Skip to content

One card or three vendors, the real cost of an integration programme

Comparing security hardware on unit price misses most of the cost. The real figure includes slots, vendors, driver stacks, integration programmes, boundaries crossed and certification packages maintained — and those scale with the number of separate components rather than with their price.

Counting the same capability two ways

Take a system needing an entropy source, a root of trust with key storage, and cryptographic acceleration. Assembled from three specialist vendors it is three parts, three driver stacks, three support relationships, three qualification efforts and three certification scopes to keep current.

The same capability in one boundary is one part, one driver, one relationship, one qualification and one scope. The unit prices may favour the three parts. The programme cost rarely does.

The costs that do not appear on a quote

  • Integration engineering. Each component needs a driver, a test plan and a place in the boot sequence. This is the dominant cost in most programmes and it is paid in schedule as well as money.
  • Boundaries crossed. Every interface between security components is a place key material or entropy moves, and each has to be protected and justified to an evaluator.
  • Certification scope. More components mean a larger evaluation boundary or several, and both cost more than one.
  • Ongoing maintenance. Firmware updates, errata and end-of-life notices arrive per vendor, and each has to be assessed against your evaluation.
  • Board area and slots. Real in embedded designs and in dense server deployments, where a slot has an opportunity cost.

Where the arithmetic flips

At low volume with an existing integration, three parts often win: the incremental cost is one more component in a design that already exists.

At volume, or on a new design, or where evaluation is required, consolidation wins — and it wins by more than the component prices suggest, because the costs that scale with component count dominate.

The honest caveat

Consolidation is a single-vendor dependency. That is a real risk and it should be priced: second sourcing, escrow, roadmap commitments and end-of-life terms are legitimate questions to ask us and anyone else.

The argument is not that fewer components are always right. It is that the comparison is usually made on the one number that matters least. The chiplet is this argument taken to its conclusion.