OpenAI funding fears hit memory chip prices
Posted by LordAlfredo@reddit | hardware | View on Reddit | 131 comments
Non-paywalled link: https://finance.yahoo.com/sectors/technology/articles/openai-funding-fears-hit-memory-154500748.html
pfak@reddit
> could spell trouble for chip suppliers
Could spell trouble? They aren't bringing much (if any) new capacity online, hence the insane price increases. They knew this was going to be a short lived capacity constraint.
StarbeamII@reddit
There's a bunch of new fabs being built - e.g. Micron's New York fab, Micron's Singapore fabs, new SK Hynix fab in Yongin, but most won't be online for years.
Mr_Lucidity@reddit
I worked for one of these for almost 2 decades, the fabs had a big boom in announcing construction when the CHIPS act was signed, all competing for that gov funding. More fabs announced 2 or 3 years later after for the AI boom. It takes a minimum of 2-3 years to construct the build and another 2-3 to bring it up to full production. The ASML photolithography tools cost near $400 mil each and have a 3 year lead time, all the other tools are usually 18mo lead time, and of the 100-200 tools per fab construction can only install a handfull at a time.
DavidsSymphony@reddit
The ASML bottleneck in the chip industry is crazy, it's worse than the TSMC one.
smarlitos_@reddit
The EU should force ASML to open source their patents for the one time š or buy them out and do it (open source their IP), so that other companies can make the machines if possible and reduce that bottleneck. Theyāre really slowing down civilizational advancement.
frezzzer@reddit
Too bad ASML uses America parts to even work.
Eu people act like they hold the cards. You donāt USA still does. Could change but EU is declining with migrant issues and aging population.
smarlitos_@reddit
Nice rw reply š
StarbeamII@reddit
The EU is not going to give away one of its very few technology companies that are actually at the forefront of technology.
smarlitos_@reddit
Another EU L smh
ekeryn@reddit
ASML is one of the most important assets the EU - if not the most important asset - has in its pocket at the moment and you call it an L not wanting to give it up to everyone else? Their work wouldn't be picked up by just other Euro companies, it would be picked up by everyone else including the US and China.
It would cripple ASML and strip the EU of a security asset.
smarlitos_@reddit
So true
Wait_for_BM@reddit
Telling me that you have no idea what those terms means nor the implication of "open sourcing".
Open source does not mean no license restrictions either as you have the dozen of Open Source licenses ranging from MIT to GPL with different level of restriction of use.
Patents are a form of "open source" as they publish what/how their invention works in exchange for legal protections (contrast to Trade Secret in which nothing is disclosed i.e. closed sourced)
What you should ask for is RAND (Reasonable and non-discriminatory licensing) where by other companies can license some of their technologies fairly.
What good is having blue prints to build something when you don't have the experts behind the machines on setting things up or support?
smarlitos_@reddit
They should just post them on Reddit
Open source it to the fullest extent possible, you know what I mean. No legal protections, just market conditions and complexity on their side to maintain their monopoly.
Other companies will figure it out if ASML is forced to give up all the info they can reasonably cough up without hiring away their experts to other companies.
Thereās plenty they can write some docs on and post it to GitHub Reddit or whatever bs
Every government can do whatever, just depends on their IP laws. The Dutch government could force ASML. Plenty of good reason to do it and plenty of factors that would keep ASML in the game and able to pay off their upfront investment in all of this/the profit incentive to keep innovating remains.
Also, plenty of innovation happens without profit motive.
BlueSwordM@reddit
Won't work even if they do.
There is so much interconnected supply, so many companies helping each other, and such insane expertise that even open sourcing the entire stack would still be rather... complex to replicate.
smarlitos_@reddit
Sounds like they should do it anyway since ASML would continue to dominate anyway thanks to that interconnected supply and precise expertise. Sounds like they can keep their incentive to innovate while everyone else benefits from the IP.
BlueSwordM@reddit
I don't disagree, but not ASML, Denmark and the US would agree.
Jerithil@reddit
China has been trying to do it for a decade with billions invested and they still are years behind.
TechTechTerrible@reddit
These things are so complicated even with the designs the vast majority of manufacturing companies couldnāt produce a working example.
smarlitos_@reddit
Great reason to open source it anyway š on the off-chance someone can. If they canāt, then ASML retains their āincentive to innovateā according to people who are pro-monopoly and pro-IP/copyright.
ryanvsrobots@reddit
Ah yes just hand the keys to China very smart
smarlitos_@reddit
Yes, more competition is good š
censored_username@reddit
This isn't an issue you can solve like that. ASML isn't artificially slowing down supply, and also not the single responsible company. They're trying to expand as we speak with an entire second location, but it takes time, a lot of very expensive equipment, and highly educated people. Furthermore they also don't make everything in house. Their supply chain is already spread through many countries. You'll have to increase the production capacity of everyone along the chain.
Heck, many of the patents required aren't even owned by ASML but licensedfrom various research groups. It also took a monstrous amount of investment to even get EUV litho working at scale. Just opening it up wouldn't suddely mean that's not still required. If you want them to scale up, that investment is much more the issue than just the patents/IP.
And that opening up would be for what? The only reason the AI bros are competing so hard is because they all want to claim this new market. It's not to actually advance anything. That's why you don't see them investing in the lithography sector like the traditional players. They don't care about the advancement, they just want to be the only one who gets it when it happens.
smarlitos_@reddit
Theyāre not artificially slowing down supply, they just canāt produce enough machines by themselves. Need more competition, competitors will come up with more efficient methods if their IP is democratized.
SnooPets1826@reddit
Those aren't being built (just) with AI in mind and were likely inevitable.
Z3r0sama2017@reddit
Yep. Capacity for predicted growth, not chasing AI fairy unicorn bubble.
FlyingBishop@reddit
Acting like AI investment is a unicorn fairy bubble is part of the problem. People think LLMs == AI and ignore shit like Waymo which is actually growing and requires RAM same as all the other myriad AI systems that are actually being deployed but don't get lumped in with "AI."
JQuilty@reddit
LLMs and image/video generation are all the cocaine addled MBAs and stock traders care about and insist on forcing into every facet of life to impress their golfing buddies. That's why they're the poster child.
PastaPandaSimon@reddit
Some of them are being built largely to accommodate the AI bubble. For instance, Samsung went crazy with the P5 expansion, reviving a huge stopped project to chase the AI unicorn. It'll be Samsung's biggest and most advanced memory fab.
ITaggie@reddit
More accurately they're being built in responses to the oodles of cash being burned for the sake of AI. Selling tools in a gold rush and all that.
Jeep-Eep@reddit
And because if they can't get cost per unit down, they're going to be pantsed by the chinese memory operations, especially as they can already fab dies (though not necessarily at scale yet) comparable to the current gold standard, Hynix A die.
deep_chungus@reddit
kinda yes and no, before ai there was bitcoin, ai is just the current bubble inflating the shit out of prices.
after this there'll be others, trust
kittymoo67@reddit
yeah the fabs if anything were doing the smart thing and people act like this is gonna doom them?
greiton@reddit
yeah, they are actually looking at self driving vehicles and more IOT expansions according to a recent press release.
imaginary_num6er@reddit
Also Micron is not selling to consumers
StarbeamII@reddit
Hynix and Samsung don't sell DRAM directly to consumers either. All 3 will sell DRAM ICs to your Corsairs and G.skills of this world, who then put it on sticks and sell to consumers.
kittymoo67@reddit
plus theyre still selling to like dell etc
woopwoopscuttle@reddit
https://www.youtube.com/watch?v=jVzeHTlWIDY
May I remind you that these POSs have been a cartel forever.
pfak@reddit
I need no reminder. I've been through every single DRAM price fixing saga, and RAMBUS.
capybooya@reddit
I remember, Rambus for sure attracted a ton of drama and hate. But there were bastards on both sides of that conflict.
trowawayatwork@reddit
the biennial fire on the Thai factory lmao
Ratiofarming@reddit
biannual*
trowawayatwork@reddit
ackshually biannual means happens twice a year not once every two years..
Nuxij@reddit
It means both
toolverine@reddit
Biennial and biannual aren't synonyms becuase they describe different amounts of tine.
Biannual means twice per annum (a single year).
Biennial means every two years, which stems from gardening/ horticulture (annual, biennial, perennial) and was originally used to refer to the amount of years before replanting becomes necessary.
Nuxij@reddit
Biannual is like biweekly and bimonthly, it means twice in X or once every X. The meaning is ambiguous so there are other terms like biennial, bi-yearly, semi-annual, etc.
You're right there are terms like triennial and quadrennial and perennial, so biennial is naturally a great term when you want to be unambiguous.
reductase@reddit
Merriam-Webster: "a word that specifically refers to something that occurs every two years or that lasts or continues for two years."
Ratiofarming@reddit
TIL I guess
kingwhocares@reddit
It's going to be even more short-lived as there's over $20 billion in guaranteed sales in the Middle East that isn't likely to happen.
hitsujiTMO@reddit
Yeah, but OpenAI are backing out of the deal. So those wafers they bought aren't going anywhere and could potentially be sold at a loss.
This could be a huge loss for RAM vendors who just uprooted the RAM market for nothing.
W0LFSTEN@reddit
This is a short lived capacity constraint?
JimJimmington@reddit
In relation to the strategic planning and multiple year long build times, yes.
Creepy-Bell-4527@reddit
It's almost as if it was a bad business decision taking such an insane order from a customer that entirely depends on investor handouts to not crumble under its own extreme negative cash flow.
hackenclaw@reddit
I believe it when I see it, until then will not use chatGPT, will not buy new computers.
shroudedwolf51@reddit
Honestly, you shouldn't be using ChatGPT or any of the other regurgitative "AI" grifter tools regardless of prices of computer hardware.
Many_Career_9035@reddit
The frontier models are all excellent aids for many functions, but especially for software engineers. Ignore them at your own peril.
work-school-account@reddit
I've yet to get ChatGPT or Gemini to be useful for my research, and I've tried a lot. On the other hand, it perfectly solves problem sets and exams that I give to my students.
ResponsibleJudge3172@reddit
Research requires more research oriented tools.
For basic research, perplexity is far better, its more grounded in publications and always cites its sources.
Notebook LM is also quite good for fact finding and other preliminary stages of fact finding, etc
LordAlfredo@reddit (OP)
"Adapt or die". It's been true in technology for a very long time. Sears ignoring the internet is a case study in business school for a reason.
FlyingBishop@reddit
For the past couple years I think we've been in a continuous state where you might say "AI isn't useful for that" and 3 months ago you would've been right, but that's not true anymore, it's useful for that now. There are still lots of things it can't do but you can't assume you know what it can't do.
pmjm@reddit
I understand your position from an ethical perspective, but that just isn't realistic. These tools are here to stay and will be an integral part of everyone's lives basically forever.
Every day we're all exposed to AI generated text whether we realize it or not, and that's only going to get more prevalent in the coming decades, not less.
BaysideJr@reddit
Ideally with things like better compression we can run good enough models locally and not need these giant excessive data centers.
pmjm@reddit
I agree that it will get better, but the path to that requires what we're going through now. People need to use it, research needs funding, and we have to build it inefficiently first so we can optimize it later.
FlyingBishop@reddit
The path to running these things locally isn't likely to be compression, it's going to be the same miniaturization and power efficiency improvements that have been driving new capabilities in computing for decades.
And there are a lot of interesting models that can run locally but stuff equivalent to the frontier models, I think that's always going to require at least a terabyte of RAM, probably 5 years from now the typical frontier model will have more than that. And maybe we'll be running terabyte models locally but only if RAM prices come down due to demand for this sort of thing.
OverlyOptimisticNerd@reddit
I keep track of the system that I built over the summer via a PCPartPicker list. It was about $1,650ish before taxes. At its peak it hit near $2,300. Itās now under $2k. Most of the drop was the memory kit used.Ā
So prices are coming down, at least temporarily, but they arenāt close to actually recovering.Ā
Due-Cupcake-255@reddit
the cheapest 5090's are still well above 3k eur and climbing here.
07bot4life@reddit
What about storage pricing??
OverlyOptimisticNerd@reddit
In my latest look, storage was still elevated but flat compared to the last comparison. The majority of the price drop for my system was the memory.Ā
pmjm@reddit
Unfortunately the consensus seems to be that peak storage pricing hasn't happened yet.
OverlyOptimisticNerd@reddit
I agree, and I wasnāt trying to imply otherwise.Ā
fpsfiend_ny@reddit
Built a 8k usd pc a couple years ago.
The 300 dollar ram alone is NOW worth 1000.
Fuck this market and fuck these pedophile billionaires
Paed0philic_Jyu@reddit
It is not funding fears but the realisation that all those deals about booking supply are about pumping stock prices, not physical goods being acquired after paying the receipts.
twothreetoo@reddit
Micron says it has presold every chip they can make for the next two years and left half their orders unfilled. TSMC says similar. At the underlying hardware level, the stock prices are legitimate/undervalued.
xNailBunny@reddit
Yep, it's all just letters of intent and misleading announcements. The often cited "deal" to buy up a large percentage of the global DRAM supply is no different than Nvidia's "deal" to invest 100b into OpenAI.
DerpSenpai@reddit
Or Musks terafab. Musk has made BS the new norm. This should be highly ilegal
ITaggie@reddit
Anyone else remember Musk's "Hyperloop" promises? lmao
arahman81@reddit
"Man on Mars in 10 years"
FlyingBishop@reddit
It doesn't need to be illegal, deals do not happen overnight, there's due diligence etc. People need to not confuse letter of intent with a signed contract and funding secured.
(Now, Musk did say funding secured that one time and that was an illegal lie, but I don't think OpenAI or Nvidia has lied, people have just drawn bad conclusions from what they said.)
Due-Cupcake-255@reddit
nvidia is as conservative and drama free as you can be for the position they are in. Jensen is remarkably stable imho.
Due-Cupcake-255@reddit
it is illegal. He was infront of a court for it. but instead of getting a lifetime without parole he got nothing.
Loose_Skill6641@reddit
* openAI says it's buying 40% of all global RAM
* Nvodka says it's forecasting massive revenue uplift due to OpenAI
* RAM maker share price rise 50% on forecasts of supply shortage
* OpenAI announces it's unable to get money to buy 40% of global RAM
* Sharemarkets enter a depression and tank
Wide_Lock_Red@reddit
This all breaks down if people just dont listen to outlandish statements by tech CEOs.
Due-Cupcake-255@reddit
but they want to listen because they are profiting from it just as much as openai.
work-school-account@reddit
Tinkerbell economy.
The problem is that if everyone else listens, it doesn't matter whether you listen.
BioshockEnthusiast@reddit
Look man Forbes has to make their money by scaring the ever loving shit out of anyone with a 401k by regurgitating the lunatic ramblings of the greediest and most delusional people on the planet on a daily basis OK it's fine and it's good journalism please click on the ad now thank you.
kittymoo67@reddit
so the steammachine is back on the table then
cederian@reddit
Altman just sent some Letters of Intent to buy RAM and the market went nuts... not a single proper contract was signed.
GhostReddit@reddit
Memory manufacturers are starting to demand payment upfront for HBM/DRAM and flash, the days of everyone booking defensive orders everywhere and then half the order book disappearing when the market turns are over.
OAI's "announcement" isn't a purchase, but the actual buyers can't create the huge phantom demand they used to.
Criss_Crossx@reddit
Maybe it is just me, the tech crunch is a problem with a fix or workarounds.
Not enough storage space? Delete something. Your game is too large? Design around it and make it smaller. People don't have enough memory in their computers to handle everything? Well, it is time to overhaul the application and find ways to reduce the overall footprint. Maybe less ads?
The modern tech overhead to operate isn't optimized, we've been lax with it and oversold something new and more spacious. Take a look at browser tab usage, wow.
Ultimately, how much of it is unneeded anyway? Not everything needs to exist in the digital space.
/end rant
tgwombat@reddit
Most of what you listed would need to be performed by the software developers and they don't have a financial incentive to do so. There's a reason we got to where we are now, and it wasn't just because hardware was cheaper.
Criss_Crossx@reddit
I am aware. Companies notice when people stop using their products. They definitely have the capacity to make changes. We've been pushed constantly towards 'the upgrade' to monetize these trends.
We can not use things or choose to seek alternatives. I ditched microsoft easy enough, it wasn't that difficult to do. Historically the company has a track record for falling out of grace of the users with its products.
Humans are the product but not always by force.
tgwombat@reddit
Most of the worst offenders happen to also be the ones that have their respective industries locked down though. It might be easy for you or me to ditch Microsoft Office (or whatever they call it these days) or the Adobe suite, be we aren't where they make their money.
Criss_Crossx@reddit
Fully understand. It doesn't require that we fulfill a company's investment however. This isn't a life sentence towards a company and we don't have to always support what is used. We (the individuals) don't require every service out there to fulfill our lives. These problems we face aren't exactly equal across the planet either. Meaning, to some companies certain locations are their main target markets. But not everywhere, because it doesn't make monetary sense.
The example argument being: I own a car because I need it to get around. But not all places that exist require a vehicle, some folks get around just fine without and live full lives. A car is just the entry point for a life designed to use it. Infrastructure with massive amounts of money under every square foot, stretching globally. Entire industrial ecosystems piggyback off of its expanse just to stay available.
You mention MS Office 365-whatever, which is another very good example. Governments are beginning to push away from the MS ecosystem, some even have their own computing hardware (a-la RISC-V) they design systems around.
We have a choice. Even if it doesn't present itself easily, we have something.
To me that is the idea that keeps pushing me to find a way and make a change. Even if it means turning something off and never turning it back on.
tgwombat@reddit
All Iām trying to say is not to expect the hardware shortages to be solved by software optimizations where itās most needed.
Criss_Crossx@reddit
All I'm trying to suggest is there are options on the table. Follow the choices made available. Expect nothing to change.
tgwombat@reddit
I understand your ideological stance. But the issue with it is similar to the issue with preaching personal responsibility in the face of climate change. Itās nice, but itās not a solution to the actual problem.
Criss_Crossx@reddit
Personal responsibility does very little for change. Every company has you outclassed and outmatched by a huge amount.
You really don't get it.
tgwombat@reddit
Yes, this is my point. I'm going to stop responding now.
hackenclaw@reddit
we kinda stuck at 512GB to 2TB range for the same price range since 2011.
The only big diff we change is; we moved from HDD to nvme SSD; other than that storage for consumer space pretty much stagnant for 15yrs.
alexforencich@reddit
Tbh I would say that actually it is largely because hardware is cheap. If end users didn't have the resources to run the applications in question, then the software developers wouldn't make many sales. If memory and storage are cheap and available, then it's cheaper to omit various optimizations because people will still use the software, buying some more RAM or an extra SSD as necessary. But if storage is expensive, then less people will purchase and use the software in the first place. If RAM and storage costs remain high for long enough, I think eventually you will see real financial pressure to do some more optimization.
Intelligent_Bison968@reddit
I agree, some companies are making it actively worse. Whatsapp for example had windows app that used native windows libraries for GUI. It consumed around 100mb when app was used. They now changed it to use a webview and it's basically starting whole browser when you open the app and consuming at least 1gb of ram. And up to 3gb when using the app.
awwc@reddit
All fine solutions for people already owning computers.
GlammBeck@reddit
Has anyone read the article? It's nonsense. There is no source saying OpenAI has cancelled any DRAM orders, just speculation about it cutting back on costs after it closed down Sora. The sole data point for memory prices going down is that there is still a single 32GB DDR5 kit going for $370, presumably the same Crucial kit I've seen on Amazon for that exact price for months.
LordAlfredo@reddit (OP)
In terms of the broader business world, killing Sora is already enough to trigger speculation and adjusting value pricing. In a similar vein, look at how the market reacted after Nvidia clarified their OpenAI deal, they didn't even say there was no deal they just said it may not be $1b and that alone was enough for a small correction.
GlammBeck@reddit
I'm not saying there aren't warning signs or market movement, I'm saying the evidence they cite for falling prices is just a single kit on Amazon that hasn't actually fallen in price. There's just literally no news in this article.
GenZia@reddit
For one thing, I donāt think the Big Three (Micron, Samsung, and Hynix) have made any meaningful investments in new infrastructure or manufacturing capacity to meet demand, demand that may not even be sustainable to begin with.
They just raised prices, screwed over customers, and called it a day.
While I usually avoid throwing around the term ācartel,ā the DRAM industry often behaves like one. After all, the post-COVID DRAM oversupply is still fresh in my memory, no pun intended.
Frankly, I wouldnāt be surprised to see a similar glut emerge once the mania surrounding LLMs inevitably cools down.
Of course, Iāve been wrong before.
StrategyEven3974@reddit
You're so confidently uniformed and wrong it's wild.
nosurprisespls@reddit
Yeah, they're not building any new plants to meet AI demand. Whatever they're building or planed to build were planned before OpenAI buy up all the RAM.
soggybiscuit93@reddit
OpenAI's intent to order was done long after the AI boom started. The big 3 haven't been building out all this capacity over the last few years randomly - they forsaw increased AI demand early on, and the AI boom has been going on since before October 2025
StarbeamII@reddit
They're still opening a bunch of new fabs, though they'll take a while to come online:
Micron: * New York fab * 2 fabs in Boise, Idaho * Singapore expansions * Taiwan PSMC fab acquisition and conversion, though won't come online until H2 2027
SK Hynix: * Yongin fab * Cheongju fab rampup
Samsung: * Pyeongtaek P4 and P5
mdnkork@reddit
They all have been announced before the crazy price increase no? Maybe accelerated because of it but they are not a reaction to current prices.
soggybiscuit93@reddit
Well, they did Forsee increased demand due to the AI boom. Shortages can create panic orders that beget more shortage. Doesn't mean that's the true underlying demand. It just means customers could be bringing forward a whole year worth of orders to a single quarter so that they're not the ones caught with their pants down in a shortage.
Alot of this new capacity is in part due to the AI demand. balancing supply with demand is very difficult in a market where adding supply takes years while having too much supply can be devastating to your bottomline.
Vushivushi@reddit
Micron's PSMC fab acquisition was announced in January and they closed the deal in 60 days.
SK Yongin is a cluster with 4 fabs planned by 2050 so I guess you could say that.
It only had $6.5b committed for the first fab with one clean room funded. It now has $21.5b funded for all six clean rooms by 2030 and the local government allowed them to bump the total clean room area by an additional 50%.
So it's accelerated, but also expanded. Because there's a clean room shortage, existing fab upgrades are happening faster too.
Paliknight@reddit
The problem is storage was already relatively overpriced before this crisis considering tech advanced so much and it would have gotten cheaper. When prices come back down, it doesnāt mean theyāll be cheap or affordable like before unfortunately. For example, even if the 4TB 9100 pro was reduced by 50% (from 800 to 400), thatās still extremely overpriced when you compare it to the 4TB 990 pro that was selling for 200 on sale at one point.
soggybiscuit93@reddit
But was that $200 price even sustainable? Because this market goes through floods and droughts. Memory makers were pinching pennies because of post-COVID oversupply that saw some pretty unsustainably low prices too
FlyingBishop@reddit
Consumers really don't have much need for more than 4TB. Most consumers can make do with 1TB. Now, obviously, you can use limitless amounts but increasingly it's not "I'm going to pay 2x as much for 2x as much storage" but I will pay 2x as much for storage that's 2x as fast.
pmjm@reddit
I see what you're saying but that example isn't really a fair comparison. It stands to reason that the gen-5 drive would be significantly costlier than the gen-4 one. Every part of it from the nand to the controller is more complex to produce and there are more difficult problems to solve.
The prices fall when the newer generation comes out (with the weird exception of Nvidia 4000 gpus) and as the manufacturing process improves its yields.
soggybiscuit93@reddit
They just raised prices, screwed over customers, and called it a day.
Damned if they do, damned if they don't? What then what was the correct course of action for them to take? Build capacity for "demand that may not even be sustainable" or raise prices?
noahloveshiscats@reddit
As opposed to not raising prices, running out of stock and now a bunch of Wall Street suits pocket a 200% price difference that Micron, Samsung and Hynix could have taken instead to increase production?
mrandish@reddit
Well, yes. The RAM chip makers thought the same thing you did. That why they didn't make "any meaningful investments in new infrastructure or manufacturing capacity to meet demand"
They realized this was likely to be a 6 to 12 month blip and wasn't sustainable. The RAM industry has gone through fairly constant 'shortage/glut' cycles for decades. It takes years to bring new capacity online and they've learned the hard way not to react to shorter-term market fluctuations.
R-ten-K@reddit
You are, in fact, wrong.
Micron is building a \~$100 billion campus in upstate NY, and \~$30 billion facilities in Asia.
SK Hynix is fast tracking 3+ fabs in Korea.
Samsung is building new DDR lines in Korea and Texas
The entire industry is shitting bricks building lines to meet demand.
KARMAAACS@reddit
The grift is coming to an end, what will the YTers do once the cycle is over?
ITaggie@reddit
Whatever macguffin that ends up causing the next "tech boom"
Hsensei@reddit
Plenty of shady practices in tech. I'm sure they will be fine
KARMAAACS@reddit
Probably.
Bitter_Leather_8319@reddit
Contract vs spot prices people... that is it.
Contract prices will climb an additional 40% in Q2.
That makes Asus or Sony raise the prices.
And what we see now, is some panic (more like a need correction) in the spot prices.
hsien88@reddit
OpenAI never bought 40% of global memory capacity, it was mostly made up by bots on twitter (because Elon's feud with Sam Altman). Now the same bots are blaming OpenAI for something they never did.
TinFoilHat_69@reddit
Between SK and Samsung they are supposed to reduce production by 500,000 this year. While exiting consumer space entirely for data centers they forced Chinese companies to increase their prices because they donāt have competition in the consumer space..
bubblesort33@reddit
People said the RTX 5070ti would go out of stock, yet it's still on sale at my local retailer in Canada, MemoryExpress, for almost MSRP. Maybe 10% over. Either sales are incredibly low, or things were never really that bad.
Corpalz@reddit
Damn where do u see that, i wanna buy a 5070 ti but cant find any below $1400 on mem express or canada computers
nightstalk3rxxx@reddit
In germany prices are going down aswell, earlier this year 5070ti wasnt found for less than 1kā¬, now dropping back to "normal" prices slowly with cheapest models beginning at 900.
CaptainDouchington@reddit
Uh no. It spells disaster for openai and normalcy for the rest of the world.
Piss off.