You would think that RAM manufacturers would ramp up production. Hopefully the fact that they aren’t means the bubble is going to burst within a couple of years.
I doubt it will take a couple years. They’re burning through so much cash right now that they’ll be bankrupt in a couple years and despite sunk cost fallacy they won’t let it get that bad. At some point they’ll cut their losses and pivot to some other new fad. The small handful of uses that make sense will stick around and a few companies will be in just the right place to make it turn a profit but the vast majority won’t. Some will go bankrupt (if we’re lucky Meta and/or X will be one of them) and some will just write it off as a failed experiment. Either way just as hard as prices spiked we’ll see them cratering before they rebound back to normal. Six months would be highly optimistic, but a year probably isn’t out of the question.
Of course all of this might be moot if Shitler manages to start WW3 by attacking Greenland. If that happens RAM prices will be the least of everyone’s worries.
Of course all of this might be moot if Shitler manages to start WW3 by attacking Greenland. If that happens RAM prices will be the least of everyone’s worries.
Which WW3? You really mean European militaries won’t back out when threatened? Suppose Shitler kills a few Danish hikers in camouflage with professional pride. Nothing will happen, he’ll just take it.
I swear. For a big war based on alliances and rules to happen you need principles and institutions of the military sort.
Practically all DRAM production is sold out for next year, there isn’t going to be any significant new supply into the consumer market in that timeframe. Also, when these massive piles of memory end up sitting in a warehouse because the bubble pops, this stock will never make it to the consumer market, because AI data centres use HBM, which is soldered directly to motherboards in such a way that it’s impossible to turn it into modules again. Even if all the DRAM producers start selling to the consumer market jan 1st 2027, the pent up demand will still probably keep prices high.
All that to say, I don’t think prices will crater.
It takes time to build factories, the big three RAM manufacturers apparently think AI is a bubble, although that could just be an excuse for price fixing.
China’s cxmt is apparently ramping up production, but they’re still relatively small.
China is pushing hard to make their domestic brands the new standard world wide so they’re not worried about whether the bubble pops or not. They want to drive prices down even if that means selling at a loss because they know that’s what it’s going to take to dislodge the entrenched players. For better or worse it’s likely a winning strategy because the existing players are more concerned with maximizing their quarterly profits rather than meeting any kind of consumer demand or indeed even selling to consumers at all.
If I as a customer pay, idk, 10€ per gb of RAM and an AI data center pays 20, do you really think they are going to consider how they are feeding a bubble that in the not-so-long term will burst and fuck them up too?
Nah, mate, they are capitalists, they only care about the money they can make in the extremely short term.
What do you think costs more? Building RAM sticks for individual users who’ll pay $200-300 per 8/16 DRAM modules, or building sticks for enterprise customers who in one single order will spend millions of USD? Also latter requires no fancy packaging, marketing, or variations in the product.
You would think that RAM manufacturers would ramp up production. Hopefully the fact that they aren’t means the bubble is going to burst within a couple of years.
I doubt it will take a couple years. They’re burning through so much cash right now that they’ll be bankrupt in a couple years and despite sunk cost fallacy they won’t let it get that bad. At some point they’ll cut their losses and pivot to some other new fad. The small handful of uses that make sense will stick around and a few companies will be in just the right place to make it turn a profit but the vast majority won’t. Some will go bankrupt (if we’re lucky Meta and/or X will be one of them) and some will just write it off as a failed experiment. Either way just as hard as prices spiked we’ll see them cratering before they rebound back to normal. Six months would be highly optimistic, but a year probably isn’t out of the question.
Of course all of this might be moot if Shitler manages to start WW3 by attacking Greenland. If that happens RAM prices will be the least of everyone’s worries.
Which WW3? You really mean European militaries won’t back out when threatened? Suppose Shitler kills a few Danish hikers in camouflage with professional pride. Nothing will happen, he’ll just take it.
I swear. For a big war based on alliances and rules to happen you need principles and institutions of the military sort.
Practically all DRAM production is sold out for next year, there isn’t going to be any significant new supply into the consumer market in that timeframe. Also, when these massive piles of memory end up sitting in a warehouse because the bubble pops, this stock will never make it to the consumer market, because AI data centres use HBM, which is soldered directly to motherboards in such a way that it’s impossible to turn it into modules again. Even if all the DRAM producers start selling to the consumer market jan 1st 2027, the pent up demand will still probably keep prices high.
All that to say, I don’t think prices will crater.
Those shenzen mofos can figure out a solution to utilize that ram mark my words
Im sure china will and the tradional manufacturers will cry about losing market share while begging for tarrifs and handouts.
It takes time to build factories, the big three RAM manufacturers apparently think AI is a bubble, although that could just be an excuse for price fixing.
China’s cxmt is apparently ramping up production, but they’re still relatively small.
China is pushing hard to make their domestic brands the new standard world wide so they’re not worried about whether the bubble pops or not. They want to drive prices down even if that means selling at a loss because they know that’s what it’s going to take to dislodge the entrenched players. For better or worse it’s likely a winning strategy because the existing players are more concerned with maximizing their quarterly profits rather than meeting any kind of consumer demand or indeed even selling to consumers at all.
They are ramping up production. But it all goes to AI data centers.
Why can’t anyone go against the current and sell to individual customers only?
If I as a customer pay, idk, 10€ per gb of RAM and an AI data center pays 20, do you really think they are going to consider how they are feeding a bubble that in the not-so-long term will burst and fuck them up too?
Nah, mate, they are capitalists, they only care about the money they can make in the extremely short term.
But bulk orders typically cost less
Selling in bulk is convenient for manufacturers too.
Because consumer sales is slow and expensive.
What do you think costs more? Building RAM sticks for individual users who’ll pay $200-300 per 8/16 DRAM modules, or building sticks for enterprise customers who in one single order will spend millions of USD? Also latter requires no fancy packaging, marketing, or variations in the product.
Profits.
And if say SK Hynux decides to only sell to consumers, Samsung and the others will sell even more to data enters because there’s more cash to be made.
No small ask. New factories, new supply chains, more staff, etc.