Numbers are a little suspect. Capacity is 120k, and they say up to 60k damaged? Inventory being what it is, they're doing great if they have half the wafers in process at any given time. So if there's a power bump, 60k is an absolute maximum. Furthermore, almost none of these lots are a dead loss. The ones in diffusion furnaces and wet hoods are likely 100% scrap, but CMP, CVD, implant, and dry etch mostly process one wafer at a time. Some of the wafers may be dusted by vacuum loss, but that's usually recoverable. Most photo layers give you redos -- strip off the resist and try again. I'd estimate the actual scrap from this event at no more than 10k wafers. The rest will be delayed production, but not a dead loss.
Edit: oops, I left out metals and test. Metals is a chamber process, wafers in process will be killed but unlikely it's the whole lot. You can lose wafers at test if they're in burn-in or being lasered.
Edit: I confused capacity and monthly output. My numbers are probably 3-4x low in terms of total impact on production, although they're likely reasonable in terms of their effect on one month's worth of output. 60k still seems like an overstatement.
The numbers aren't suspect, the journalist just segued into a second topic without being clear.
The outage happened at the Pyeongtaek plant, which according to Wikipedia has capacity of 450,000 chips per month[1]. Making the maximum estimated damage only 13% of monthly production capacity, not 50%.
The Xi'an plant is a completely different plan, which is beginning an expansion at the end of the month that should add 200k/month of additional capacity at that plant. Nifty fact, but not immediately relevant to the outage that happened.
5 comments
[ 5.1 ms ] story [ 25.1 ms ] threadEdit: oops, I left out metals and test. Metals is a chamber process, wafers in process will be killed but unlikely it's the whole lot. You can lose wafers at test if they're in burn-in or being lasered.
Edit: I confused capacity and monthly output. My numbers are probably 3-4x low in terms of total impact on production, although they're likely reasonable in terms of their effect on one month's worth of output. 60k still seems like an overstatement.
The outage happened at the Pyeongtaek plant, which according to Wikipedia has capacity of 450,000 chips per month[1]. Making the maximum estimated damage only 13% of monthly production capacity, not 50%.
The Xi'an plant is a completely different plan, which is beginning an expansion at the end of the month that should add 200k/month of additional capacity at that plant. Nifty fact, but not immediately relevant to the outage that happened.
[1] https://en.wikipedia.org/wiki/List_of_semiconductor_fabricat...