Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Shells Virtual Desktop
BMail.ag - Secure Email Service
Server.net
CPLicense.net
VPS Server
Buy VPN
Vultr
VMs for AI
HostDare
ReliableSite White-Label Dedicated Hosting for Resellers
InterServer VPS
BMail.ag - Secure Email Service
Best VPN
High-Performance Bare Metal Server Solutions
Karvl.com
Server Mania Cloud Hosting
DataWagon Hosting
AlphaVPS Hosting
Evoxt.com
Clouvider
VPS Hosting with NVMe
Residential IPs in the US & 4G Mobile Proxies in EU & US with Unlimited Bandwidth
ReliableSite White-Label Dedicated Hosting for Resellers
Rabisu - Hosting Solutions
Shells Virtual Desktop
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

$364 DDR5 now $2,256 USD - costs more than my existing node

2»

Comments

  • fiberstatefiberstate Member, Patron Provider
    edited January 3

    @OpaqueRegistrant said:

    @fiberstate said:
    Many AI players will likely go bust just in shear CAPEX to revenue. However, AI gets better with scale, therefore demand for power and scale will not diminish likely ever. Arguably, compute is the true currency of the future. We are still in the vacuum-tube era of the transistor, exploring how far brute force increases in clock cycles and linear algebra can take humanity.

    You can't get to the moon by climbing taller and taller trees. The power of LLMs is not that they can do things well. The power is that they are good con men.

    LLMs have gotten better with each iteration, this is a fact. Where are LLM's being useful to the modern digital infrastructure being con men? In the last two years, searches have integrated LLM's into results, content creation has been automated, deep analytics have been vastly improved, agents have been created to handle tasks and the world has turned to them for productivity in most professions. Wouldn't call that a con. Amount of capex speculation? Possibly, but there will be winners and losers in this race, like anything else important.

    Thanked by 1Shakib
  • Selling these, bulk only if anyone is interested.
    1024x 128GB M321RAGA0B20-CWK Used (few months only) (Lifetime warranty)
    260x 96GB MTC40F204WS1RC48BT NEW 2Year Warranty

    Thanked by 1Shakib
  • @Blazingfast_IO said:
    Selling these, bulk only if anyone is interested.
    1024x 128GB M321RAGA0B20-CWK Used (few months only) (Lifetime warranty)
    260x 96GB MTC40F204WS1RC48BT NEW 2Year Warranty

    deadpool incoming?

    Thanked by 1Shakib
  • RubbenRubben Member

    @MikeA said:
    It be how it do bro. Get on your knees and bow to our AI overlords and billion dollar companies hogging memory contracts for the next 5 years. Us peasants are well screwed.
    Edit: Talk to ya when that same memory is not $364, is not $2256, but is $5000.

    got on my knees for you now what do i do Mike?

    Thanked by 1Shakib
  • @fluffernutter said:

    @Blazingfast_IO said:
    Selling these, bulk only if anyone is interested.
    1024x 128GB M321RAGA0B20-CWK Used (few months only) (Lifetime warranty)
    260x 96GB MTC40F204WS1RC48BT NEW 2Year Warranty

    deadpool incoming?

    Not really. Just an investment/business idea that ended up not going forward and since the market exploded we have decided to sell all of them.

  • emghemgh Member, Megathread Squad

    @OpaqueRegistrant said:

    @fiberstate said:
    Many AI players will likely go bust just in shear CAPEX to revenue. However, AI gets better with scale, therefore demand for power and scale will not diminish likely ever. Arguably, compute is the true currency of the future. We are still in the vacuum-tube era of the transistor, exploring how far brute force increases in clock cycles and linear algebra can take humanity.

    You can't get to the moon by climbing taller and taller trees. The power of LLMs is not that they can do things well. The power is that they are good con men.

    The amount of time I know that I save not writing boilerplate code anymore tells me otherwise.

    Thanked by 1Shakib
  • Hopefully we see the real return of Low End Boxes with 128MB RAM now

    Thanked by 2Shakib tentor
  • sliixsliix Member

    @hostnoob said:
    Hopefully we see the real return of Low End Boxes with 128MB RAM now

    We'll need custom/stripped version OSes for RAM as low as this right?

    Thanked by 1Shakib
  • ShakibShakib Member, Patron Provider

    @ralf said:

    @Shakib said:

    @MikeA said:
    It be how it do bro. Get on your knees and bow to our AI overlords and billion dollar companies hogging memory contracts for the next 5 years. Us peasants are well screwed.
    Edit: Talk to ya when that same memory is not $364, is not $2256, but is $5000.

    I have had so many plans for 2026, I am just sad now. :'(

    If you're making decent bank now and can have a comfortable lifestyle, it's OK to have a year or two without expansion. It's only if you think RAM prices will continue to stay this high for the long term that you need to just bite the bullet and pony up.

    At the moment everyone is gearing up because they are starting to invest in AI capacity. I suspect there's going to be a fairly sudden drop off in demand when they've reached saturation in customer base and revenue stops growing, and companies realise how much capacity they already have and that it's "good enough" for their future development as well.

    I will see if I can find complete systems with RAM at previous prices used or new.

    If not, I will just enjoy life and expand other services.

  • uekixuekix Member

    As much as I hate and blame AI for every problem today, i need it for my deepfake addiction.

    Thanked by 1Shakib
  • TerranodeTerranode Member, Host Rep

    Just as I was thinking about selling some servers with 384 GB of RAM that I acquired recently, I checked on eBay and the memory alone costs $1,800 — it’s crazy

    Thanked by 1Shakib
  • @fiberstate said:

    @OpaqueRegistrant said:

    You can't get to the moon by climbing taller and taller trees.

    LLMs have gotten better with each iteration, this is a fact.

    These facts are not in contradiction.

    Thanked by 1Shakib
  • Waiting until I can sell half my computer's RAM to cover the price of the rest of the computer.

    It won't happen though. This will be a temporary (months long) blip just like COVID toilet paper.

    Thanked by 1Shakib
  • @OpaqueRegistrant said:
    Waiting until I can sell half my computer's RAM to cover the price of the rest of the computer.

    It won't happen though. This will be a temporary (months long) blip just like COVID toilet paper.

    Toilet paper literally grows on trees. RAM not so much

  • emghemgh Member, Megathread Squad
    edited January 5

    @OpaqueRegistrant said:

    This will be a temporary (months long) blip just like COVID toilet paper.

    No

    Temporary yes

    A few months no

  • jonbeardjonbeard Member, Patron Provider, Veteran

    We have been amortizing the RAM market increases for customers over payments up to 50% of their term limit, as an NRC.

    If you need some help with infrastructure during this time, let me know. Happy to share numbers with you.

    Our OEM vendors are advising expect 2-3% increase almost DAILY up to 30% increase each quarter, until atleast quarter 1 of 2027.

    Thanked by 2Shakib Noct
  • Its my belief that there is already surplus GPU supply. My reason being that it has been months since I have seen any sort of rate limiting on OpenAI or Anthropic. While AI certainly has carved out its use cases, companies like Openrouter are seeing largely stagnant token usage growth. There is a lot of GPU capacity coming online in Q1 this year, and I am hoping that once OpenAI sees their GPU utilization running at 45%, it will produce a shakeup in the markup that brings things like RAM prices back down.

    Thanked by 1Shakib
  • servarica_haniservarica_hani Member, Patron Provider

    @eezcloud said:
    Its my belief that there is already surplus GPU supply. My reason being that it has been months since I have seen any sort of rate limiting on OpenAI or Anthropic. While AI certainly has carved out its use cases, companies like Openrouter are seeing largely stagnant token usage growth. There is a lot of GPU capacity coming online in Q1 this year, and I am hoping that once OpenAI sees their GPU utilization running at 45%, it will produce a shakeup in the markup that brings things like RAM prices back down.

    based on my understanding openAI does not take all RAM because they need it
    they take it because they dont want competitors to use it

    They are trying to starve them

    Also it is fishy that non of the ram manufacturers are ramping up production for this crazy demand

    it feels this scarcity is somehow engineered to starve some players in the AI market and we are just flies in this game

    Thanked by 4Shakib Noct eezcloud ralf
  • AlexBarakovAlexBarakov Patron Provider, Veteran
    edited January 6

    @servarica_hani said: Also it is fishy that non of the ram manufacturers are ramping up production for this crazy demand

    They very well know that the bubble will blow. In previous component deprivation periods, different manufacturers have ramped up production, leading to eventual oversupply and losses when they have to sell at cost. They don't want to sell at cost, they prefer to keep scarce stock and fully predictable income. It's not always about momental wins, it's about predictability of the business.

    Thanked by 1Shakib
  • hostnoobhostnoob Member
    edited January 6

    @sliix said:

    @hostnoob said:
    Hopefully we see the real return of Low End Boxes with 128MB RAM now

    We'll need custom/stripped version OSes for RAM as low as this right?

    Maybe

    Back in the day, Debian (9/10 maybe) + lighttpd + PHP + mySQL/MariaDB could easily be tuned to run on a tiny VPS (maybe 64MB but definitely 128)

    There was no crap with containers or needing other apps requiring GBs of RAM to install your apps

    It was apt and you would configure everything your self

    Thanked by 1sliix
  • fiberstatefiberstate Member, Patron Provider
    edited January 6

    @servarica_hani said:

    @eezcloud said:
    Its my belief that there is already surplus GPU supply. My reason being that it has been months since I have seen any sort of rate limiting on OpenAI or Anthropic. While AI certainly has carved out its use cases, companies like Openrouter are seeing largely stagnant token usage growth. There is a lot of GPU capacity coming online in Q1 this year, and I am hoping that once OpenAI sees their GPU utilization running at 45%, it will produce a shakeup in the markup that brings things like RAM prices back down.

    based on my understanding openAI does not take all RAM because they need it
    they take it because they dont want competitors to use it

    They are trying to starve them

    Also it is fishy that non of the ram manufacturers are ramping up production for this crazy demand

    it feels this scarcity is somehow engineered to starve some players in the AI market and we are just flies in this game

    More likely the fact that the only three major memory manufacturers in the world saw the profits in HBM and decided to re-tool and focus the majority of production on HBM ram. Inference will continue to also drive demand for standard ram. This shortage will continue to get worse until providers are in a dystopian desert fighting for the last 8G DDR3 modules mad max style all while LET users hunt for the $7 yr deals.

    Thanked by 2Shakib ozonti
  • icemaniceman Member

    @fiberstate said:

    @servarica_hani said:

    @eezcloud said:
    Its my belief that there is already surplus GPU supply. My reason being that it has been months since I have seen any sort of rate limiting on OpenAI or Anthropic. While AI certainly has carved out its use cases, companies like Openrouter are seeing largely stagnant token usage growth. There is a lot of GPU capacity coming online in Q1 this year, and I am hoping that once OpenAI sees their GPU utilization running at 45%, it will produce a shakeup in the markup that brings things like RAM prices back down.

    based on my understanding openAI does not take all RAM because they need it
    they take it because they dont want competitors to use it

    They are trying to starve them

    Also it is fishy that non of the ram manufacturers are ramping up production for this crazy demand

    it feels this scarcity is somehow engineered to starve some players in the AI market and we are just flies in this game

    More likely the fact that the only three major memory manufacturers in the world saw the profits in HBM and decided to re-tool and focus the majority of production on HBM ram. Inference will continue to also drive demand for standard ram. This shortage will continue to get worse until providers are in a dystopian desert fighting for the last 8G DDR3 modules mad max style all while LET users hunt for the $7 yr deals.

    Well, not sure how you do it but, you are the only one in LET providing most cheapest prices ever. And because of such prices, don't blame let users for demanding $7 offers. If you can offer a dedicated server for $40+/month with 64/128G ram, 2-4TB NVMe and unmetered traffic, why is then so terrible requesting a vps with 2-4GB ram, 20-40GB NVMe for $7/y?

  • fiberstatefiberstate Member, Patron Provider

    @iceman said:

    @fiberstate said:

    @servarica_hani said:

    @eezcloud said:
    Its my belief that there is already surplus GPU supply. My reason being that it has been months since I have seen any sort of rate limiting on OpenAI or Anthropic. While AI certainly has carved out its use cases, companies like Openrouter are seeing largely stagnant token usage growth. There is a lot of GPU capacity coming online in Q1 this year, and I am hoping that once OpenAI sees their GPU utilization running at 45%, it will produce a shakeup in the markup that brings things like RAM prices back down.

    based on my understanding openAI does not take all RAM because they need it
    they take it because they dont want competitors to use it

    They are trying to starve them

    Also it is fishy that non of the ram manufacturers are ramping up production for this crazy demand

    it feels this scarcity is somehow engineered to starve some players in the AI market and we are just flies in this game

    More likely the fact that the only three major memory manufacturers in the world saw the profits in HBM and decided to re-tool and focus the majority of production on HBM ram. Inference will continue to also drive demand for standard ram. This shortage will continue to get worse until providers are in a dystopian desert fighting for the last 8G DDR3 modules mad max style all while LET users hunt for the $7 yr deals.

    Well, not sure how you do it but, you are the only one in LET providing most cheapest prices ever. And because of such prices, don't blame let users for demanding $7 offers. If you can offer a dedicated server for $40+/month with 64/128G ram, 2-4TB NVMe and unmetered traffic, why is then so terrible requesting a vps with 2-4GB ram, 20-40GB NVMe for $7/y?

    Oh agreed. Not stating users shouldn't be looking for deals. Just bemoaning the current hardware situation like everyone else. Don't hate the player, hate the game.

    Thanked by 1Shakib
  • icemaniceman Member
    edited January 6

    @fiberstate said:

    @iceman said:

    @fiberstate said:

    @servarica_hani said:

    @eezcloud said:
    Its my belief that there is already surplus GPU supply. My reason being that it has been months since I have seen any sort of rate limiting on OpenAI or Anthropic. While AI certainly has carved out its use cases, companies like Openrouter are seeing largely stagnant token usage growth. There is a lot of GPU capacity coming online in Q1 this year, and I am hoping that once OpenAI sees their GPU utilization running at 45%, it will produce a shakeup in the markup that brings things like RAM prices back down.

    based on my understanding openAI does not take all RAM because they need it
    they take it because they dont want competitors to use it

    They are trying to starve them

    Also it is fishy that non of the ram manufacturers are ramping up production for this crazy demand

    it feels this scarcity is somehow engineered to starve some players in the AI market and we are just flies in this game

    More likely the fact that the only three major memory manufacturers in the world saw the profits in HBM and decided to re-tool and focus the majority of production on HBM ram. Inference will continue to also drive demand for standard ram. This shortage will continue to get worse until providers are in a dystopian desert fighting for the last 8G DDR3 modules mad max style all while LET users hunt for the $7 yr deals.

    Well, not sure how you do it but, you are the only one in LET providing most cheapest prices ever. And because of such prices, don't blame let users for demanding $7 offers. If you can offer a dedicated server for $40+/month with 64/128G ram, 2-4TB NVMe and unmetered traffic, why is then so terrible requesting a vps with 2-4GB ram, 20-40GB NVMe for $7/y?

    Oh agreed. Not stating users shouldn't be looking for deals. Just bemoaning the current hardware situation like everyone else. Don't hate the player, hate the game.

    True, seems you hate the game from your previous replies here, but you keep playing it. So, despite price increase on RAM/NVMe etc your game goes on, but you keep spitting on that game while playing it. What am i missing :)
    Btw, i have 3x of your dedicated servers at $47/month and like it.

    Thanked by 2fiberstate Shakib
  • common 640KB is more than enough for your PC.

    Thanked by 1Shakib
Sign In or Register to comment.