The Bridge and the Redesign - How to manage parts obsolesce

MANAGEMENT BRIEF: We discuss how to manage parts obsolescence. Companion to the Rationed by the Boom.

From Reaction to Platform Transition Instrument

“Rationed by the Boom” established the problem: AI-driven wafer allocation has sunset the memory standards beneath long-life OT platforms while those platforms sit at peak deployment, and the standard mitigations — last-time buys, qualified crosses, redesigns — each carry costs the generic playbooks understate. This brief addresses the question that follows: how the mitigations should relate to each other.

The conventional framing treats the last-time buy and the redesign as competing alternatives — stock the shelf or change the platform, with a spreadsheet deciding between them. That framing is wrong, and in the current market it is expensively wrong. For any platform with meaningful service life remaining, the LTB and the redesign are not alternatives. They are two halves of a single coupled system: the last-time buy is the financing instrument for the platform transition. Procurement buys the redesign its runway; the redesign schedule defines procurement's buy quantity. Neither number can be set alone.

The reason the coupling works — in either direction the AI cycle breaks — is structural. If fab pressure persists, the redesign onto DDR5-native or integrated-memory platforms escapes a segment that will remain rationed through the decade. If AI capital expenditure contracts instead, the relief arrives in the active product lines — DDR5, LPDDR5 — while the decommissioned legacy nodes stay decommissioned: no memory maker reinvests in a retired process to serve a shrinking installed base at post-correction prices. A capex bust makes the destination cheaper; it does not resurrect the origin. The redesign therefore pays in both branches of the scenario tree, and only its magnitude of payoff varies. The genuine exposure is not choosing wrongly between LTB and redesign — it is running them uncoupled.

Sizing the Bridge to the Schedule

Once the coupling is accepted, the bridge sizing changes character. The textbook LTB formula — remaining product lifecycle years, times annual consumption, times a 1.2–1.5 safety factor — assumes you are stocking to sunset. The bridge model replaces the lifecycle horizon with the transition horizon:

•    Bridge quantity = consumption across the redesign duration (18–24 months), plus qualification and ramp (3–6 months), plus an explicit slip allowance — a 24-to-30-month horizon in total — plus a separately calculated service-stock tail for the fielded legacy base.

•    What this buys down: the stock-to-sunset objections. A schedule-bounded buy is smaller, easier to defend to finance, less exposed to shelf-life limits, and less likely to strand capital in depreciating inventory if the market loosens.

•    What it demands in return: the buy must be executed inside the LTB window — typically 90–180 days — against lead times now exceeding 30 weeks. The bridge is built once, at a moment the supplier chooses. A redesign decision that has not been made by the time the discontinuation notice arrives is a bridge that cannot be sized.

One boundary condition from the companion piece bears repeating: this logic holds where the redesign amortizes. Platforms within five to seven years of retirement never justify the transition in either market scenario; they are pure bridge-and-sunset cases, sized on the classical formula. The portfolio answer is a split — redesign the platforms with runway, bridge the ones without — with the AI cycle determining only how expensive the bridges turn out to be.

Governing the Burn-Down

Coupling the two systems creates a new obligation. Once inventory underwrites a schedule, redesign slip is no longer only a program-management problem — it is a supply-continuity event. Every quarter of NPD delay consumes bridge stock with no replenishment path behind it. An organization that sizes the bridge and then manages the redesign on ordinary program discipline has quietly converted schedule risk into line-down risk without telling anyone.

Once the last-time buy underwrites the redesign schedule, every quarter of program slip is consumed in silicon that cannot be repurchased.

The governance that fits is a burn-down structure, not a status report. Inventory-months-remaining becomes a first-class program metric, reviewed alongside schedule and budget, with pre-agreed thresholds that escalate to program leadership before options expire. Behind each threshold sits a pre-negotiated contingency, decided in calm conditions rather than improvised in crisis: a secondary buy through vetted authorized-independent channels, a scope reduction that pulls the memory migration forward of other redesign content, or a partial-fleet transition that stretches the remaining bridge across the platforms that need it most. The test of the governance is simple — at any review, the program can state how many months of bridge remain, what event triggers the next decision, and what that decision already is.

The Recurring Cadence

The final adjustment is to stop treating this as a crossing and start treating it as a rhythm. The structural fact beneath both papers is that component availability horizons — now 5–10 years for active devices, and shrinking — will remain shorter than OT field lives indefinitely. Wafer economics guarantee it, with or without the AI cycle that made it acute. Every long-life platform therefore carries, from the day it ships, a planned successor redesign and a planned bridge to reach it: availability engineering as a standing lifecycle discipline, resourced and scheduled like safety engineering, rather than an incident response mounted when the discontinuation notice lands.

The organizations that operate this way exhibit a recognizable operating signature: obsolescence buys that reference an NPD roadmap; NPD roadmaps that reference a component-availability forecast; and a single owner accountable for the coupling. The ones that do not exhibit an equally recognizable signature — procurement and engineering surprising each other, serially, with each function's rational local decision creating the other's emergency. Which signature an organization carries is discoverable in an afternoon of the right questions, and changing it is a matter of operating structure.