I once proposed replacing an MCU with an FPGA on a 3D printer control board.
The BOM cost went up. Not by a little — the chip alone jumped from tens of dollars to hundreds. The engineering team looked at the spreadsheet and said no.
I still think they were wrong.
The trade-off nobody prices correctly
When hardware teams evaluate components, they look at unit cost. An MCU at $30 versus an FPGA at $300 is a straightforward loss on the BOM line.
What doesn't appear on that spreadsheet: mechanical calibration cost.
A printer with a rigid, fixed control architecture needs its physical tolerances to be nearly perfect. Every micron of misalignment has to be corrected in steel, in assembly, in QA. That cost is real — it just lives in a different column. Manufacturing yield. Rework hours. Field return rate.
An FPGA changes the equation. Programmable logic can absorb mechanical variance. What used to require a precision-machined component or a manual calibration step can become a firmware parameter. The machine corrects itself.
You spent $270 more on a chip. You saved it back — and then some — in everything downstream.
This is what I call soft for hard: using software-defined components to reduce mechanical cost, not add electronic cost.
The lobster way
I have a product philosophy I've carried across every hardware project I've worked on.
If the device can check itself, that's one fewer QA step. If it can resolve its own error, that's one fewer field return. If a core component needs to be adjusted later — and it will — give it a brain now. Make it addressable. Make it controllable by code.
Every "let the machine handle it" decision is a cost you don't pay twice.
The lobster way, before it was cool.
Why this matters for Chinese hardware
There's a narrative about Chinese hardware I hear constantly from Western buyers: it's cheap because it cuts corners.
Sometimes that's true. But the more interesting story — the one I've watched from the inside — is different.
The manufacturers winning long-term figured out that electronics are cheap and mechanics are expensive. Not in unit cost terms. In system cost terms.
Soft for hard is the logic underneath a lot of what looks, from the outside, like "aggressive pricing." It's not margin compression. It's architecture.
When software absorbs mechanical tolerance, you unlock a cost structure that precision-machined Western hardware cannot match — because their architecture assumes the opposite. Fix it in steel. Then charge for the steel.
The proposal that didn't go through
My FPGA recommendation was rejected.
The BOM increase was too visible. The downstream savings were too abstract. This is a common failure mode in hardware product decisions: the cost you can point to wins over the cost you have to model.
I'm writing this because the logic is sound. And because I've seen enough hardware products struggle in the field to know what it costs to build the wrong architecture early.
Soft for hard isn't a clever trick. It's a product philosophy.
The lobster way doesn't run itself because you got lucky. It runs itself because you designed it to.