On Friday, Western Digital disclosed that a factory operated by Toshiba Memory Corporation as part of the two companies’ joint production operations in Japan’s Yokkachi region had been affected by a power outage on June 15. According to to the storage manufacturer, the outage not only affected the facilities and tools used to make its flash storage, but six exabytes of Western Digital NAND that was in production was lost.
According to Reuters, the plant has been running at less-than-full capacity since the outage occurred, and won’t be able to resume full production until the middle of July. Western Digital says that this will affect the company’s bottom line during the first fiscal quarter, which runs July through September. Western DIgital had just finalized plans to invest in Toshiba’s “K1” factory that’s currently under construction in the Kitikami region of Japan in May. Construction on that facility won’t complete until fall, and WD expects that it will begin producing 96-layer 3D NAND chips in early 2020.
Between the in-production NAND that was lost and as a result of the diminished capacity, global supplies of flash memory could be reduced as much as 25% between August and October, according to a report from Blocks and Files. The six exabyte figure published by WD is just for that company’s production. The total damage to the global supply chain could be far greater. Wells Fargo analyst Aaron Raker told Blocks and Files that he believes around 60% of the factory’s production to be for Toshiba, dealing that company a nine exabyte loss, too.
The last time a big ol’ chunk of NAND was lost due to a power outage, Samsung lost around 3.5% of the world’s flash memory production for the year back in March of 2018. My back-of-the-napkin math suggests that this would be around twice that percentage in Toshiba and Western Digital’s case. As a result of the loss, TrendForce expects 2D NAND prices to rise and 3D NAND prices to start leveling off. If you’ve been eyeing a shiny new SSD and haven’t already jumped on it, there might be no time like the present.
Share
Question & Answers (0)
Leave a Comment
Ben Funk
Depends on whether they can pause the process between each mask layer (I don’t know if they can or not, maybe someone else here knows). If they can pause after a single mask layer, then it could take up to a day or so. If they need to finish all of the steps to complete the wafers currently in process, you are potentially looking at weeks or even months.
Making a modern chip is a very lengthy process. The only way they are able to maintain high production volumes is by making thousands of them in parallel (hundreds per wafer, with hundreds of wafers in the pipeline at any given time).
Chicks can still be hot irrespective of their looks.
FYI “Japs” is considered derogatory. Only marginally better than “Nips”.
Problem is, “acceptable” (according to you) equates to “flash chips being sold at a loss”. Current pricing was not sustainable given current flash densities. Western Digital’s stock price was UP significantly on this news, so semiconductor industry pundits and investors obviously agree that they (and the flash industry as a whole) had an excess of capacity.
Sorry to all the conspiracy theorists but its not just the Japs and the Koreans who make SSD any more. There are large plants in China who will be only to happy to supply you with any and all the flash you need. I reject your conspiracy theory and replace it with my own. A group of malingering Chinese “Tourists” were near the plant, and then poof! a substation blew up. Or it could have been Bulgarian “Tourists” who are known to do the behest of others.
The chinese NAND may not be the latest and greatest but the price per Gb is among the lowest, drevo, GoldenFir, TCsunbow etc etc.
And how long does that take?
And if you do scrap them, can you at least avoid damage to the equipment?
TR’s own article profiling the now-dead GloFo 7nm facility noted that they used flywheels, batteries, and diesel generators to provide 40 MW of backup power. Maybe you guys are all right and they were fools to do that — maybe that’s why they gave up on 7nm. Maybe the smart fabs all just plug into the grid and hope for the best.
is she? seems like an objectification. plus, you’ve never even seen her.
This happens exactly when prices of SSDs were becoming significantly more acceptable for the offered capacity.
Makes me wonder: 1. Was it really an accident ? 2. It it was, will they (they as in all SSD manufacturers) use it as an excuse to profit a lot from that by increasing prices ?
People like to believe in conspiracy theories even when the problems are due to genuine, serious issues that take a lot of hard work to fix, because they want to believe that we have control.
And, it happens more with people that have no knowledge about the issue at hand.
Let me wake you from that illusion. You *don’t* have much control of your life as you believe.
Governments and large organizations believe this. You probably have less control the bigger the organization gets. This is why government programs have minimal affect on mitigating the problems, and usually make it worse.
Things like the Business Cycle exist. Lots of things that seem to happen due to malice are merely coincidences. Not all, but lots.
This, included.
This.
I regularly visit an offsite datacenter in London where we keep some DR servers and that place is vast, yet an insignificant spec compared to the size of Toshiba’s NAND facility.
The ISP datacenter is home to Microsoft, Cloudflare and one other large anonymous customer who dwarfs the other two in floors occupied (so basically either Amazon or Google). Three Caterpillar gas generators provide up to 18MW of on-site power and the building behind them is, I’m told, filled with nothing but batteries and cooling for those batteries.
Any idea how much power the facility uses regularly? Fuel-cell generators range from 1-4MW for a single “unit” and there looks to be room for more than a few large deployments on the site.
30MW seems feasible, though I would be curious to learn the cost of keeping so much power generation on standby versus just throwing up ones hands and writing off the material that was in various stages of production.
Pretty sure there is no “safe shutdown” for wafers currently in the production line. You finish ’em or scrap ’em.
Have to wait for SMR’s
Is 3 hours not enough time to safely shut down?
exactly
Yeah, not really the same thing.
Just like in 2012 when Seagate, WD and other HD manufactures teamed up to flood the country of Thailand to ‘drive’ up the cost of Hard drives…
I wonder how many poor souls they drowned just to make a few dollars more…
…and the fact that if the outage lasts more than ~3 hours, those Powerwalls don’t do you any good anyway. And if you go 10 years WITHOUT an outage, you’ll STILL need to replace all the batteries since lithium ion batteries degrade even if not used.
Crap, I had plans to update all my drives at the end of the year, as part of my Windows 10 migration.
Can’t do it right now… just bought a fridge, and one of the cars needs tires and an alignment. Curse you, real life! Foiled again!!!
You win 100 Internets
Don’t forget the additional switchgear to handle those megaamps, and the now-doubled infeed capacity from the utility to charge them. And the physical space to house those 40,000 power walls.
So kinda like branch prediction. Too bad they took the wrong branch
A Tesla power wall can output 5kw and costs about $7k.
40,000 powerwalls would output 200 MW.
Let’s assume Tesla gives discounts on bulk purchases, so maybe $5k per powerwall , so $200 million.
Compared to the cost of this complex of fabs, that sounds like a reasonable price to me
I don’t know how many times I’ve told those boys, [url=https://www.youtube.com/watch?v=AqeatUvRu68<]never call chicks bro[s<]ad[/s<]s[/url<].
These fabs draw megawatts. HUNDREDS of megawatts. Any backup system that can sustain them is not a battery bank or generator, it’s a dedicated power station.
As for why they don’t just shut down noncritical systems: they’re ALL critical systems! Lose the enormous AC filtration plant? Wafer contamination, bin everything and deep clean all equipment before recommissioning. Chemical plant goes down? Can’t guarantee purities and mixes are correct, bin everything and deep clean all equipment before recommissioning. And so on.
Power goes out, you lose every wafer in flight and need to effectively recommission the plant.
Earlier this week we had a whole mess of flash, now we have a whole flash mess.
Chicks can be bros too!
But before you can conspire with your friendly “competitors”, a great excuse is needed.
There wasn’t room in the budget for a roll of gaffe tape so WD made this gaffe instead.
Thank you for the mental image involving vast expanses of snow and a highly suspect substance obtained from an ox.
dink!””silence… Did someone forget to charge the backup batteries?
Revenue targets for their competition, yes.
Auxy ain’t a bro, bro. She’s one hot chick.
Because desire for higher SSD prices.
Good timing as recently picked up 12TB for under £700 which included a 4TB Intel PCIe U.2 data centre drive. 🙂
Toshiba capping their supply artificially to drive up demand / prices kind of puts them in some legal hot water which has some pretty hefty costs down the line.
Losing 1/4 of your quarterly production but selling the remainder at double cost adds up pretty quickly.
“Oops, I tripped over the power cord!”
[url=https://youtu.be/LntQvMZuwEk<]NAnd NAnd Na NAnd Nand Nand[/url<] Soon they will be Lovin, Touching, Squeezing, Another.
Hmm… I’m a little skeptical that the price increase is big enough to offset their production loss. This sounds analogous to Saudi Arabia setting a major oil field on fire — sure the price goes up, but wouldn’t it have been better to just cap wells in a controlled way and say they’re down for maintenance or whatever?
My uninformed guess is that this is a case of a desperate company (Toshiba) cutting corners in order to stay afloat. Maybe it’s a good thing their nuclear power business went belly up — this could have been worse.
That’s the key really. Not only will they recoup their costs by prices increasing in the short term, but they’ll ride out those higher prices for at least a year after supply has normalized again. Obviously everyone else gets this benefit also (without the loss of revenue) but that’s probably part of the plan anyway.
Well, YOU would. IMHO, that’s the least surprising thing that will happen this year.
No no no. That’s Shill [i<]Club[/i<]! Oh wait, I talked about it!!
1. You do not talk about shill training.
2. You [i<]do not[/i<] talk about shill training.
And then the whole management team went out and had a crazy party, knowing revenues would be up and bonuses would be fat.
Exactly. I don’t know the power usage of these fabs, but I can’t imagine it’s so huge that they can power it with batteries but somehow can’t spin up some generators.
I guess there’s no real incentive for backup generators since they can just jack up prices, though…
Maybe they drank too much Musk-aid.
Backup batteries are great for these situations:
1. Transient power hiccups.
2. Long power outages to give the generators enough time to turn on when you absolutely need to keep running (like this fab). OR
3. Long power outages where they give you enough time to turn things off in a controlled and non-damaging manner.
In Shill training 101 they make sure to tell us exactly which products will be canceled in the event of a power outage so that we are prepared for any and all contingencies!
Yeah, there’s something a bit fishy here. I read that their backup batteries ran out, but if it can be powered by batteries, it can absolutely be powered by backup generators.
Great post, bro.
Exactly
Read my post #1
Huh. I had the impression that because companies realize the tremendous cost of a power outage, they invest in multiple backups, including batteries and diesel generators, to prevent exactly this kind of a catastrophe. Am I wrong about that?
Or is this a case of [url<]https://www.youtube.com/watch?v=XqGnAIgzUTY[/url<] ?
I’m really curious as to how this happens exactly. Fabs always have backup power systems for exactly this reason and they usually pay increased rates to be the last place that loses power when a grid fails and the first place to be repaired when it starts coming back online.
So what happened? The main grid went down and then the backup didn’t work? Some sort of problem with the internal building electrical system? The article says that the regional grid went down, but then why no backup power?
“Thanks for no memories!”
— Everyone else in the cartel
OH COME ON!!!! COME THE HELL ON!!!
You don’t REALLY believe this is ANYTHING but an excuse to raise the rock-bottom NAND prices, do you? SERIOUSLY? Are you THAT gullible? This happens, JUST as a whole wave of new SSDs are about to hit the market? People won’t notice the price increase because they’ll just attribute it to the cost of PCIe 4.0 development. Such bullcrap!!!
you’re srsly late, bro
In before tipfoil types claim that the whole thing was intentional self-sabotage.
Say goodbye to rock-bottom SSD prices.
I guess that’s one way to increase prices. I can almost hear the CFO yelling, “REVENUE TARGETS WILL BE MET. PULL THE SWITCH!!”