Buy reserved capacity for blob storage

This is one of the cleaner savings levers in Azure: the capacity is already there, already being paid for at full rate, and a reservation simply re-prices the committed portion downwards. There is no migration, no engineering rework and no service risk. The only decision is how much of the steady baseline to commit and for how long, because the discount applies automatically to matching usage once the reservation is live.

The commitment is real but bounded and reversible. You pay either upfront or in equal monthly instalments at no extra cost, and a reservation can be exchanged for another Azure Storage reservation at any time with no fee, or refunded for a prorated credit on the unused term. Microsoft caps refunds at 50,000 US dollars per year, which is the one figure to keep in view at large scale. The downside of getting the size slightly wrong is therefore small and recoverable, not a stranded cost.

Frame the decision around the durable baseline, not peak usage. Reserve the capacity you are confident will still be there in twelve months, take the smaller discount of the one year term if the workload is young, and step up to three years only for storage you know is permanent. The savings are real recurring spend off the bill, the commitment is refundable within limits, and the payback is immediate because the discount starts the moment the reservation is purchased.

Priya is the finance partner working with the platform team on Azure run-rate. The Advisor recommendation lands with an estimated annual saving on the data lake, and the instinct is not to approve a three year commitment on the headline number; it is to check how durable the underlying capacity really is. A reservation discounts the steady baseline, so the question is how much of the current footprint has been present for the whole year and is certain to remain, versus how much is recent growth that might churn.

Priya asks the team to separate the durable baseline from the volatile top. The data lake has sat above a few hundred terabytes for over a year with no sign of shrinking, so that slice is a safe candidate for a longer, deeper-discounted commitment. Recent project data that could be archived or deleted is left on pay-as-you-go for now. The pack reads: 'We reserve the proven baseline at the three year rate for the deepest discount, hold the volatile top at pay-as-you-go, and revisit in a quarter. The commitment is refundable for a prorated credit if we are wrong, so the downside is bounded, and the discount starts the day we buy.'

The arithmetic here is unusually clean. The capacity already exists and is already billed; a reservation simply re-prices the committed slice downwards, so the saving is recurring spend removed from the bill with no offsetting cost beyond the commitment itself. The probability of realising it is high because the baseline is observable: you can see exactly how much storage has been present continuously before you commit a gigabyte.

The risk is bounded on every side. Payment can be upfront or in equal monthly instalments at no premium, so the commitment does not force a cash-flow hit. If the workload changes, the reservation can be exchanged for another Azure Storage reservation with no fee, or cancelled for a prorated refund on the remaining term, with Microsoft capping refunds at 50,000 US dollars per year. The finance role is to size the commitment to the proven baseline, choose the term that matches confidence in the workload, and track the captured discount as recurring savings rather than a one-off.

Azure Advisor is telling us we are paying full retail for a large block of blob storage that never goes away. A reservation discounts that same capacity in exchange for a one or three year commitment. No data moves, no service changes, the bill simply drops on the committed slice. This is a pure savings decision, not a risk one.

The leadership question is not 'is this safe?' It is 'how much of our steady storage baseline are we confident enough to commit, and for how long?' Committing capacity we will keep regardless is close to free money. Over-committing capacity we might shed is the only real risk, and even that is bounded: reservations can be exchanged at no fee and refunded for a prorated credit, subject to an annual refund cap.

The defensible position is that the durable baseline is reserved, the term matches our confidence in the workload, and the recurring discount is captured automatically. The only way to lose here is to keep paying retail for capacity we were always going to hold.

At a cost review the CFO noticed that storage was one of the larger recurring lines and asked the obvious question: if this much data is permanent, why are we paying the on-demand rate for it every month? The honest answer was that nobody had committed it, so it defaulted to retail pricing even though the baseline had not moved in a year.

The team's response was to treat the durable baseline as what it is, a permanent commitment, and price it accordingly. They confirmed with Advisor and their own usage history how much capacity had been present continuously, reserved that slice on a term that matched their confidence in the workload, and left the volatile top on pay-as-you-go so they kept the flexibility to shed it. The next review showed the same data, the same service, and a visibly lower storage line, with the reservation refundable within limits if the workload ever changed. The shift, from paying retail for storage we always intended to keep to committing it deliberately for a discount, is the saving this recommendation delivers, and it carries no operational risk.

This is the rare cost lever with upside and almost no downside. The saving is real, recurring and immediate, the capacity is one we intend to keep regardless, and the commitment is bounded and reversible. There is no migration, no service risk and no change to the data. The only way to lose is to keep paying retail for storage that was always going to be permanent.

The decision leadership owns is how much of the durable baseline to commit and for how long. Reserve the capacity you are confident will still be there, match the term to that confidence, and let the discount recur automatically. Even an over-sized reservation is recoverable through exchange or a prorated refund, so the worst realistic outcome is a small, bounded correction, not a stranded cost.