Is it better to survive the crash and suffer, or just die in the initial turmoil? At this point I dont know.

The fact that none of these "AI" companies are making any profit whatsoever and are being kept afloat by having their debt traded and burning through investor funding means that something is going to give.

It reminds me very much of the dot com bubble. Everyone wanted in on it, but the vast majority of players were making nothing but enormous losses.

There are some useful avenues for "AI" - we run the high performance computing facility for an English university and there are definite, specific workloads that can benefit.

But the way it is currently being sold as a universal panacea and touted as the saviour (or destroyer) of mankind? Crazy, absolutely crazy.

Once, men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them.

megatron-uk wrote on 2025-12-27, 18:10:

The fact that none of these "AI" companies are making any profit whatsoever and are being kept afloat by having their debt trade […]
Show full quote

The fact that none of these "AI" companies are making any profit whatsoever and are being kept afloat by having their debt traded and burning through investor funding means that something is going to give.

It reminds me very much of the dot com bubble. Everyone wanted in on it, but the vast majority of players were making nothing but enormous losses.

There are some useful avenues for "AI" - we run the high performance computing facility for an English university and there are definite, specific workloads that can benefit.

But the way it is currently being sold as a universal panacea and touted as the saviour (or destroyer) of mankind? Crazy, absolutely crazy.

No further analysis than this is needed. AI bros simply bookmark this post and come back to it in 10 years when you have forgotten your own complicity in the madness.

Oh it's 110% going to be a thing. The cat is oit of Pandora's box and the general public is FAR too lazy to NOT use Ai slop. Nor do they care. It is what it is. The companies are fighting each other to ne the next google. Thats it. Once the bubble pops, it determines who our new overlord is.

Ive given up trying to steer the ship, I'm putting my energy into survival now.

luckybob wrote on 2025-12-28, 01:31:

Ive given up trying to steer the ship, I'm putting my energy into survival now.

Ain't no LLMs running on no Quad-PPro workstation, amirite???

😄

If i can shoehorn a modernish gpu into one... I probably could run some small models.

luckybob wrote on 2025-12-28, 02:03:

If i can shoehorn a modernish gpu into one... I probably could run some small models.

Yeah, after I typed that I thought about how much RAM one of those old rigs can probably hold and I'm sure it'd be enough for a very small model.

Challenge accepted? 😮

Im missing 2 OD processors. I keep losing ebay auctions as they go above my maximum bid of $420.69.

In the uk , i bought 64gb 6000 cl 30 for 230 pounds ; this was in Fubruary 2025. Now , a year later almost , the same 64gb goes for 840 punds. That is MORE than triple.

It is insane.

For gaming , i don t need 64 gb , but i thought to get 64 for future proofing. Glad i did.

Now , depending on what you want to do with your pc , you could probably get a ddr3 or 4 machine and still be more than fine.

The situation is a real one, and is also affecting enterprise supply chain at NVME level. I work in IT enterprise, and one of my customer is telling me both Dell and HPE can't guarantee NVME disks, but stopped dealing with SSDs, so in case or storage you now are SAS, SATA (!!!) or out of luck. This is a big problem not just with new orders, but for stocking and replacement, as NVME tend to degrade overtime.
Somebody talked about ai land lords in the first posts on this thread: I cannot disagree.

Bet my 50+ 1Mb 36pin modules resting in the basement for the last 30 years will pay for my retirement, 🤣

This is all insanity and what is going to happen is the AI bubble will burst and all of these companies will have to deal with the massive hangover from imbibing the strong drink of reconfiguring the entire computing economy for the sole benefit of what should be a small percentage of it.

luckybob wrote on 2025-12-27, 18:13:

Once, men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them.

Can't remember where I read this (probably Slashdot) about the gambit afoot:

By making hypothetic catastrophe the center of public discourse, architects of AI systems have positioned themselves as humanity’s reluctant guardians, burdened with terrible knowledge and awesome responsibility. They have become indispensable intermediaries between civilization and its potential destroyer; a role that, coincidentally, requires massive capital investment, minimal regulation and concentrated decision-making authority.

And I for one welcome our new over-hyped-lords.

I guess people went insane too because someone just snatched my 16 GB Fury 6400 module at the warehouse of the shipping company.
"Reason of non-delivery: Lost parcel under investigation"
Someone just couldn't resist that black gold. 😁

luckybob wrote on 2025-12-28, 02:03:

If i can shoehorn a modernish gpu into one... I probably could run some small models.

and everyone said I was crazy when I started buying up the Radeon HD4650 and Geforce 9500/210/610 PCI cards 🤣

J/K only got a couple, the geforce though, nVidia's dev system for embedded AI vision models was basically one of those bolted to an atom.

Though actually depending on model it doesn't even take much of a CPU, because they've got some for ARM, ESP32, even Arduino AVR boards. But those maybe as dumb as fuck and replicated in function with a playing card sized analog computer board. For some things involving velocities and trajectories, digital was a dumber way of doing it, dumber still in high level language environments, and absolute amoeba level stupidity to be doing in AI. (and expecting repeatable real time output)

Well... I will have 6 "cores". 4gb of ram... I should at least be able to run a significant text LLM. I know I can shoehorn some later nvidia video card into the thing, even if its like a 3050 or something....

just when you thought it had peaked

https://pcpartpicker.com/trends/price/memory/ … r5.6000.2x16384

Wow, now I dunno whether to install the 2x16 I got in the summer, or cash in, or wear them on a chain around my neck 🤣

Funny how the 5200 looks a tad above the 5600 though.

SSD looked like they might have doubled also.

Edit: I would only bet on a 6 month duration though... A few years back, there was supposedly a lack of 20nm fab capacity in February, with predictions in all the industry blogs and journals that I could find that nobody much would have fabs online before that time the following year, and due to backlog, supply would still be constrained... anyway, what happened was that some fabs came online in late Q2, Q3 the dam burst, and Q4 had spare capacity going cheap. So I am expecting much the same thing, some fabs spinning up right now to take up slack for RAM dies, and June or July the supply loosens and turns into a glut by December... maybe. Post covid world is a bit strange though so who knows.

So, according to that graph, a quintupling in cost since Sept 2025? $500+ for 32gb and over $1000 for 64gb... absolutely batshit insane. This is worse than the hyperinflation in Germany after WWI and seems to be a very, very bad thing.

Back in Black Friday 2024, Amazon offered a 4TB TLC SSD with DRAM cache for $150 -- a good quality drive. Now, 4 TB QLC without DRAM cache (bad quality) drive sells for $449+ on Amazon. I'm not even sure if it's possible to buy a TLC SSD any longer. Ridiculous.