Supply-Chain Criticality Hierarchy
Rank every component by how catastrophic losing it is — then pre-buy the killers.
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 3
- Confidence
- 90%
In hardware a single missing part stops the whole product, but not all shortages are equal. Kalinowski sorts components into a hierarchy: mechanical parts like die-cast pieces are recoverable in months, while silicon and RAM shortages force a catastrophic redesign of the product's guts. The strategic responses are to identify long-lead, hard-to-substitute parts up front, pre-buy constrained commodities like memory to ride out price spikes, and pursue vertical integration where feasible.
Origin
Caitlin Kalinowski; the memory-price warning was prompted by Matic CEO Mehul Nariyawala's 'meteor' framing. Tesla/Starlink verticalization cited as the exemplar.
Core principles
- 01One missing component stops everything, but recovery cost varies enormously by part type
- 02Die-cast/machined parts are recoverable in 3-5 months; silicon and RAM shortages are catastrophic redesigns
- 03Constrained commodities (memory) can only be paid for or pre-bought — there's no third option
- 04Vertical integration lets you absorb supply shocks by redesigning around what you can get
How to run it
- 1
Map components onto a recoverability hierarchy
For each part, ask how bad losing it is. Mechanical/die-cast parts: recoverable, re-source in ~3-5 months. Silicon or RAM in a specific form factor: catastrophic — you must redesign the board or the entire guts, re-secure supply, rebuild the line, and redo reliability testing. Track the longest-lead items (typically silicon and display) explicitly.
Watch out A chip or memory shortage isn't a delay, it's a full redesign cascade — treat those parts as strategic, not routine.
- 2
Pre-buy constrained commodities to ride out spikes
For components under demand pressure (Kalinowski's live example is memory/DRAM, driven by AI data-center demand), stock enough inventory to survive price spikes and shortages, if you can afford it. During COVID her team pre-bought memory for exactly this reason.
Pro tip Weigh the pre-buy against the risk that prices fall — but when a key commodity is constrained, your only real options are pay up or have already pre-bought.
Watch out Consumer-electronics makers get outbid by less cost-sensitive data-center buyers, so waiting for the market to soften can leave you unable to build at all.
- 3
Verticalize where a shock would otherwise be fatal
Bringing components in-house lets you adapt to shocks — Kalinowski cites Tesla redesigning its PCB in record time to accept different silicon during a shortage, which would have been catastrophic for a company with a classic outsourced chain.
In the wild
During COVID's supply-chain disruptions, getting enough memory was a core challenge, so the company Kalinowski was working with pre-bought memory to ride out the shortage. She now advises startups to hold enough memory in stock to survive the current AI-driven price spikes.
→ Continued production through a shortage that would otherwise have halted builds; the same playbook is her standing advice for today's memory 'meteor.'
When silicon was hard to find, Tesla — having verticalized much of its supply chain — redesigned its PCB in record time to adapt to newly available chips.
→ Absorbed a shock that would have been catastrophic for a company dependent on a classic outsourced supply chain.
Common mistakes
Treating all component shortages as equal delays
A missing die-cast part is a months-long re-source, but a missing chip or RAM form-factor forces a redesign of the entire product guts — mis-ranking them leaves you unprepared for the catastrophic ones.
Refusing to pre-buy a constrained commodity to save money
Less cost-sensitive buyers (like AI data centers) will pay the higher price and take the supply, leaving a consumer-hardware maker who waited unable to build at all.
Is it for you?
Best for
Hardware founders and ops leaders planning BOM risk for a product entering or scaling in a constrained-component market
Not ideal for
Low-volume or hobby builds where component shortages don't threaten a production line
From the transcript
“If your silicon goes out, if you can't buy your silicon, you can't buy your chip now.”
“you now have to redesign the entire guts of your product”
“advising startups and companies to pre-by memory”
From the episode
Why we’re at the beginning of the AI hardware boom
Caitlin Kalinowski (ex–OpenAI, Meta, Apple)