top 71 comments

sorted by: hot top controversial new old
[–] 77 points 5 months ago (3 children)

Reminds me of that time Samsung wouldn't sell ram to Samsung because they needed to compete with the demand from Samsung.

  • source
  • hideshow 3 child comments
  • [–] 9 points 5 months ago (2 children)
  • [–] 21 points 5 months ago

    Samsung is a Korean jaebeol. It's a massive conglomerate all owned by the one family, but each department operates very independently and is incentivised largely to maximise its own profit independent of the rest of the conglomerate.

  • source
  • parent
  • [–] 21 points 5 months ago

    Fuck Microsoft with an umbrella, but in this case it's not just only Microsoft, there are a lot more and bigger players on this shit dish

  • source
  • [–] 101 points 5 months ago (29 children)

    They could start manufacturing ram? Or fund startups trying to make ram?

    Seriously is there no way to get out of having only 2-3 chip and memory makers?

  • source
  • hideshow 29 child comments
  • [–] 129 points 5 months ago (4 children)

    You cant just "fund a startup trying to make ram". Chip fabrication is probably the most difficult and capital intensive production process there is. What manufacturing more ram looks like is investing tens of billions of real money (not the you give us stock we let you use our GPU deals the AI companies have been doing) and then waiting 5-10 years before the fab you funded starts to make chips, and hope prices are still high by then.

    That's why the existing manufactures are slow to scale up, they arent sure that the current spike in demand will still be there by the time their scaling up increases production.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 1 point 5 months ago

    It's actually worse than that, the whole ecosystem needs to be set in place, it's not just one company, it's the raw materials it's the logistics, the experienced people living in the area etc.

    It's more like creating a city from the ground up than create a company.

  • source
  • parent
  • [–] 33 points 5 months ago

    Honestly, no. There is no fast way to spin up fabs for this stuff. A lot of lithography equipment for the top tier stuff is made by 1 supplier, stocking a shitload of fabs with the right gear just isn’t something they can do.

    IMHO, the fastest way out of this mess would be for governments to regulate how supply is spilt between consumer and enterprise products.

  • source
  • parent
  • [–] 22 points 5 months ago (2 children)

    That still might take 10y because lithography printers aren’t cheap, their location isn’t cheap, and so much more. Once you watch an Intel, TSMC, or Texas Instru… chip factory tour.

    Wait, why doesn’t Texas Instrument cash in or did they offshore their production too?

  • source
  • parent
  • hideshow 2 child comments
  • [–] 14 points 5 months ago (5 children)

    They're going to start up an entire RAM company to fill a temporary shortage?

  • source
  • parent
  • hideshow 5 child comments
  • [–] 10 points 5 months ago (4 children)

    Yes, then they'll be shortage resistant

  • source
  • parent
  • hideshow 4 child comments
  • [–] 15 points 5 months ago (1 child)

    It can't really work the way to want it to.

    You have let's say Samsung who can make money selling a chip as long as the price > $50. And historically the price of the chip has averaged $100.

    But the demand is crazy and they can't keep up and the price of the chip is $500. They are making money hand over fist but let's say they feel a moral obligation (hahahahaha) to lower the price by increasing capacity.

    So they invest a billion dollars to increase capacity. Now that's a huge cost that reduces their margin on all chips. Between loans and maintenance, now they have to sell a chip for $90 to break even. But that's fine because they are making $410 per chip instead of $50!

    Except now you fix the supply issue and demand falls to normal. You've just cut your profit from $50 to $10. You have to sell 5x the volume to make the money you were making!

    Except it's even worse, because now you have all these extra chips you're building and nowhere to put them. Supply exceeds demand, pushing prices lower so instead of $100, they are selling for $80. Now Samsung loses $10 on every chip and they go bankrupt trying to pay back a billion in loans.

    So it's not really in their interest to build capacity to meet a temporary demand. Unfortunately.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 3 points 5 months ago

    That is not how it works at all. RAM is a necessary part for every compute device, and there is no way to recover any part of selling at a loss once it's acquired by the final owner. Thus, they will never be sold without a profit margin. This is very different from, for example, the video game consoles market. Once you have the console, you still need games, so, whatever loss the manufacturer is assuming has a chance of being filled by the customer "buying" (renting) games.

    Having said that, the rest of your logic is sound. However, I don't see prices dropping back to what they were before this AI bullshit even if the market is suddenly flooded by triple the offer vs demand. These corporations would certainly manufacture a fictional shortage somehow, makeing people rush-purchase ram for fear of the prices going too high again, and release all the production into the market to maintain the illusion of low offer.

  • source
  • parent
  • [–] 5 points 5 months ago

    There were loads of ram companies in the eighties, so many that when meager times came, half went bankrupt.

    RAM making is a brutal market. It's very different to chip making, it's "just" billions of capacitors.

  • source
  • parent
  • [–] 9 points 5 months ago (1 child)

    Honestly, we should probably just have a state owned and ran chip fab company. If the US is serious about security and/or innovation, that's what needs to happen. There's no way it's going to happen though sadly, but that's what should be done. >

  • source
  • parent
  • hideshow 1 child comment
  • [–] 5 points 5 months ago (6 children)

    The problem is it doesn't solve the core issue. What's to stop an AI data center buying up all the startup's stock and future stock also.

  • source
  • parent
  • hideshow 6 child comments
  • [–] 2 points 5 months ago (5 children)

    If you're the one funding the startup you can just reject that

  • source
  • parent
  • hideshow 5 child comments
  • [–] 1 point 5 months ago (4 children)

    Doesn't that go against Microsoft's fiduciary responsibilities to shareholders to maximize profits? At the very least, it would be contentious among the shareholders. Leaving money on the table? This is the kind of thing would need a very brave (or cavalier) CEO to risk a personal lawsuit.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 1 point 5 months ago (3 children)

    You don't have shareholders until you ipo. Startups have not ipoed.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 1 point 5 months ago (2 children)

    But Microsoft does. And those shareholders would want to know why money that could be going to them is frittered away on a startup that is constitutionally set up to leave money on the table.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 1 point 5 months ago (1 child)

    I mean Microsoft invested in OpenAI back when they were still claiming to stay non-profit. There are ways such as just treating them as an exclusive subcontractor in order to keep your main business. Microsoft is not a chip company afterall but needs them to have product at prices people will pay.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 5 months ago

    In that case they would be locking down their subsidiary even further to just one client. I guess it all comes down to the plausibility of their argument for making this investment, rather than, say increasing dividends or ploughing that money back into other investments for the business. I guess they were able to justify the OpenAI investment along these lines. If I could hazard a guess, it would be something along the lines of:

    This is a new market. We need to get in at the ground floor.

    I suspect, OpenAI presented itself as a turn key solution to get a foothold in this brave new world and maybe steal a march on Google.

  • source
  • parent
  • [–] 3 points 5 months ago

    It is an extremely long and complicated process to do so. China has been trying to enter chips manufacturing for decades and are now almost there (they don't have the capacity to produce the most advanced chips yet, but are still working on it).

    However the moment they got a decent DDR4 production going they announced they are going to "phase them out" in favour of DDR5 chips (of which they have not really perfected the process yet) so they can be used for Chinese AI-datacenters.

  • source
  • parent
  • [+] 2 points 5 months ago (2 children)
  • [–] 19 points 5 months ago (3 children)
  • [–] 18 points 5 months ago*

    If there was a decent governement behind they would have never left this reaching the point where the consumer market is left with nothing

  • source
  • [–] 55 points 5 months ago (7 children)

    Episode 7893 of corporations artificially creating problems to make profit by selling the solution. Just to then fail at solving the problem of their own creation.

    Brilliant economic system. You can marvel at its ingenuity and just get lost in it.

  • source
  • hideshow 7 child comments
  • [–] 15 points 5 months ago (6 children)

    They aren’t artificially creating this problem, they are inadvertently creating this problem.

    AI companies need RAM and other compute components to feed the AI beast which drives up demand, and cost but reduces supply for everyone, including their Xbox and Surface divisions.

    Given their enterprise divisions make vastly more money than their consumer divisions, nothing will change.

  • source
  • parent
  • hideshow 6 child comments
  • [–] 12 points 5 months ago (1 child)

    The problem is artificial, because 95% of all this 'AI' is useless, pointless, and a massive waste of resources. Nobody asked for it, and practically nobody is using it.

    Even in cases where a chatbot is helpful, a small, local model is more than enough. Not a single soul needs a 50T parameter model for anything. Nobody on the planet uses AI image enhancement or item removal.

    Almost all of this shit is purely made to inflate the bubble, and serves no other purpose. All the while ruining many other industries, people's hobbies, ease of finding factual information online, etc.

    The problem is artificial, because it was created pretty much deliberately, for no particular reason other than profit.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 5 months ago

    Just because you and your friends don’t use it, doesn’t mean it isn’t being used. If 95% of AI is useless then why are hardware prices so high? The demand must’ve dropped off by now if all the hardware they bought to date is going unused, because why would they be buying more?

    Companies are definitely using AI, and not just as chatbots. I use it everyday as a search tool, to manipulate data and perform other tasks which saves me time because I no longer have to do it manually.

    Most mid-large sized companies have terabytes or even petabytes of data that their AI tools are reasoning over, which requires more than the processing power of your small local model running on a Mac Mini.

    I don’t disagree with you that it is impacting other industries and consumers, but as long as customers have a use for it and are willing to pay, the AI industry will continue.

  • source
  • parent
  • [–] 6 points 5 months ago (3 children)

    But does anyone need ai companies?

  • source
  • parent
  • hideshow 3 child comments
  • [–] 1 point 5 months ago (2 children)

    The fact that the major players are still in business means someone thinks so.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 5 points 5 months ago (1 child)

    Are they still 'in business' though? They're burning through money they don't even have at a never before seen rate. Just to make barely useful software that works 20% of the time at best.

    I feel like it's the same as saying the banks and real estate companies were 'still in business' in 2008. Naturally, time will tell if it's the same cycle repeating. So far everything seems to be pointing in that direction.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 5 months ago

    Many companies operate without making a profit, sometimes it’s because they aren’t making enough money and sometimes they invest it back into the business.

    Regardless, SpaceX is about to go public with a valuation of $1.8 trillion, and they arent profitable.

  • source
  • parent
  • [–] 37 points 5 months ago (4 children)

    Asha Sharma standing on the glass cliff.

    She and Xbox will be gone inside of two years.

  • source
  • hideshow 4 child comments
  • [–] 25 points 5 months ago (3 children)
  • [–] 9 points 5 months ago

    They're still planning on making a new Xbox? Sometimes I forget there's an Xbox console anymore.

  • source
  • [–] 9 points 5 months ago

    They could just cancel all yet to be manufactured orders for starters...

  • source
  • [–] 8 points 5 months ago
    [–] 6 points 5 months ago

    Just... lemme help folks out with the corpo speak here.

    'leading-end'

    Ok now watch this ->

    'lead-ending'

    mind explosion gif

  • source
  • [–] 22 points 5 months ago (6 children)

    I'm all for shitting on Microslop, but let's not pretend they're solely responsible for the memory pricing crisis.

  • source
  • hideshow 6 child comments
  • [–] 21 points 5 months ago (3 children)

    let’s not pretend they’re solely responsible for the memory pricing crisis.

    Nobody is claiming that. Did you even read the literal first sentence of the article?

    the memory pricing crisis that its own AI ambitions are helping cause

  • source
  • parent
  • hideshow 3 child comments
  • [–] 7 points 5 months ago (2 children)

    The headline is claiming that. You can't just lie and then walk it back once someone clicks. There is a limit to how much inaccuracy you can put into a headline to simplify it, and this is definitely past it. To save one word they make it this wrong.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 3 points 5 months ago

    Agreed, the headline is clearly pointing fingers at Microsoft even if the whole story involves other corpos. What's the point of getting aggressive with people if the problem is the headline spewing misinformation?

  • source
  • parent
  • [–] 4 points 5 months ago* (2 children)

    I wonder if we’re going to see a console generation with expandable memory for the first time since the N64.

    Would make some sense if the current climate persists for a few more years. Sell variants with multiple ram configs at different price points, and when ram prices come down you can wack a ram stick/module in there.

    Obvious problems: consoles usually use unified memory, which probably won’t really work with expandable storage, so would need a different architecture. Also, if they use non-standard dimms, it’s unlikely the manufacturers would drop prices of those modules anywhere close to the actual amount of theoretical ram price drops. And this would require cooperation with game developers to make games that work with different ram configs, and give a tangible benefit for having more ram, without breaking compatibility for the base model units.

  • source
  • hideshow 2 child comments
  • [–] 6 points 5 months ago (1 child)

    You think RAM prices are ever going to drop?

    Sweet, summer child..

  • source
  • parent
  • hideshow 1 child comment
  • [–] 5 points 5 months ago*

    Yeah, I do. For two reasons.

    1. Buildout will start to slow down when the speculative investment slows down. Which is going to happen. None of the big ai companies are actually making money yet, and even if they were, eventually growth will start to level off. Investors are always looking for the next big thing, so once AI isn’t the shiny new thing, the magic investor money will start to dry up.

    2. Even if demand continues as it is, more memory will be manufactured. China is already producing memory, and I’m expecting them to scale fast. Also other countries and regions are realizing how important tech independence is, so I expect we’ll see more fabs pop up at least in the EU and US. One of the big reasons we’re in this mess is that the Korean fabs have been hesitant to expand production because even they don’t expect this to last.

  • source
  • parent
  • [–] 3 points 5 months ago

    Good, then they can finally be the kings of the shithole they're digging.

    sits on his sbc pcs, sticks of ddr4 and stack of old hard drives like a dragon

  • source
  • [–] 1 point 5 months ago