A multi-year commitment was signed weeks before a re-platforming halved the fleet it covered - what are your realistic options now?

## Establish what was actually bought A "commitment" is not one thing, and the recovery options differ so much by instrument that answering before you know which one you hold is guesswork. The axis that matters most: - **Tied to a shape.** The commitment covers a particular machine profile, size or family, in a particular place. Usage only draws it down if it matches. When the covered fleet goes away, nothing consumes it. - **Expressed as spend or usage per hour.** The commitment covers an amount of consumption, and any qualifying usage anywhere in the account or organisation draws it down. The replacement architecture may consume it without anybody doing anything. Providers differ on which forms they sell, on how wide the matching rules are, and on whether unused commitment in one part of an organisation can be consumed by another. Find out before you plan. ## The recovery routes, in the order worth trying 1. **Redirect consumption.** If the instrument is flexible, look for other steady workloads that could move onto the covered shape or draw down the covered spend - a neighbouring service, a batch tier currently on metered capacity, the replacement platform itself. This is the only route that is usually available immediately and usually costs nothing. 2. **Exchange.** Some platforms let a commitment be exchanged into a different shape, typically with conditions on term length and total value. Where it exists it is the cleanest fix for a shape-tied instrument; it is not universal. 3. **Resale or transfer.** Some providers run a route for reselling or transferring certain instruments. It applies only to some instruments, usually at a discount to what you paid, and it should never be assumed as part of the original purchase decision. 4. **Renegotiation.** Where the relationship is large enough that terms are negotiated rather than clicked, a stranded commitment can sometimes be rolled into a larger or longer arrangement. This trades a known loss for a bigger promise, so it is a decision to take deliberately rather than as a rescue. 5. **Accept it and sequence around it.** If none of the above applies, the term is sunk cost. The correct response is to schedule the migration so that the covered fleet's retirement lands as close as possible to term end, and to keep the remainder visible in planning so nobody rediscovers it as a surprise at renewal. ## The decision that is actually being tested The trap in this scenario is the sunk-cost one: **a remaining commitment is not a reason to keep running the old architecture.** The money is spent either way. The only comparison that matters is the cost of continuing to run the old fleet against the cost of running the new one, with the commitment excluded from both sides because it bills regardless. Where the covered shape can still be used productively, using it is a genuine saving; where using it means delaying a migration that pays for itself, the commitment has simply become a fixed cost and should be treated as one. ## What changes next time The durable fix is not a better negotiation, it is a policy about how much of the bill may be non-cancellable: - **Match term to confidence horizon.** Sign the longest term only for capacity whose existence is credible over that whole term. Discount depth is the wrong axis to choose a term on. - **Ladder, do not step.** Cover the confident part long, a further slice short, and leave a deliberately uncommitted band on metered capacity. The band is what absorbs a roadmap change. - **Prefer the flexible instrument** where the provider sells one and the price difference is small, because it survives a change of shape. - **Put a roadmap gate on the purchase.** A commitment decision should require an explicit statement from whoever owns the covered workload that no change is planned inside the term - in writing, at the time. - **Review at a fixed cadence**, not only at renewal, so a divergence between the covered shape and the running fleet is found in the quarter it starts rather than at the end. ## Saying it at the right altitude At a lead level the interviewer is listening for risk framing rather than a trick. The honest summary is that a term commitment converts a variable cost into a fixed one in exchange for a discount, that the discount is payment for the risk you took, and that the organisation's job is to decide how much fixed cost it wants and to keep that decision synchronised with the roadmap. A stranded commitment is what that risk looks like when it lands, and it is recoverable only partially, only on some platforms, and never reliably.