Anyone developing a device right now no longer plans around the most elegant circuit. They plan around what will actually be purchasable in twelve months. The semiconductor shortage has reversed the order of work in our projects, and to be honest, the planning has not suffered for it.
Where the shortage really bites
The pain rarely sits where you expect it. The processor is usually not the problem, because everyone watches it from day one. It is the small parts: a particular voltage regulator, a radio module, an interface chip worth a few cents. A component that was always in stock for three years suddenly carries a 52-week lead time, and the whole board is dead.
The nastier part is the timing. You find out late. During development the sample sits happily on the bench. Then the first small production run comes around and purchasing calls.
A second source beats a favourite chip
The engineer's instinct is to hunt for the perfect part. Exactly the right ratings, a good datasheet, a pleasant toolchain. That instinct is one we are all having to unlearn.
These days we select so that every critical function has at least two parts we could populate. Not identical, but interchangeable. That costs headroom in the design, because you have to accommodate the weaker of the two. The headroom is cheaper than a redesign under deadline pressure.
Check availability before the layout exists
Availability checks used to happen just before procurement. Now they happen before the schematic. Before a part enters the bill of materials we look at how many distributors carry it, how deep the stock runs, whether the manufacturer has announced a successor and how long it intends to keep producing it.
That half hour per critical component is the best money in the project. A layout is quick to draw and expensive to change.
Design for swapping
A footprint that accepts two package variants. A pad set that serves both pinouts. A module that plugs in rather than being soldered down, so it can be replaced without touching the board. A few unpopulated resistor positions to reconfigure between variants.
None of this is pretty. It is a compromise and it costs board area. But when one regulator vanishes in autumn and the other is available, you populate the alternative and keep shipping instead of losing four months. That is how we build devices in our IoT lab in the Tegernsee valley, where we work mostly on small production runs and rarely have the volumes that buy priority at a manufacturer.
Lead times belong in the schedule
A 40-week lead time is not a line in the risk register. It is the schedule. If you order in February and populate boards in December, engineering has to know in spring what it will be soldering in December. Component decisions therefore land earlier than they used to, and they become binding earlier. If you do not explain that to your customer, you are promising a date the supply chain cannot deliver.
Stockpiling: opportunity and trap
Purchasing pushes for buffer stock in times like these, and often they are right. If you buy two years of demand you can ship while your competitor waits. That is a genuine advantage.
It is also capital sitting on a shelf, ageing. And it locks in a design decision. Nobody with 5,000 regulators in the warehouse will change that circuit again, even when there are good reasons to. So we recommend buffering a small number of genuinely critical positions rather than the whole bill of materials.
The prototype you cannot build
The most expensive self-deception of these months is a working prototype built on a part nobody can buy any more. It looks like progress. Everyone is happy, the customer sees a device, the device does what it should. Then it turns out the sample was assembled from the last few pieces in the lab drawer.
So we now build samples deliberately with parts that are realistic for production, even when a better alternative is sitting in the cupboard. A prototype is meant to prove the production run. Not itself.
Frequently asked questions
How long will the shortage last?
Nobody can say that honestly. The new fabs that have been announced take years to produce anything. So we do not plan on relief arriving soon. We plan a component selection that survives if it does not.
Is it worth starting a new development at all right now?
Yes, if you put the component question at the front. Whoever develops now and designs for availability will be able to ship sooner than whoever waits and starts later.
What about parts from brokers?
Possible, but handle with care. Price and origin are often hard to verify, and counterfeit or desoldered parts are a real problem. For a production run, the supply route has to hold up.
If you are planning a device and the bill of materials worries you, we will go through it with you before the board is drawn. A conversation, with no obligation and no sales pressure.
More on our work with devices and small production runs: Hardware development.