Power disruption at Micron DRAM facility in Japan

Conspiracy theories? I would never accuse the dram manufacturers of price fixing. (To be clear that was sarcasm before anyone searches up a thread with me accusing the dram manufactures of price fixing).

Power outages happen though.
 
Conspiracy? They don't need to. Power outages happen. Their choices are to have a backup power source which would be obscenely expensive or raise prices and bill the loss to the customers. Pretty easy choice if you ask me.
 
Stuff certainly can happen. But I think you have to be naive to not think that some of the things that do happen with RAM chip production are deliberate ploys. The disruptions seem to happen like clockwork when the supply is high and the prices low. The joke is so old and overused that it's not even really amusing anymore, it just sounds worn out... yet it isn't just a joke, it's based on actual happenings. This stuff should be investigated by outside agencies.
 
Conspiracy? They don't need to. Power outages happen. Their choices are to have a backup power source which would be obscenely expensive or raise prices and bill the loss to the customers. Pretty easy choice if you ask me.

you dont have a clue

AMD Fab 25 here in austin had dual power station feeds and that was in the 90's. one substation was right across ben white blvd ... about 1/2 mile away
there was another substation ON the amd property!

Samsung here in austin - oh wait whats that the only one outside of korea ... until the 2nd is built in taylor ... not only has a substation less than a mile away but also has DUAL power feeds

1 day of disruption of the wafers is enough to pay your "obscenely expensive" bill ... it happen last year at samsung during the ice storm...my buddy is a manager in charge of a line and he said ercot cut them off and they lost thousands of wafers (stuck in machines mid cycle)....wafers...not chips
 
its always this shit..... its always power loss or some flood... billions at stake but lets leave it chance....

1 day of disruption of the wafers is enough to pay your "obscenely expensive" bill ... it happen last year at samsung during the ice storm...my buddy is a manager in charge of a line and he said ercot cut them off and they lost thousands of wafers (stuck in machines mid cycle)....wafers...not chips
I've been in datacenters or working very near to them since 2007. I'm no rookie. Every datacenter in the country built in the last 15 years or greater in some cases can run COMPLETELY on generators. My facilities in Ashburn Virginia were required to run off generators only one year for two weeks during the summer as the grid couldn't handle datacenters and larger facilities. Mind you, we're talking about datacenters only, so everyone can access pornhub and their yahoo mail.

Yet major production facilities with umpteen millions of dollars on the production line continually have environmental issues affect their production.

At *SOME* point this will become unbelievable. I don't know when we'll get there, but we'll get there.
 
"Micron expects production output loss and associated cost impact from both productivity and wafer scrap across Micron's fourth quarter of fiscal 2022 and first quarter of fiscal 2023.

"Micron noted it would be working to utilize existing inventory and its remaining DRAM factory network to meet customers' critical requirements. The company is still examining the impact on near-term supply."

Sometimes a power disruption is just a power disruption. Or part of a larger nefarious scheme. :sneaky:
 
I've been in datacenters or working very near to them since 2007. I'm no rookie. Every datacenter in the country built in the last 15 years or greater in some cases can run COMPLETELY on generators. My facilities in Ashburn Virginia were required to run off generators only one year for two weeks during the summer as the grid couldn't handle datacenters and larger facilities. Mind you, we're talking about datacenters only, so everyone can access pornhub and their yahoo mail.

Yet major production facilities with umpteen millions of dollars on the production line continually have environmental issues affect their production.

At *SOME* point this will become unbelievable. I don't know when we'll get there, but we'll get there.
I've seen 2 datacenters in my career lose power when doing a power failover test to generators. 2 very large DC companies. It can happen.
 
Why wouldn't they just announce some market downtime? That is common in other industries.
 
Congrats! You win 30% more expensive DRAM!
FUCK YEAH!
I'd like to first thank the academy and all of the wonderful members who this would not be possible without.
I'd also like to thak my IT coach and deliver guys who were always rock solid during this time getting everything to the table when it was needed even during COVID issues.
I have to throw some mad props out also to my 2nd Grade teacher Ms. Barrett who told me I could really clap together some mother fucking erasers.
That then gets me back to my family who has had to cut back on expenses like food while I have been shopping for components.....those waist lines are looking sick ya'll! I mean sick!
And I have to finally thank myself for having the courage to play the same losing hand over and over and over without ever having any semblance of chance before.

Bronze_Medal_Meme_Banner.jpg
 
I've seen 2 datacenters in my career lose power when doing a power failover test to generators. 2 very large DC companies. It can happen.
Of course it can happen. Planes fall out of the sky to no fault of the pilot. Things happen. But would you be arguing that these facilities electrical issues are purely due to construction or implantation issues? Or were they simply ill designed from the beginning?
 
Yeah a lot depends on the infrastructure. It doesn't help to have two substations if they are both fed from the same unreliable supplier.
And running a data center on generator power is one thing, a large factory is probably 2 orders of magnitude more power. Factory I worked at pulled over 200 megawatts. How many near line generators would that take?
 
Yeah a lot depends on the infrastructure. It doesn't help to have two substations if they are both fed from the same unreliable supplier.
And running a data center on generator power is one thing, a large factory is probably 2 orders of magnitude more power. Factory I worked at pulled over 200 megawatts. How many near line generators would that take?
I completely agree that a production facility would be a larger ask to back up. But it can be done. Any when the stakes are high enough it's worth it. I guess losing millions isn't worth it.
 
I completely agree that a production facility would be a larger ask to back up. But it can be done. Any when the stakes are high enough it's worth it. I guess losing millions isn't worth it.
Its a pretty simple calculation. Cost of lost production versus cost of mitigation.

Its also possible that they carry insurance to cover these types of disruptions, or even have a deal negotiated with the power company.
 
Yeah a lot depends on the infrastructure. It doesn't help to have two substations if they are both fed from the same unreliable supplier.
And running a data center on generator power is one thing, a large factory is probably 2 orders of magnitude more power. Factory I worked at pulled over 200 megawatts. How many near line generators would that take?
Certainly, but in that case then segment it out, put critical systems on generators, such as saving your wafer's so you don't have to bin them and once out of the process so they can be stored, done, let everything else shutdown.
But as noted, sure the costs vs insurance and other things have been considered and if they have a 99.999% or what ever up time....
 
I just don't think they would have to manufacture a crisis to manipulate thier selling price. They can slow down production without wrecking thier factory.
 
Certainly, but in that case then segment it out, put critical systems on generators, such as saving your wafer's so you don't have to bin them and once out of the process so they can be stored, done, let everything else shutdown.
But as noted, sure the costs vs insurance and other things have been considered and if they have a 99.999% or what ever up time....
Saving your wafers means having every machine processing wafers on backup, which means having 99.9...% of the facility power be backed up. Getting them to a safe state means completing the current layer, and each layer takes hours.

At several hundred megawatts, you're talking about having a full scale natural gas power plant on site as a backup. Medium term I suspect as grid-scale batteries become cheap enough to go from doing frequency stabilization and storing 1 or 2 hours of power to storing 6-24 hours worth we might see large semi-conductor facilities I suspect we'll see them co-located with the fabs.

Until then though, the people who run the fabs actually can run the numbers on how much it would cost to have enough backup power on site to do a safe shutdown vs how much losing every in process wafer and a month or two of yearly revenue costs. They've all concluded that backup power is even more expensive than the risk of loss.

And no they couldn't decide otherwise but decide not to do it for decades as a way to generate artificial scarcity any more than they could keep any of the price fixing schemes they get busted for every 5 or 10 years when someone decides to blow the whistle and retire on their 1% share of the collected govt fines.
 
Saving your wafers means having every machine processing wafers on backup
Depending on how many different machines are running at once, it might be worth looking at keeping the ones that are close-ish to done running, to avoid wasting all that sunk time.
 
Depending on how many different machines are running at once, it might be worth looking at keeping the ones that are close-ish to done running, to avoid wasting all that sunk time.
I'm sure all of this has been planned out to the nth degree. I dont work in a fab but im in the semi industry for test/dev. We have backup plans, risk mitigation strategies, all this stuff and its not that critical when we lose power, chilled water, ln2, air, ect. Worst for us is we lose time on a test flow or a wafer gets scratched or a probe card gets mangled.
 
Saving your wafers means having every machine processing wafers on backup, which means having 99.9...% of the facility power be backed up. Getting them to a safe state means completing the current layer, and each layer takes hours.

At several hundred megawatts, you're talking about having a full scale natural gas power plant on site as a backup. Medium term I suspect as grid-scale batteries become cheap enough to go from doing frequency stabilization and storing 1 or 2 hours of power to storing 6-24 hours worth we might see large semi-conductor facilities I suspect we'll see them co-located with the fabs.

Until then though, the people who run the fabs actually can run the numbers on how much it would cost to have enough backup power on site to do a safe shutdown vs how much losing every in process wafer and a month or two of yearly revenue costs. They've all concluded that backup power is even more expensive than the risk of loss.

And no they couldn't decide otherwise but decide not to do it for decades as a way to generate artificial scarcity any more than they could keep any of the price fixing schemes they get busted for every 5 or 10 years when someone decides to blow the whistle and retire on their 1% share of the collected govt fines.

You clearly know a hell of a lot more about this than I do! Appreciate the info.
 
You clearly know a hell of a lot more about this than I do! Appreciate the info.
Mostly I'm assuming the person above who gave a 200MW figure for a fabs power consumption is in the right ballpark. That's a full up natural gas powerplant in size. If the actual value is an order of magnitude lower, a collection of semi-trailer sized generators doing a few MW each could carry the load. (As long as the machines could be UPSed long enough for the generators to start anyway.) The fact that they don't do this suggests to me that the 200MW figure is at least in the right ballpark.

I think I was wrong in assuming that they'd need 1 day class grid scale batteries to be useful though, even the standard ~2h type currently being installed would be useful. Reserve 50% of the capacity for on site emergency use, while making the remainder available to the grid for frequency stabilization and peaking uses (to help offset the cost); the 1-2h of power available would still be enough to ride out shorter outages and if the word from the power company about restoration time is bad triage the in process wafers to salvage as many as they can with the available stored power.
 
Another thing to mention is that all of these facilites are audited depending on the standards their products adhere to. The auditors ask for things like risk mitigation, force majeure, all that. All of this has to be developed and documented if you want to sell parts into automotive, DoD, ect.
You also have to think about not just the scrap cost, but reputation and market share damage. Stock prices get affected by this type of news, which is always the CEO's #1 priority regardless of what they say.
 
Back
Top