I’m not much of a conspirationist, but this memory “shortage” doesn’t add up for me; it’s like AI companies went from buying 10 memories as usual to 1000000, all of a sudden! It’s too fast, and sus.
it’s like AI companies went from buying 10 memories as usual to 1000000,
I mean, they basically did. OpenAI announced a few months ago that they reached deals to buy 900,000 wafers of DRAM per month, representing 40% of global production capacity. For a single company. There are several other companies competing at those scales, too.
It’s a production issue mostly. And one that has been predicted for a while now.
High Bandwidth Memory is what the AI companies want. Which while it isn’t the same as consumer DDR RAM, it does use the same chips.
It’s the production of the chips that are the bottleneck. Combined with the recent and rapid growth in AI data centers. That’s where the price increase and shortages come from, corporate buyers have bought out production, before it hits the consumer market.
Demand has risen sharply, and supply isn’t able to keep up (and it isn’t a simple thing to just make more chips).
I wonder if they realized there weren’t enough GPUs, so they decided lets just build a massive ram cpu/farm to do the job at 1/100th the speed and waste money on the inefficiency.
I’m not much of a conspirationist, but this memory “shortage” doesn’t add up for me; it’s like AI companies went from buying 10 memories as usual to 1000000, all of a sudden! It’s too fast, and sus.
I mean, they basically did. OpenAI announced a few months ago that they reached deals to buy 900,000 wafers of DRAM per month, representing 40% of global production capacity. For a single company. There are several other companies competing at those scales, too.
It’s a production issue mostly. And one that has been predicted for a while now.
High Bandwidth Memory is what the AI companies want. Which while it isn’t the same as consumer DDR RAM, it does use the same chips.
It’s the production of the chips that are the bottleneck. Combined with the recent and rapid growth in AI data centers. That’s where the price increase and shortages come from, corporate buyers have bought out production, before it hits the consumer market.
Demand has risen sharply, and supply isn’t able to keep up (and it isn’t a simple thing to just make more chips).
Once AI hype goes away they can keep the prices as it was during the shortages and make extra money from there.
I wonder if they realized there weren’t enough GPUs, so they decided lets just build a massive ram cpu/farm to do the job at 1/100th the speed and waste money on the inefficiency.