How many parts to buy on a last-time buy

The notice arrived. A chip your product depends on is going away, and the distributor is offering a last chance to order. Now someone has to type a quantity into a purchase order, and it cannot be undone. Buy too few and you stop shipping. Buy too many and you hold stock that ages, ties up cash and may never be used.

This note is about that number: how to build it, what moves it, and which parts of the notice you should read twice before you commit.

Read the notice before you do any math

A discontinuance notice is a contract-shaped document. The dates in it set the shape of your buy.

Take a real example. NXP’s PCN 202606009DN was issued in June 2026 and withdraws roughly a hundred orderable part numbers across several families, listed as full withdrawal, sole source, with no replacement part named. It gives two different dates: a last time buy date in October 2026 and a last time delivery date in July 2027. It also says that limited-availability products are sold first come, first served until inventory or manufacturing capacity runs out, so orders may be filled before the last time buy date.

Three things follow from that one notice, and they apply to any notice you receive:

• The order date and the delivery date are different clocks. You must order by the first. You may receive parts up to the second. Your cash and planning move on the order date.
• “Last time buy” is not a guarantee. If availability is limited and allocation is first come, first served, waiting until the deadline can mean getting nothing. Reading the notice early is part of the plan.
• The notice may be silent on cancellation. The NXP notice, as published, does not discuss order cancellation. Your terms live in the distributor’s or manufacturer’s order conditions, not in the PCN. Ask before you order whether a last-time-buy order can be changed, rescheduled or returned. Assume it cannot until someone tells you otherwise in writing.

Other vendors differ. A discontinuance announced by a microcontroller vendor may give only a final order date and say final deliveries continue through a following year. Treat every notice as its own document. For the broader question of what to do when a controller you rely on is withdrawn, see our note on microcontroller end-of-life.

The sizing formula, in words

The calculation has five parts. Keep each one as a separate line in a spreadsheet so you can argue with each assumption on its own.

1. Demand. Units you expect to build per year, times the number of years you need to bridge, times parts per unit. Be honest about the bridge. It is not “until we redesign.” It is the time until a redesigned product is built, validated, certified if needed, and actually replacing the old one in the field.
2. Process loss. Parts consumed by assembly that never ship: placement errors, reel-end remnants, damaged parts, test and rework losses, engineering builds. Use your own history. If you have none, say so and carry a visible allowance rather than a guess dressed up as data.
3. Service and warranty spares. Parts you will need to repair units already in the field. Products with long service lives need this line more than products that are replaced.
4. Margin. An explicit buffer for demand you cannot predict. Make it a stated decision, not a hidden fudge in the other lines.
5. Minus what you already hold. Stock on the shelf, stock at your assembler, parts on open orders, and parts sitting on boards in finished goods you could rework.

Line	What goes in it	Common mistake
Demand	Units per year × bridge years × parts per unit	Using the redesign start date as the end of the bridge
Process loss	Your own scrap and rework history	Forgetting engineering and test builds
Spares	Field repair coverage	Leaving it at zero
Margin	A stated buffer	Burying it in other lines
On hand	Shelf, assembler, open orders	Missing parts at the contract assembler

The result is a number, but the useful output is the range. Run it three ways: a plan where the redesign lands on time, one where it slips, and one where demand is higher than forecast. Then look at how far apart they are. If the gap is small, buy the middle and move on. If it is large, the argument is about the redesign schedule, not the purchase order.

Shelf life is part of the quantity

Parts you buy now are parts you will use over years. That raises a question the sizing sheet usually leaves out: will they still be good when you reach the last reel?

TI’s product shelf life page is a useful vendor statement of how this works. It says shelf life depends on the materials in the device, manufacturing conditions, moisture sensitivity level, whether moisture barrier bags are used, the amount of desiccant, and the customer’s storage conditions. It also says how a customer stores the product is critical. That is the vendor telling you the burden of storage is on you.

For a buy meant to last years, that means:

• Find the vendor’s own shelf-life statement for your part, not a rule of thumb from a forum. TI also points to per-part shelf-life information in its handling note, AN-2029. Other vendors publish their own storage guidance, and the numbers differ.
• Keep parts sealed. Moisture-sensitive packages stay in their barrier bags with desiccant and humidity indicator cards until the day they go on a line. Do not split reels into convenient quantities and reseal them casually.
• Check the indicator card when you open a bag, and plan for baking if it shows exposure. We cover that in when SMT parts that sat open need baking before reflow.
• Decide where the stock lives. A single shelf in one building is a single point of failure. A contract assembler holding your parts may store them under conditions you have never seen. Ask.

Shelf life also argues against buying the whole bridge in one lump if the vendor will allow staged deliveries. Not every notice does. When it does, deliveries spread across the delivery window cost you less in storage risk than one large drop.

Counterfeits and where you buy

A last-time buy is when gray-market sellers get busy. If the authorized channel runs out, parts appear from brokers at confident prices. Buying from them is a different risk from buying from the manufacturer or an authorized distributor, because provenance, storage history and authenticity become your problem.

Our advice is plain. Place the bulk of the buy through the authorized channel while it exists. If you must go beyond it, treat incoming parts as unproven: inspect them, test a sample electrically on a representative board, and keep records of where each lot came from. Do that before you commit them to a production run, not after.

Set the redesign trigger now

A last-time buy is a bridge, not a destination. The most common failure is quietly treating the buy as the solution and never starting the redesign. Put three things in writing on the day you place the order:

• A start date for the redesign, chosen so that the new design is in production before the bought stock runs out, with room for the slip scenario above.
• A consumption check. Compare actual use against the plan on a fixed schedule. If you are burning parts faster than forecast, the date to start the redesign moves earlier. Do not wait for the shelf to look empty.
• A candidate list. Start identifying replacement parts early, and check every candidate’s drawing and electrical requirements against your board. A part that looks like a drop-in often is not. Our note on checking a chip’s longevity commitment before you design it in covers how to avoid choosing a replacement that is itself near the end of its life.

Redesign work has real time cost: layout changes, firmware ports, re-validation and any compliance retesting your product needs. The buy exists to pay for that time, so size it against the time, not against optimism.

How we approach it

We start from the notice, your build plan and your current stock, work the sizing sheet with you, and scope the redesign alongside it so the buy has an end date. Send us the notice, the part numbers and your rough yearly volume.