For those looking to upgrade your computer. Be prepared for a shock.
Prices for memory and storage are going through the roof due to AI demand sucking up all the supply.
RAM I bought a few months ago is now selling for 217% the price I bought it. Storage I bought is not 23% more after 5 months. Some analysts predict this will get worse over the next year or two.
It might be worth looking to the second hand market (e.g. ricardo) since most people price their stuff on what they paid and not necessarily the market price. Another option would be to buy complete systems and plundering it for parts since most can’t be bothered. I remember during COVID selling a PC, the buyer told me that he would be happy for me to just post him the GPU since that was all he wanted.
Yeah. I made this mistake during covid. I sold my GPU for just less than what I bought it for and didn’t realise market prices meant that the value had doubled. This was problematic as when I was trying to buy a replacement, the people were selling them at market prices.
Anyone remembers the flooding in Thailand when HDD prices exploded and the world moved on to SSDs? Wondering if software developers could write something that runs in less expensive software.
The shortage is having knock on effects. It started out with certain RAM and then all RAM, then fast NAND storage and moved onto HDDs too.
Now I read that even your smartphone is expected to cost more as AI look to use smartphone RAM instead of their usual RAM due to shortages:
I’m having a beefy PC, 192GB RAM, Ryzen 32 threads, I’ve been thinking of getting rid of it since ~3 years. It was an excitement to build it but I never made use of it. Hoping to sell it at decent price at last 
I’ve already bought a nice 20cm x 20cm cube which fulfills all what I really demand from a beefy system, so yeah I hope the crisis will last a bit so I could get my money back
PS. yeah I’ve been laughing about it, but I’ve already sold the GPU from it with a gain! One day I’ve ready the article about GPU shortage so I thought let’s try and it was sold back on galaxus next day as 130% of the price 
What? Why?
I remember getting a computer with 4 GB RAM and it worked better than anything before. Then 8, and 16 brought only marginal improvements. Current work laptop has 32 GB and can’t tell it’s better than 16GB from the user experience perspective.
The software I usually use benefits from parallel threads until 16. 32 threads is only like 15-20% faster than 16 threads. 64 threads is basically the same as 23, no benefit at all. My impression is than software has been developed as fast as hardware.
Caveat: I connect to a virtual machine (hosted in a server) somewhere in a basement in Zurich to do calculations or compiling. The laptop at home could do it, but lots of noise, worrying about having data locally.
Software bloat for one thing. Just the amount of RAM that Chrome uses is insane.
3 year old PC, can’t be worth too much, esp. without a GPU, I’ll give you 200 CHF for it 
I use firefox, it allows to track memory use. Also, it sends tabs to sleep after a threshold of RAM is passed, called “unloading” or something like that.
I’m curious how much benefit if any compressed RAM gives. With NVMe 4 or even 5 the compressed RAM seems a gimmick, except if you care to prolong the SSD life a little bit.
what do you mean by compressed RAM? this reminded me of the DOS days “Ram Doublers”.
It’s sort of like the ram doublers but different. The “unused” used ram is being compressed instead of going right away to the disk. I think the original idea behind this was that the compression/decompression is faster than read/write to disk. Both mac and windows use this technique nowadays.
Well, Linux also has it, albeit not turned on by default like the other oses
I’m not sure if it can eventually swap to real disk though, have to read more about it
I found an article which explained why this happened:
In short, OpenAI, the company behind ChatGPT, secretly cornered the DRAM market by simultaneously striking deals with the 2 biggest DRAM manufacturers without letting anyone else know this was happening.
When the deals were announced and people realised that 40% of the memory was effectively taken off the market by OpenAI, panic set in, without these essential chips, companies would not be able to make and sell their products.
In a few months, RAM prices have tripled, product launches have been cancelled and the chaos is growing.
There’s some conspiracy theory too, since OpenAI bought unfinished wafers that they can’t even use yet, there’s speculation they did this to cripple their competitors:
And now time for the biggest twist of all, a twist that’s actually public information , and therefore should be getting discussed by far more people in this writer’s opinion: OpenAI isn’t even bothering to buy finished memory modules! No, their deals are unprecedentedly only for raw wafers — uncut, unfinished, and not even allocated to a specific DRAM standard yet. It’s not even clear if they have decided yet on how or when they will finish them into RAM sticks or HBM! Right now it seems like these wafers will just be stockpiled in warehouses – like a kid who hides the toybox because they’re afraid nobody wants to play with them, and thus selfishly feels nobody but them should get the toys!
And let’s just say it: Here is the uncomfortable truth Sam Altman is always loath to admit in interviews: OpenAI is worried about losing its lead. The last 18 months have seen competitors catching up fast — Anthropic, Meta, xAI, and specifically Google’s Gemini 3 has gotten a ton of praise just in the past week. Everyone’s chasing training capacity. Everyone needs memory. DRAM is the lifeblood of scaling inference and training throughput. Cutting supply to your rivals is not a conspiracy theory. It’s a business tactic as old as business itself. And so, when you consider how secretive OpenAI was about their deals with Samsung and SK Hynix, but additionally how unready they were to immediately utilize their warehouses of DRAM wafers – it sure seems like a primary goal of these deals was to deprive the market , and not just an attempt to protect OpenAI’s own supply…
It’s a bubble.
It’s great to increase production capacity to serve new customers. It’s weird to abandon current customers for new ones that might not exist next year.
Crucial is shutting down — because Micron wants to sell its RAM and SSDs to AI companies instead
Seems that AI has forced a Crucial turning point upon us where we will literally have no more memory. Perhaps only a recollection of it all.