PB reaches $200,000+. Providers offer no standard waiver during exit; it is billed like any commercial transfer. That makes the portability window the single most valuable exit clause.
Anchor the window to the migration, not the calendar. A 90–180 day window at zero or reduced egress, scoped to a defined volume threshold, converts an open-ended data-gravity tax into a bounded, plannable cost. Pair it with a documented transition-support commitment so extraction assistance is contractual rather than a goodwill favour that evaporates the moment you give notice.
06 Exit as leverage
Most exit value is captured without leaving. The threat of departure resets renewal terms — but only when the alternative is real. Vendors run sophisticated credibility checks and ignore performative exit talk.
Credibility requires evidence: architecture and cost modelling on a second cloud, a technical proof-of-concept, and workloads genuinely running elsewhere. Organisations holding 20–30% of spend on a second provider report 3–8 points of additional incumbent discount versus single-cloud peers at equivalent spend. The window matters too — exit positioning is most potent 18–24 months before renewal, early enough to show real progress, late enough to avoid signalling theatre.
On-premises is a credible alternative again. Andreessen Horowitz's 2023 analysis found stable, at-scale workloads run 2–3x cheaper on-premises than in cloud on a proper TCO basis, and repatriations like 37signals/Basecamp ($3.2M annual savings) and Apple's iCloud shift prove the option is real. Even with no intent to repatriate, an articulated on-premises path strengthens exit bargaining power — providers now price to keep the conversation from starting.
07 Exit framework
Four considerations shape any exit or exit-threat strategy. Weight them to your situation before opening a renegotiation.
Commitment exposure
Map exactly what you owe on early exit. The committed-spend shortfall, not egress, sets the true floor on exit cost — and determines whether no-acceleration relief is worth fighting for.
Architectural portability
Audit how much of the estate is bound to proprietary managed services. Portable primitives make exit a migration; deep serverless coupling makes it a rebuild with $250K–$2M+ of refactoring.
Alternative credibility
A second-provider or on-premises footprint of 20–30% of spend registers on the vendor's risk model. Below that, the threat is theatre and prices accordingly.
Timing to renewal
Exit positioning peaks 18–24 months out. Vendor pricing committees close in November/December — present competitive work before that window for maximum bargaining power.
08 Our recommendation
You are moving all workloads off the incumbent, usually to on-premises. Lead with committed-spend relief and aggressive egress waivers — the two costs that decide whether the exit pays for itself.
You move some workloads or business units while staying resident. Prioritise commitment-scaling flexibility and the right to reduce committed amounts mid-term without penalty acceleration.
You threaten a move between clouds. Secure commitment-transfer provisions and competitive price-match language — the re-platform threat generates as much leverage as a full exit at a fraction of the disruption.
09 Exit sequencing
The single highest-value process choice for exit leverage is when you start building it:
Build leverage first Recommended
Stand up genuine second-provider workloads and secure exit clauses 18–24 months before renewal. You arrive at the table with a priced, credible alternative and win 3–8 points of additional discount — whether or not you ever leave.
Negotiate at exit time Weaker
Wait until you want out, then ask for relief. Every protection is now a concession the vendor has no reason to grant, egress is billed at list, and the commitment accelerates. Leverage is at its lowest exactly when you need it most.
Plan the exit before you need it
Our Cloud & FinOps practice builds exit protection, commitment relief and multi-cloud leverage into your contract before signature.
The Licensing Edge
Weekly cloud and licensing intelligence for enterprise IT leaders. 3,000+ subscribers.