Look at it from the bright side. Manufacturers are building massive new capacity for demand that will never come. Already produced chips can’t be repurposed but machinery can, easily. In a few years RAM will be dirt cheap.
Too late, I’m afraid. The supply for DRAM basically can’t adjust, so even if Oracle, Meta, and OpenAI went bankrupt tomorrow, it will take some time to catch up.
flamingo_pinyata@sopuli.xyz 2 days ago
brucethemoose@lemmy.world 2 days ago
That’s not true, from what I’ve read:
www.trendforce.com/…/20251113-12780.html
despite higher ASPs boosting profitability across the memory industry, capital spending on DRAM and NAND Flash is only anticipated to increase modestly in 2026. This limited investment growth is unlikely to significantly affect bit output.
Memory makers are skeptical, too.
xep@discuss.online 2 days ago
They also have a history of forming cartels and colluding to fix ram prices, so I doubt prices will normalize for a while.
rumba@lemmy.zip 2 days ago
I don’t actually buy it. It won’t fit on memory modules as they are today, but in the end, it’s just faster, higher density, prob has some extra features, but nothing you can’t rework a motherboard or modules to support. of course, min quantity will be like 128GB :)
brucethemoose@lemmy.world 2 days ago
That’s not true at all. It’s not economical to transplant RAM ICs once they’re packaged and soldered onto something.
And if they’re produced as, say, HBM modules, they absolutely cannot be repurposed for, say, DDR5 or LPDDR5 CPUs, or GDDR GPUs. There’s no reworking, the memory buses on processors simply do not support them, and altering them would have a massive development cost.