Three sentences that reprice a portfolio
The vendor’s statement ran to three sentences. Elevated component costs; an average price increase of roughly forty percent across most of the portfolio; subscription customers on longer contracts limited to about ten percent. It was posted with one date, updated five days later “to reflect the final, implemented pricing”, and said nothing about outstanding quotes or orders in flight.
The component in question is flash memory, and the supplier side has stopped pretending capacity exists. The largest memory maker reported flash revenue up 169 percent year on year and told investors that demand now significantly exceeds available supply “for the foreseeable future”. Its gross margin, at over seventy percent, is the kind of number that in previous cycles belonged to software companies. New fab capacity arrives in 2027 and 2028. There is no short-term fix, and every storage vendor is buying from the same constrained pool, so this will not be the last reset of a price list.
For an organisation that budgeted a warehouse refresh, a data-platform migration or an AI storage tier on last year’s numbers, the plan is now out of date.
Read the spread, not the headline
The forty percent is the headline. The strategic content is the thirty-point spread between it and the subscription increase. A customer buying arrays outright absorbs two-fifths more; a customer extending a multi-year as-a-service term absorbs a tenth. That asymmetry converts a supply crisis into a funnel: revenue locked in for the vendor, and customers locked in who might otherwise have shopped the next refresh against three competitors — all of whom, again, buy from the same fab.
None of this makes the subscription the wrong choice. For many estates it is the right one. But it should be negotiated as what it is: a multi-year commitment made at the top of a price cycle, with the term, the capacity flexibility and the exit conditions priced against what a capex purchase will cost in 2028, not against today’s inflated list. And a customer holding a pre-increase quote should treat its validity as an open question for the account team, in writing, this week.
What to do with the plan
The instinct in a shortage is to buy before prices rise further. In our experience that is usually the second-best move. The best one is to find the capacity you already own.
- Audit the copies. Most estates carry multiple full copies of production data — replicas, test refreshes, analytics extracts, the sandbox nobody deleted. Each is capacity you are about to pay forty percent more to duplicate again. The inventory alone usually recovers a meaningful fraction of the planned purchase.
- Consolidate before you extend. Fragmented silos each carry their own protection overhead and their own idle headroom. A consolidation onto the governed platform frees both, and improves the governance while it does so.
- Size AI storage from measurement, not forecast. Vector indexes, retrieval caches and persisted model state all consume flash, and every vendor has a forecast for how much. Run the pilot, measure the actual footprint at your context lengths and reuse rates, and size from that. Buying AI storage at the peak of a memory cycle on a vendor’s projection is how a business case dies quietly.
- Re-base every assumption. Any cost model for modernisation or migration built before this quarter should be re-run with current pricing and a scenario for further increases through 2027. A plan that only works at last year’s prices is not a plan.
- Negotiate the term, not just the price. If the subscription route is right, the concessions to seek are capacity flexibility, a defined price path and an exit before the shortage ends — the things that will matter in 2028 when the vendor is the one who wants the renewal.
The cycle, and the estate
The memory business has swung between glut and shortage for forty years. This swing is out of scale with the earlier ones, which is exactly why the discipline matters more: decisions made at the peak have a long tail. The organisations that come through it well will be the ones that treated the shortage as a reason to understand their estate rather than a reason to expand it — and who signed the multi-year deal, if they signed it, on their own terms rather than the vendor’s timetable.
