Comments Locked

62 Comments

Back to Article

  • jrs77 - Friday, March 16, 2018 - link

    I don't believe for a second, that it really happened. It's just another one of those stories to inflate the prices.
  • phantom505 - Friday, March 16, 2018 - link

    Ah, another tin-foil hat Libertarian in the computer biz that thinks he knows everything including the notion "he can see all the conspiracies". Moving on.
  • Flunk - Friday, March 16, 2018 - link

    11% of Samsung’s monthly output from a 30 minute power outage sounds really unlikely. If that was the case wouldn't they have multiple redundant battery backups? If true, the potential loses for Samsung are in the millions, I can't see the accountants signing off on a risk like that.

    To sum it up, the reports are very suspicious.
  • DanNeely - Friday, March 16, 2018 - link

    11% implies roughly 3.3 days to complete a wafer (80 hours if the factory is running around the clock; less if they only run 1 or 2 shifts and/or close on weekdays). If that's not a realistic number I'd expect them to be called on it by actual semi-conductor engineers (but remember the many additional layers in vnand should take longer to produce than planar chips).

    UPSing/generatoring an entire factory isn't going to be cheap, if power outages are rare enough insuring against the loss instead of making a massive infrastructure investment is going to be the way to go.
  • peevee - Friday, March 16, 2018 - link

    "UPSing/generatoring an entire factory isn't going to be cheap"

    Except you don't need to keep the entire factory operational just to make sure the shutdown is soft, without damaging anything already in production.
  • saratoga4 - Friday, March 16, 2018 - link

    >Except you don't need to keep the entire factory operational just to make sure the shutdown is soft, without damaging anything already in production.

    They'll have enough backup to shut everything down safely, but anything half way processed is trash at that point. If you're half way through deposition, and the machine powers down, it's not like you can just deposit a second layer on top to make up the difference. That'd be a dead chip anyway.
  • Morawka - Friday, March 16, 2018 - link

    which is why the actual lithography machines should be running on redundant power sources. Whomever designed that factory surely did not overlook this.
  • Samus - Friday, March 16, 2018 - link

    You would only hear about something like this from Samsung. I've never heard of a FAB losing production due to a power loss in the 30+ years I've been paying attention to semiconductor production.

    Sure, you could say the likes of Intel would keep it hush hush out of embarrassment (because this is an embarrassment) but this is the semiconductor manufacturing equivalent of Fukushima (an embarrassment) because its completely avoidable. The difference here is this is a relatively new, state of the art fabrication plant, and Fukishima a 60's era GE BWR that had a backup generator system last updated in the 90's.

    I mean, how does a 30 minute power loss really happen in new construction. I've installed permanent natural gas generators in businesses and homes for a decade and the longest a power outage last is 10 seconds while the generator starts up and the switch panel swaps the poles. For those 10 seconds, various UPS's keep critical equipment operational during the switch over.
  • Hul8 - Friday, March 16, 2018 - link

    If you accept that Samsung suffered a power outage (therefore technology sometimes fails), why assume that they had no UPS instead of their UPS systems failing or being misconfigured?
  • Ushio01 - Saturday, March 17, 2018 - link

    Well global foundries fab 8 needs 80MW's to run and the 40MW back ups aren't enough to save in-production wafers if there is a power cut. Fabs are huge and very power hungry a couple of seconds of power loss and everything in-production is scrap.
  • darkman99 - Saturday, March 17, 2018 - link

    Korean media has coverage of this: http://www.yonhapnews.co.kr/bulletin/2018/03/16/02...

    Battery powered emergency power did kick in. However, some areas had to be shut down because of the amount of power that battery power can cover.
    Stuff that cannot be shut down were kept operational.

    Not all the wafers will be thrown away. They are checking which are safe to use.
    Maximum damage is estimated at about $50 million.
  • brodock - Saturday, March 17, 2018 - link

    if a power outage in that level makes then loose that much, buy tesla powerbanks for the factory is probably MUCH MUCH CHEAPER
  • iter - Friday, March 16, 2018 - link

    Generators are not all that expensive. And get cheaper as you scale up. Running a high cost chip production line without backup power scheme is highly irresponsible, but then again, that's just my opinion, and I am not a corporation.

    Reality is kinda sad thou... They likely don't care all that much, as they have the option to capitalize and make money on that, so why spend money preventing it?

    They will without a doubt make more money on that single power outage than it would cost to supplement the factory with on-site emergency power generation.

    Let's not forget there is also the variable where does the problem occur. If the fab's own power system gives out then no amount of backup would help, further diminishing the point of investing in one.
  • saratoga4 - Friday, March 16, 2018 - link

    >They will without a doubt make more money on that single power outage than it would cost to supplement the factory with on-site emergency power generation.

    This is absolutely wrong. They lost a fortune on that power failure, and yes, they do care. Someone is probably getting fired over this.
  • iter - Friday, March 16, 2018 - link

    They have insurance that will completely cover any loses. The even tighter supply will allow them to justify even higher selling prices, boosting up the margins to more than make up for what production was lost, even if the losses are paid by the insurance.

    A megawatt of backup power plus the UPS buffering comes at 200k $ from vendors that sell at a premium. About 100k if you go el cheapo.

    To give you a basic idea on how much power 1 megawatt is, that's enough to power 800k average homes in the US.

    In this context, I highly doubt it would have taken more than a couple of million $ to provision the factory against power outages, which is next to nothing considering they spent like 30 billion on that fab.
  • iter - Friday, March 16, 2018 - link

    The only genuine concert samsung might have is to not get sued in cases they fail to meet their contract obligations. That however will likely be handled by the production they keep at stock.
  • IntelUser2000 - Friday, March 16, 2018 - link

    "To give you a basic idea on how much power 1 megawatt is, that's enough to power 800k average homes in the US."

    That's not a proper context considering modern fabs use 50-100 times that, or 50 to 100 Mega Watts.
  • PixyMisa - Friday, March 16, 2018 - link

    A megawatt isn't enough to power 800 homes, let alone 800k.
  • futrtrubl - Friday, March 16, 2018 - link

    Lol, he's saying an average home in the US uses 1.125 Watts of power? My phone uses more than that.
  • saratoga4 - Friday, March 16, 2018 - link

    >They have insurance that will completely cover any loses.

    I doubt their revenue is insured. Probably just damages.

    >boosting up the margins to more than make up for what production was lost

    Not how sales work. If your margins actually increased by more than the value of what you were considering selling, you would never sell anything. Losses from selling one less unit have to be positive, and for something that is very profitable like semiconductors, they are very large.

    >To give you a basic idea on how much power 1 megawatt is, that's enough to power 800k average homes in the US.

    It is also a tiny fraction of what a fab requires. Individual machines can use more than that. The problem isn't supplying 1 MW of power (which wouldn't be enough to run anything), it is continuing to supply it until the power comes back on. An outage could easily consume hundreds of thousands of gallons of generator fuel. At some point you run out because there is a limit to how much gasoline you can store.

    >In this context, I highly doubt it would have taken more than a couple of million $ to provision the factory against power outages,

    Larger fabs consume more power than some US states. You clearly have no idea how much these things actually cost.
  • Samus - Friday, March 16, 2018 - link

    Actually insurance may deny this claim if the underwriting had a rider requiring backup generators. TEPCO in Japan has been declined by all insurance claims for Fukushima because after investigation, they failed to meet some basic insurance requirements, in particular a warning from the 90's when a backup generator flooded and they were told to move them above ground. Obviously, that didn't happen, so future claims related to flooding were naturally denied.

    The same thing happened to a client of mine years ago when their previous IT dept failed to identify the tape backup system malfunctioned and the carbonite subscription expired and stopped running current backups. When the RAID array crashed, there most recent backups by that point were 2 months old and they had some huge project deadlines coming up. Data recovery cost $17,000 for the array drives since all backup systems were out of date.

    They tried to file a claim with the insurance for $17,000 and it was denied due to negligence.

    If something is avoidable, and not a true accident or an act of God, insurance will find a legal way to deny a claim.
  • Ushio01 - Saturday, March 17, 2018 - link

    How much power do you think they need? Even small fabs need a constant 10's of MW's of power big ones need 80MW's plus.
  • saratoga4 - Friday, March 16, 2018 - link

    >11% implies roughly 3.3 days to complete a wafer

    It takes several months to complete a logic wafer, and about 1-2 months for a NAND wafer. 3.3 days is too low because not all wafers are destroyed by a power outage, likely only those that were actually being etched or deposited were lost.
  • Calin - Friday, March 16, 2018 - link

    This is 64-layers flash. I would assume they brick the entire wafer if one of the layers go bad. So, I don't know how long (or short) it takes to work on a single layer, but you must multiply that by 64 to get the actual number of affected wafers.
  • psychobriggsy - Friday, March 16, 2018 - link

    64 layers times however many fab exposures per layer (probably several), plus base logic exposures. There will be a long fab pipeline, and multiple pipelines, so that Samsung can make enough NAND to make the whole thing worthwhile. Those pipelines appear to have 3.3 days worth of wafers in process at any given time.
  • edzieba - Friday, March 16, 2018 - link

    "If that was the case wouldn't they have multiple redundant battery backups?"
    Fabs use 100+ MEGAwatts of power during operation. That's not something you can feasibly run off any sane battery array for more than a minute or two.
  • edzieba - Friday, March 16, 2018 - link

    And when it comes to "just run only essential equipment": the essential equipment is ALL of it. Cut power to the FEoL/BEoL equipment mid cycle, and Bad Things happen to the wafers inside and potentially the machines themselves. Cut power to the chemical processor and anything in the deposition process is similarly effected. Cut power to HVAC and everything in the cleanroom needs to be considered contaminated.
  • mark4asp - Saturday, March 17, 2018 - link

    Battery backup may be able to keep an office running for ½ hour. Not a factory requiring high temperatures.

    Your reply is very suspicious. Maybe, you vastly underestimate the amount of energy needed to make silicon wafers.
  • iter - Friday, March 16, 2018 - link

    Well, I highly doubt that they don't have power generators or at the very least insurance that outta cover the damage.

    As highly as I doubt they will miss on using that as an excuse to jack prices up.

    The semiconductor industry is gradually adopting the "make more money on doing less work" technique, using an increasing array of excuses to jack up prices while simultaneously sacking hard working people. In the long run that doesn't bode well for anyone save for corporate executives.

    There probably was a power outage, and samsung probably doesn't have to use that as an excuse to jack up prices, and they will probably jack them anyway, and the price jack will probably extend far beyond making up for any alleged company losses.
  • iter - Friday, March 16, 2018 - link

    And of course, other, completely unaffected flash makers will use that as an excuse too.

    "Oh see, 3.5 % of a month's flash production was lost, lets jack prices up by 10% for the next 4 months"
  • rpg1966 - Friday, March 16, 2018 - link

    Jeebus, tinfoil hat alert, anyone?
  • iter - Friday, March 16, 2018 - link

    And that, ladies and gents, is all rpg1966 has to offer. So thorough, substantiated and convincing. Did you think of that yourself?
  • bji - Friday, March 16, 2018 - link

    Your arguments are lame and unconvincing. Basically you are saying that this event is something that Samsung would use to jack up prices, and even if they don't, they'll jack prices up anyway. You provide no evidence for your claim, you're just complaining like someone who expects companies to produce products for you at whatever price you feel like playing. I'm going to wager a guess that you're squarely in the "entitled millenial" demographic.

    Did you consider that Samsung paid money to partially produce flash chips that they now cannot sell? That now makes their average cost to produce the chips slightly higher. You act like raising prices to cover costs is some kind of sinister plot.
  • lmcd - Monday, March 19, 2018 - link

    No one needed convinced by RPG as he is only repeating common sentiment.

    The burden of proof is on you, the owner of the tinfoil hat.
  • iwod - Friday, March 16, 2018 - link

    Considering this is coming from Samsung, I wouldnt trust it wholeheartedly .

    And the news broke out BEFORE any Korean news site had it mentioned.
  • Kristian Vättö - Friday, March 16, 2018 - link

    This news broke out in Korea already a week ago. It's surprising it took this long for English media to pick it up.
  • billybeer321 - Friday, March 16, 2018 - link

    While I doubt this is some conspiracy, we are talking about an industry that has paid me three separate checks for class action lawsuits specifically due to price fixing over the last 20 years. Losing that much inventory over a 30 minute power outage does seem pretty remarkable.
  • bji - Friday, March 16, 2018 - link

    WTF does Libertarianism have to do with conspiracy theory? I don't think you actually know the meaning of the words you are using.
  • Xebec22 - Saturday, March 17, 2018 - link

    Libertarianism supports free market economics. You've used the wrong term.
  • deil - Friday, March 16, 2018 - link

    shit happens.
  • baka_toroi - Friday, March 16, 2018 - link

    It's a possibility but not probable enough to rule out all other possibilities.
  • eddman - Friday, March 16, 2018 - link

    "The US gov and/or intel sent ninjas to sabotage the fab in order to keep South Korea and/or samsung behind." - this is what one of the posters here would say. You know who you are.
  • Hurr Durr - Friday, March 16, 2018 - link

    US is a neocohen war machine hell bent on starting nuclear war, so sending ninjas to screw with Samsung is nothing special at all.
  • bji - Friday, March 16, 2018 - link

    All it takes to start a nuclear war is one button press, you retard, so obviously, if "the US" (as if there really even is such a thing as "the US" as a single coherent entity in this discussion) wanted to start nuclear war, it would have been started already.
  • tipoo - Friday, March 16, 2018 - link

    Oh dear.

    This doesn't bode well with the other price issues.
  • PeachNCream - Friday, March 16, 2018 - link

    We can't be looking at a huge price bump that loiters for a long time. It seems like there's enough inventory available so that consumers won't see a significant jump in the already inflated prices of solid state storage.
  • Hereiam2005 - Friday, March 16, 2018 - link

    I am living in South Korea, reading South Korean papers, and there's no mention of this at all.
    A quick scan of several Korean newspapers, also no mention.
    http://english.chosun.co.kr/
    http://www.koreaherald.com/
    http://www.koreatimes.co.kr
    Another Fake News trying to short stock?
  • bji - Friday, March 16, 2018 - link

    Someone else said it was reported two weeks ago, so maybe it's cycled its way out of the news sources there by now. I have no idea if this is true or not, but it could be the reason.
  • Hereiam2005 - Saturday, March 17, 2018 - link

    There is a nearly identical news, but it was in 2016 and in China instead.
    http://english.chosun.com/m/svc/article.html?conti...
    Wafer damage, power outage, etc.
    Someone recycled the news and everyone felt for it.
  • Sahrin - Friday, March 16, 2018 - link

    This is such a pack of lies. At Uni I worked in a power-dependent lab, and we spent hundreds of thousands building a power system that was essentially immune to medium term outages. The idea that somehow Samsung built a multi-billion dollar fab with a system that was capable of losing *hundreds of millions of dollars* of risk wafers is...laughable.
  • Reflex - Friday, March 16, 2018 - link

    It is cheaper to insure against losses than it is to build what would effectively be a small power plant to keep a facility of this scale online. Especially since the power plant would be rarely used yet would still require ongoing maintenance, testing, etc.
  • CoreLogicCom - Friday, March 16, 2018 - link

    If you're going to build a power plant to keep the facility online during a power outage, why wouldn't you just run the facility with the power plant 100% of the time? Then you don't care about street power. Not sure a power plant to run this type of facility wouldn't itself be an expensive undertaking (we talking a nuclear power plant?), but its certainly a realm of possibility.
  • close - Monday, March 19, 2018 - link

    There are gas turbines capable of generation ~600MW per unit. But building a powerplant to generate 100MW would cost well above $100 million.
  • titanmiller - Friday, March 16, 2018 - link

    Sounds like they need Tesla Powerpacks to backup their power supply. Apparently their current UPS strategy wasn't enough.
  • chaos215bar2 - Friday, March 16, 2018 - link

    Would it have killed you to add "for March" to the headline, so it actually said something meaningful?
  • Ryan Smith - Friday, March 16, 2018 - link

    That one is completely our fault and has been corrected. Thanks!
  • inighthawki - Friday, March 16, 2018 - link

    The amount of retardism in this comments section is beyond astounding.
  • IntelUser2000 - Friday, March 16, 2018 - link

    Lack of knowledge doesn't fit the definition of retarded.
  • DanNeely - Friday, March 16, 2018 - link

    Whatever you do, never look at the comments on an Intel/Amd/Nvidia/Android/Apple article on wccftech then.
  • Hurr Durr - Friday, March 16, 2018 - link

    And you`ve only posted one.
  • bji - Friday, March 16, 2018 - link

    Hey, I already called you a retard in an earlier comment before I even read this comment, you retard.
  • darkman99 - Saturday, March 17, 2018 - link

    Korean media has coverage of this: http://www.yonhapnews.co.kr/bulletin/2018/03/16/02...

    Battery powered emergency power did kick in. However, some areas had to be shut down because of the amount of power that battery power can cover.
    Stuff that cannot be shut down were kept operational.

    Not all the wafers will be thrown away. They are checking which are safe to use.
    Maximum damage is estimated at about $50 million.

Log in

Don't have an account? Sign up now