3 ms·
A 22-year-old builds 1970's-era chips with ~$1M capital in his garage. Intel builds 2020's-era chips with ~$20B capital in Ohio. Take the arithmetic mean of 1
by csense 5y ago
A 22-year-old builds 1970's-era chips with ~$1M capital in his garage.
Intel builds 2020's-era chips with ~$20B capital in Ohio.
Take the arithmetic mean of 1970 and 2020, and take the geometric mean of 1M and 20B. This very rough back-of-the-envelope calculation implies a small-ish company could build 1995-era chips with ~$141m capital.
It seems to me that $141m is a quite reasonable "insurance policy" price major automakers would be willing to pay to ensure a stable supply of microcontrollers with 1995-era specs, given that they were recently in a situation of being unable to sell a bunch of $35,000 trucks because they couldn't find a batch of $3.50 microcontrollers.
And this story repeats across a lot of different industries.
Why aren't people throwing money at quickly building cheap chip plants to churn out low-spec chips?
- fisherjeff 5y agoI mean, most MCUs are already 15+ years behind the state of the art process node and are still supply constrained. The issue, I think, is that once these market conditions are resolved, any new fabs will have ~zero value, and there is a non-zero chance that your shiny new-old fab comes online after the shortage is over. Not an easy sell to shareholders I suspect.