3 ms·

Yeah I guess I'm just surprised. It feels to me (who knows basically nothing about it) that it's been a pretty big problem for at least a few years, and it also

Yeah I guess I'm just surprised. It feels to me (who knows basically nothing about it) that it's been a pretty big problem for at least a few years, and it also seems like the combined buying power of the entire world's hardware companies would make the investment in new fabs trivially worth it. Maybe fabs take longer than a couple years to set up regardless of money spent? Maybe fabs take a level of expertise that only a handful of people in the world have, and it's a matter of ramping up promising undergrads to that level which will take like 15+ years? Maybe analysts expect demand to even out so that capital outlays right now don't make sense? But all of these are surprising to me to the degree that we still don't have enough semiconductors. I'm just hoping for a sort of comprehensive overview of the issue, if such a thing exists.

pjc50 5y ago
From my perspective in the software side of a semiconductor company, it's very much a problem of the last year. The AKM fab fire in November 2020 also made the situation worse.
- Leherenn 5y ago
  It's pretty much impossible to find GNSS chips currently, at least the ones we use (ublox). There's a massive TCXO shortage.
browningstreet 5y ago
Fabs also have to be manufactured.
akiselev 5y ago
> Maybe fabs take longer than a couple years to set up regardless of money spent? Maybe fabs take a level of expertise that only a handful of people in the world have, and it's a matter of ramping up promising undergrads to that level which will take like 15+ years? From the start of permitting to first tape out, I've heard five years for cutting edge, three to four for mid range (under 100nm), and two years for older processes. Subtract a year for expansions of existing facilities where hazmat permits and infrastructure already exist. Far more than a handful of people in the world are qualified to set up a fab but it really is hard, time consuming work that requires a lot of expertise. While most of the equipment in a fab is customized off the shelf (not totally custom but not cookie cutter), each piece has to be carefully calibrated to fit in with the whole. Stuff like HVAC, which is normally pretty predictable in commercial buildings, has to be custom designed for seasonal variations which differ from region to region while taking into account each piece of equipment's heat generation and stability needs. Tens if not hundreds of different robotic positioning and material handling systems need to be calibrated so that they can move wafers entire meters with a precision and accuracy measured in the tens or hundreds of nanometers. The design of the factory even needs to take into account regional variations in day to day humidity which usually takes a year just in data collection. Building the concrete shell for a fab is easy. It's filling it with equipment that actually works together to produce cutting edge technology that is the expensive part.

pjc50 5y ago

From my perspective in the software side of a semiconductor company, it's very much a problem of the last year. The AKM fab fire in November 2020 also made the situation worse.

Leherenn 5y ago
It's pretty much impossible to find GNSS chips currently, at least the ones we use (ublox). There's a massive TCXO shortage.

browningstreet 5y ago

Fabs also have to be manufactured.

akiselev 5y ago

> Maybe fabs take longer than a couple years to set up regardless of money spent? Maybe fabs take a level of expertise that only a handful of people in the world have, and it's a matter of ramping up promising undergrads to that level which will take like 15+ years? From the start of permitting to first tape out, I've heard five years for cutting edge, three to four for mid range (under 100nm), and two years for older processes. Subtract a year for expansions of existing facilities where hazmat permits and infrastructure already exist. Far more than a handful of people in the world are qualified to set up a fab but it really is hard, time consuming work that requires a lot of expertise. While most of the equipment in a fab is customized off the shelf (not totally custom but not cookie cutter), each piece has to be carefully calibrated to fit in with the whole. Stuff like HVAC, which is normally pretty predictable in commercial buildings, has to be custom designed for seasonal variations which differ from region to region while taking into account each piece of equipment's heat generation and stability needs. Tens if not hundreds of different robotic positioning and material handling systems need to be calibrated so that they can move wafers entire meters with a precision and accuracy measured in the tens or hundreds of nanometers. The design of the factory even needs to take into account regional variations in day to day humidity which usually takes a year just in data collection. Building the concrete shell for a fab is easy. It's filling it with equipment that actually works together to produce cutting edge technology that is the expensive part.