- cross-posted to:
- us_news@lemmygrad.ml
- usa@midwest.social
- cross-posted to:
- us_news@lemmygrad.ml
- usa@midwest.social
You must log in or register to comment.
well, at least it is computer hardware, which famously keeps its value over long periods of time
This is one of the most absurd aspects of the whole AI build out in the US. Like this tech is still rapidly evolving, and we don’t even know that current hardware is optimal for running modern LLMs. In fact, it’s almost certainly isn’t given that we already see stuff like this being developed. So, the US ended up investing astronomical sums of money into producing vast amounts of first gen chips that will likely be obsolete in a couple of years. Meanwhile, the part that is actually needed is power generation since whatever future chips are made are still going to need power, and that aspect has been almost entirely ignored. Incidentally, the situation appears to be completely opposite in China. There’s been a massive investment in the grid, but the data centre build out has been very modest.
I am convinced that it is an effort to control who has access to AI and compute in general more than it is for the AI.
They wanted to make sure that they didn’t have any competition and that you would have to rent from them.
I am 10000% in belief of this. Its not tinfoil hat. They want to own your compute. Then youre absolutely beholden to them, and they can control the narrative and sow dissent even better.
Log onto bezosnet with your faceid and blood sample please.
What’s his face from OpenAI has said over and over again, almost as a point of pride, that they view it as an utility and the goal is to put it on a meter
Information as a service .
PUUUUKKKEEEE
I’m just hoping I live to see all these GPUs go for pennies on the dollar when the bubble pops.
Gonna do a lot of video processing and rendering
yeah I’d like to pick up a couple of H200s on the cheap :)
I’m hoping that there’s eventually a massive glut of secondhand datacenter equipment for running local models.
same
Huawei accelerators are starting to show some muscle I believe GLM5.3 runs on them. The West wants to fully monopolize AI but China is starting to already catch up.
However Huawei also doesn’t seem to sell their AI accelerators the public market outside of China. Not sure why they don’t want gain market share there.
I’m guessing it’s a question of capacity. They’re obviously going to prioritize domestic market first, but I expect in a few years we’ll see exports ramp up.
Getting some chips out is important to grow a base and get software support going. Maybe deployment still requires too much specific software for now. But while China is catching up it’s a bit sad that few can get their hands on the Huawei GPU’s for now.
I think GamersNexus got one quite a while ago but didn’t manage to get it to work? Haven’t seen an update on it anyway
Given how quickly China tends to ramp tech up, I’m fairly confident that we’ll be seeing cheap hardware on the market sooner than later.
China already has the dirt on all its citizens. That’s why there’s no rush for them. Usa wants to jam as many in as possible before the bubble pops or the public boycotts them. These I bet are mostly for surveiling citizens.
lmfao imagine thinking that western regimes don’t already have dirt on all their citizens. People literally carry a spying device in their pockets everywhere they go.
My gpu is holding it’s value near msrp or above and my ram has increased in value. So… :|
So basically just inflating prices to keep the AI bubble bubbling. Amazon sponsoring Nvidia which then sponsors Amazon in return.
Surely the mark of healthy business practices!
Ed Zitron called it at least a year ago. He recently released an estimate of how many there are:
As of September, likely about $234bn worth (of the $561.5bn of Nvidia GPUs purchased since 2023) are sitting in warehouses or unpowered data centers.
I recall him mentioning it in March, and it wondering if it was just a case that the rate of installation of new compute was rising fast, and so these companies needed a big buffer of GPUs to make sure they had something to install. And so I expected that what we would see is that maybe these buffers of GPUs in warehouses would grow, but the installation rate would grow faster. But now that the numbers are in, it seems like the gap is widening in the opposite direction - warehousing is accelerating faster than the acceleration of installation of GPUs. Every month brings new absurdities.
Ed needs to be in office or something making a difference. Hes on fucking point at every turn and ive never seen him be wrong.
The people most suited to public governance are never the ones that actually want to do it. And vice versa, unfortunately.
Exactly therein lies the issue with humans.
Many people may see this and assume its just them drinking their own koolaid essentially but there is a bit of a twisted logic to it. They’re basically just buying so many of these to keep the foundries happy to make them. They do have this delusion that they’ll need more of them, but its more that they want production to increase so they are creating demand. Which is actually WORSE for the economy, because once their bet fails not only are these GPUs worthless but the foundries have built out production capacity that is now worth way less.
‘We bought 600 train engines and 8000 rail cars, because we will probably figure out how to build 800,000 miles of new rail lines, somewhere, at some point.’
‘Also the train engines and rail cars we bought are guaranteed to be horrendously inefficent within 2 years.’
‘Also also, the train engines are highly fuel intensive and no we are not planning to fund any new fuel production.’
‘We are very smart.’
‘Why yes, our future rail line traffic demand projections are based on the “thoughts and prayers” model, thanks for asking!’
The big difference is that rail lines and rail cars will have value for many decades after being built. Meanwhile, chips are a notoriously depreciating asset that’s likely to become worthless in a year or two.
Yep, the analogy isnt perfect, and it is amazing to me that in reality, its even fucking worse.
It would have to be more like a disintegrating or time delayed self destructing rail engine…
Give them a break, it’s not like they’re a big cloud services company
Just remember these fuckers want to live forever, and then everything makes sense.
They expect AI to be able to grant them inmortality, either biologically OE digitally.
Their ego is so grand that maybe existing as a AI replica of themselves is enough.
If Amazon and the rest were around during the Tulip Boom and Bust, we would still be trading tulips for houses because they can manipulate this shit indefinitely.
