DRAM Pricing Jumps 50%, Only 70% of Orders Getting Filled

“DDR5 RAM Prices Skyrocket 761% In A Year In Japan, As 128 GB Corsair Kit Now Costs As Much As A High-End Gaming PC​


Crucial-DDR5-Memory-728x485.jpg


Soaring DDR5 memory prices are affecting the hardware market significantly, making it impossible to build or upgrade PCs.

Akihabara Price Survey Reveals Another Sharp Increase in DDR5 Memory Prices With Some Kits Witnessing Over 60% Price Increase Since Last Month​

DDR5 RAM prices have gone up drastically in the last few months. Every week or so, we see most DDR5 RAM kits skyrocketing in price, and this is universal across all regions. In Japan, the latest price survey from AKIBA PC Hotline revealed that the new price hike has made PC building more difficult than ever. At around 4-5X higher prices, most DDR5 kits were already unaffordable, but now they have seen another sharp rise of up to 62% compared to last month.”

Source: https://wccftech.com/ddr5-ram-prices-skyrocket-761-in-a-year-in-japan/
 
DukenukemX said:
There's some news that the AI bubble is popping. Alphabet loses $482 Billion in market value, biggest drop in history Or is everyone being doomers?
Click to expand...
A bit of the reverse, the ai popularity is the reason Google search is not growing fast, which was always the expected and why google has been cheap the last 3 years.

People talkign to an AI instead of spending time in a browser-google search with blue link was always a big risk (and the reason why Google was keeping those LLM for themselve without a public chat access, waited for others to force them, fearing cannabilization).

I.e. the cloud-datacenter-ai part of google did phenomenal last quarter, the issue is search (under 50% of their revenues for the first time in history now) and a big challenge for them, the more AI do well, the more tradtionnal google search get hurt.

google cloud grew revenues more than 100% year over year, the profit tripled.
 
Last edited:

“Sam Altman Thinks Your PC Is Wasting Precious RAM​

Gilmoure 38 minutes ago		25
"Looks like OpenAI CEO Sam Altman wants to own all the computing resources and rent out terminals to the world as a way of propping up his business plan, now that debt-ridden AI companies are teetering on the edge," writes ancient Slashdot reader Gilmoure. Tom's Hardware reports: In an interview over beers with the YouTube channel Cheeky Pint, Altman said that local inference, and, by extension, people having DRAM in computers at home, is "quite terrible." When discussing local inference on home computers with open-source models, Altman discussed the assumption that people will be able to run local AI on their own computers, the ones they may already own or want to build.

Then he explained why that's a bad idea (at the 29-minute mark): "From a global allocation perspective, actually, it's quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer and being used 5% of the time," Altman said. "You want maximum utilization in the cloud if you're trying to deal with this very supply-constrained environment. And somehow many people's intuition is the opposite of that."

While OpenAI does have some open-source models, its primary business is in closed-source models that run in data centers, so he's really defending the company's business model. OpenAI isn't exactly threatened by some laptops and gaming rigs just yet, but local AI could mean people one day paying less for subscriptions to models like GPT-6 Luna if they have the power to run them at home. Altman's talk suggests that the company would prefer that more RAM go to OpenAI's data centers rather than home computers, which also sparks enthusiast fears of not being able to own your own hardware or software, but rather rent an agent and a computer in the cloud.”
 
erek said:

“Sam Altman Thinks Your PC Is Wasting Precious RAM​

Gilmoure 38 minutes ago		25
"Looks like OpenAI CEO Sam Altman wants to own all the computing resources and rent out terminals to the world as a way of propping up his business plan, now that debt-ridden AI companies are teetering on the edge," writes ancient Slashdot reader Gilmoure. Tom's Hardware reports: In an interview over beers with the YouTube channel Cheeky Pint, Altman said that local inference, and, by extension, people having DRAM in computers at home, is "quite terrible." When discussing local inference on home computers with open-source models, Altman discussed the assumption that people will be able to run local AI on their own computers, the ones they may already own or want to build.

Then he explained why that's a bad idea (at the 29-minute mark): "From a global allocation perspective, actually, it's quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer and being used 5% of the time," Altman said. "You want maximum utilization in the cloud if you're trying to deal with this very supply-constrained environment. And somehow many people's intuition is the opposite of that."

While OpenAI does have some open-source models, its primary business is in closed-source models that run in data centers, so he's really defending the company's business model. OpenAI isn't exactly threatened by some laptops and gaming rigs just yet, but local AI could mean people one day paying less for subscriptions to models like GPT-6 Luna if they have the power to run them at home. Altman's talk suggests that the company would prefer that more RAM go to OpenAI's data centers rather than home computers, which also sparks enthusiast fears of not being able to own your own hardware or software, but rather rent an agent and a computer in the cloud.”
Click to expand...

“OpenAI CEO says it's 'quite terrible' for your PC to have RAM — Sam Altman says data centers offer better utilization​


With data centers full of RAM, your PC is apparently the problem.”

https://www.tomshardware.com/tech-i...an-says-data-centers-offer-better-utilization
 
erek said:

“Sam Altman Thinks Your PC Is Wasting Precious RAM​

Gilmoure 38 minutes ago		25
"Looks like OpenAI CEO Sam Altman wants to own all the computing resources and rent out terminals to the world as a way of propping up his business plan, now that debt-ridden AI companies are teetering on the edge," writes ancient Slashdot reader Gilmoure. Tom's Hardware reports: In an interview over beers with the YouTube channel Cheeky Pint, Altman said that local inference, and, by extension, people having DRAM in computers at home, is "quite terrible." When discussing local inference on home computers with open-source models, Altman discussed the assumption that people will be able to run local AI on their own computers, the ones they may already own or want to build.

Then he explained why that's a bad idea (at the 29-minute mark): "From a global allocation perspective, actually, it's quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer and being used 5% of the time," Altman said. "You want maximum utilization in the cloud if you're trying to deal with this very supply-constrained environment. And somehow many people's intuition is the opposite of that."

While OpenAI does have some open-source models, its primary business is in closed-source models that run in data centers, so he's really defending the company's business model. OpenAI isn't exactly threatened by some laptops and gaming rigs just yet, but local AI could mean people one day paying less for subscriptions to models like GPT-6 Luna if they have the power to run them at home. Altman's talk suggests that the company would prefer that more RAM go to OpenAI's data centers rather than home computers, which also sparks enthusiast fears of not being able to own your own hardware or software, but rather rent an agent and a computer in the cloud.”
Click to expand...
counter point (only for nvidia GPUs as of now)

Our solution at Sogni is hilariously simple.

We built an SSH guard that tells you (or your AI agent) who’s using the machine and for how long. You reserve a time slot, and when you’re done, the GPUs go back to earning on the network.

The funniest part? AI agents treat the SSH welcome message like it’s the police. 😂

It’s literally just text in a terminal, but they respect it like some higher authority.

The future of personal compute isn’t choosing between your own AI lab and a revenue-generating GPU business.

It’s having both.

Your GPUs work for you when you need them, and earn for you when you don’t.

That’s the beauty of decentralized GPU neoclouds.
Click to expand...

https://x.com/krunkosaurus/status/2115655405338816775?a=PMUtUWB85WuhkWldyXk&s=20
 
Marees said:
counter point (only for nvidia GPUs as of now)
Click to expand...
That not really a counter point, that making the same point, that maximising cloud providing and usage if it work well. (every user of sogni, is using the cloud, every provider of sogni making one), that simply distributed cloud marketplace, not people running locally their AI.

And Datacenter level system would still be way more efficiant too, they do more work with way less hardware, while probably not able to scale for massive model at all.
 
LukeTbk said:
That not really a counter point, that making the same point, that maximising cloud providing and usage if it work well. (every user of sogni, is using the cloud, every provider of sogni making one), that simply distributed cloud marketplace, not people running locally their AI.

And Datacenter level system would still be way more efficiant too, they do more work with way less hardware, while probably not able to scale for massive model at all.
Click to expand...
but I can do personal/private stuff locally when I need it isn’t it.

it would be inefficient/costly to go to the cloud for that. & when I am not using it, I rent it to those who need it
 
erek said:

“Sam Altman Thinks Your PC Is Wasting Precious RAM​

Gilmoure 38 minutes ago		25
"Looks like OpenAI CEO Sam Altman wants to own all the computing resources and rent out terminals to the world as a way of propping up his business plan, now that debt-ridden AI companies are teetering on the edge," writes ancient Slashdot reader Gilmoure. Tom's Hardware reports: In an interview over beers with the YouTube channel Cheeky Pint, Altman said that local inference, and, by extension, people having DRAM in computers at home, is "quite terrible." When discussing local inference on home computers with open-source models, Altman discussed the assumption that people will be able to run local AI on their own computers, the ones they may already own or want to build.

Then he explained why that's a bad idea (at the 29-minute mark): "From a global allocation perspective, actually, it's quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer and being used 5% of the time," Altman said. "You want maximum utilization in the cloud if you're trying to deal with this very supply-constrained environment. And somehow many people's intuition is the opposite of that."

While OpenAI does have some open-source models, its primary business is in closed-source models that run in data centers, so he's really defending the company's business model. OpenAI isn't exactly threatened by some laptops and gaming rigs just yet, but local AI could mean people one day paying less for subscriptions to models like GPT-6 Luna if they have the power to run them at home. Altman's talk suggests that the company would prefer that more RAM go to OpenAI's data centers rather than home computers, which also sparks enthusiast fears of not being able to own your own hardware or software, but rather rent an agent and a computer in the cloud.”
Click to expand...

I'm sure he's fine destroying the world with pollution Data Centers. There is no Ram the way it is I rather see Sam converted into a PC gamer and close up shop with Open Ai.

https://videocardz.com/newz/openai-...m-a-quite-terrible-use-of-limited-dram-supply
 
Last edited:
He's pretty much admitting his business model doesn't work without an enforced captured audience through a monopoly in computing power.

It would be insanely interesting to know how much RAM is just sitting in warehouses waiting for server farms to be built, that can't be built because there isn't enough power in the grid.
 
zehoo said:
He's pretty much admitting his business model doesn't work without an enforced captured audience through a monopoly in computing power.

It would be insanely interesting to know how much RAM is just sitting in warehouses waiting for server farms to be built, that can't be built because there isn't enough power in the grid.
Click to expand...
I wondered the same about all the HW when I read the quote.
 
zehoo said:
He's pretty much admitting his business model doesn't work without an enforced captured audience through a monopoly in computing power.

It would be insanely interesting to know how much RAM is just sitting in warehouses waiting for server farms to be built, that can't be built because there isn't enough power in the grid.
Click to expand...
That's the reason Sam Altman is complaining. Remember that his company took something like 40% of the worlds ram production? It was a move meant to push users into the cloud. I'm guessing that he expected the world to move to the cloud before the AI bubble pops. He's got all the compute and he needs a market for it. It's starting to seem like this whole rampocalypse was orchestrated to push people away from ownership instead of needing it for AI.
sam altman ram at face.gif
 
DukenukemX said:
That's the reason Sam Altman is complaining. Remember that his company took something like 40% of the worlds ram production? It was a move meant to push users into the cloud. I'm guessing that he expected the world to move to the cloud before the AI bubble pops. He's got all the compute and he needs a market for it. It's starting to seem like this whole rampocalypse was orchestrated to push people away from ownership instead of needing it for AI.
View attachment 834207
Click to expand...
"You will own nothing, and be happy."
 
DukenukemX said:
That's the reason Sam Altman is complaining. Remember that his company took something like 40% of the worlds ram production? It was a move meant to push users into the cloud. I'm guessing that he expected the world to move to the cloud before the AI bubble pops. He's got all the compute and he needs a market for it. It's starting to seem like this whole rampocalypse was orchestrated to push people away from ownership instead of needing it for AI.
View attachment 834207
Click to expand...
This is one of those areas where I personally feel a ton of people are naive.

"They won't ban local LLMs!", "Streaming only is never going to take off!", etc.

OpenAI has tried to do both. So, if you can't ban local LLMs, just make it so people can't afford local LLMs. You have to be incredibly delusional to think prices are ever going to fall back to what they were. And the past couple decades should have been proof enough alone to demonstrate that yes, companies can turn products you used to own into permanent rentals, so why wouldn't computers be the same?
 
Comixbooks said:
Steve tells him to f&$@# off. I'm sure it requires a lot to run a business but Ai data centers will all be junk in 10 years needing to replace everything in 10 years.
Click to expand...

"Sam Altman meets with Samsung leadership to secure more RAM for data centers amid crushing shortages — OpenAI CEO travels to South Korea to meet Samsung leadership to discuss future AI data centers​


Samsung and SK hynix control roughly 83% of HBM revenue as AI data centers intensify competition for memory chips"

View attachment 834200

"For now, demand for data centers is facing major competition, especially from PCs, smartphones, and other electronics, for limited memory supplies (and may be nowhere close to peaking). High-bandwidth memory (HBM), a specialized form of DRAM used in AI devices, has become particularly important as companies deploy increasingly powerful computing systems.


Samsung and SK Hynix are among the world's leading memory manufacturers, giving South Korea a central role in supplying the hardware behind the AI expansion. According to Counterpoint Research, both companies controlled about 63% of global DRAM revenues in the second quarter of 2026. Their influence is greater in HBM, where the two South Korean manufacturers accounted for roughly 83% of global revenue. This has prompted other technology executives to approach the country, with AMD CEO Lisa Su visiting earlier this month to discuss securing additional HBM supplies from SAMSUNG, saying, "It is important to plan three to five years ahead with our partners."

Meanwhile, Samsung and SK Hynix have been upgrading Chinese factories to meet the surging demand for AI-related memory and storage products. However, those particular upgrades concern NAND flash production, rather than DRAM — but they show just how manufacturers are adjusting capacity as demand shifts towards data center hardware."

Source: https://www.tomshardware.com/tech-i...-leadership-to-discuss-future-ai-data-centers
 
zehoo said:
With Anthropic now rated as junk before their IPO, surely the house of cards for the AI bubble is nearing an end. And RAM prices can return to reality.
Click to expand...
It probably depends on how much fresh cash Nvidia and other entities throw at the problem. Thankfully, there's only so much money (until the US government gets involved and has more printed/borrowed).
 
  • Like
Reactions: zehoo
like this
Nytegard said:
This is one of those areas where I personally feel a ton of people are naive.

"They won't ban local LLMs!", "Streaming only is never going to take off!", etc.

OpenAI has tried to do both. So, if you can't ban local LLMs, just make it so people can't afford local LLMs. You have to be incredibly delusional to think prices are ever going to fall back to what they were. And the past couple decades should have been proof enough alone to demonstrate that yes, companies can turn products you used to own into permanent rentals, so why wouldn't computers be the same?
Click to expand...
When tech companies say they want to regulate AI, they mean they want to ban LLMs. What I find ironic is that AI isn't good enough to replace professional workers, but it is good enough to empower amateurs with LLMs. But I do believe prices will go back down. Not only that but have a rubber band effect where they will be lower than before. China's AI market just dropped massively in the past week, so we're next. I'm not saying the AI bubble bursted, but it bursted.
erek said:


"For now, demand for data centers is facing major competition, especially from PCs, smartphones, and other electronics, for limited memory supplies (and may be nowhere close to peaking). High-bandwidth memory (HBM), a specialized form of DRAM used in AI devices, has become particularly important as companies deploy increasingly powerful computing systems.
​

Click to expand...
Sam Altman is complaining that the price of DRAM is too much even for him. OpenAI owes $665 billion to over $1.4 trillion just due to long term contracts. Things are not going as planned for Sam Altman.
 
DukenukemX said:
When tech companies say they want to regulate AI, they mean they want to ban LLMs. What I find ironic is that AI isn't good enough to replace professional workers, but it is good enough to empower amateurs with LLMs. But I do believe prices will go back down. Not only that but have a rubber band effect where they will be lower than before. China's AI market just dropped massively in the past week, so we're next. I'm not saying the AI bubble bursted, but it bursted.

Sam Altman is complaining that the price of DRAM is too much even for him. OpenAI owes $665 billion to over $1.4 trillion just due to long term contracts. Things are not going as planned for Sam Altman.
Click to expand...

“Samsung Hosts OpenAI's Sam Altman for AI Seminar, No Supply Deal Announced​

by DrUsmanSaleem Today, 14:41 Discuss (3 Comments)
Two days after arguing that the DRAM in home PCswould be put to better use in data centers, OpenAI CEO Sam Altman sat down with around 50 Samsung presidents on Saturday, October 31. The group went well past the chip business, with leaders from Samsung's finance and biotech arms joining the electronics side at the company's training center in Yongin. Samsung presented it as part of a leadership seminar series for its company-wide AI push, and the talks apparently covered where AI is headed, the markets it could open up, and how it changes the way large companies make decisions. According to Samsung, Altman told the room that adopting AI alone doesn't give anyone an edge, and that the payoff comes from using it to move faster and build better products.

Nothing came out of it in terms of a new supply contract, at least publicly, and Samsung's own write-up stuck to AI strategy. Still, it's hard to look at the visit separately from the memory shortage. Samsung signed a letter of intent with OpenAI back in October 2025 to act as a memory partner for Stargate, with OpenAI's demand projected at up to 900,000 DRAM wafers per month. Separately, on October 31, SK Group chairman Chey Tae-won also reportedly met Altman in Seoul, with SK hynix and SK Telecom executives joining, according to industry sources. Last month, AMD's Lisa Su spent hours with Samsung's chip chief chasing HBM supply, and amid all these meetings, Micron's CEO expects the memory crunch to get worse in 2027 and 2028. It seems like every major buyer is knocking on the same Korean doors.“
 
erek said:

“Samsung Hosts OpenAI's Sam Altman for AI Seminar, No Supply Deal Announced​

by DrUsmanSaleem Today, 14:41 Discuss (3 Comments)
Two days after arguing that the DRAM in home PCswould be put to better use in data centers, OpenAI CEO Sam Altman sat down with around 50 Samsung presidents on Saturday, October 31.
Click to expand...
How many presidents does Samsung need?
Nothing came out of it in terms of a new supply contract, at least publicly, and Samsung's own write-up stuck to AI strategy. Still, it's hard to look at the visit separately from the memory shortage. Samsung signed a letter of intent with OpenAI back in October 2025 to act as a memory partner for Stargate, with OpenAI's demand projected at up to 900,000 DRAM wafers per month. Separately, on October 31, SK Group chairman Chey Tae-won also reportedly met Altman in Seoul, with SK hynix and SK Telecom executives joining, according to industry sources. Last month, AMD's Lisa Su spent hours with Samsung's chip chief chasing HBM supply, and amid all these meetings, Micron's CEO expects the memory crunch to get worse in 2027 and 2028. It seems like every major buyer is knocking on the same Korean doors.“
Click to expand...
Gotta remember that Sam Altman basically tanked the DRAM market but he likely bought a lot of DRAM manufacturing on the cheap. Now, because of him that same DRAM is no longer cheap and he's upset. Samsung like Hynix and Micron are not about to drastically increase production because this AI market could collapse any moment. Sucks to be Sam Altman.
erek said:
“Elon Musk mocks OpenAI CEO Sam Altman's claim that PCs having RAM is terrible — 'The people deserve to have their RAM too!'“

https://www.tomshardware.com/tech-i...ible-the-people-deserve-to-have-their-ram-too
Click to expand...
Gotta wonder if Grok doesn't use a lot of ram or Elon Musk is just fishing for brownie points?
 
mullet said:
I see alot of people taking old pc's/laptops and loading linux on them for friends and family. High PC/laptop prices along with win 11 horrible adoption rate. The Penguins revenge is upon us.
Click to expand...
I'm in that camp! Repurposed my old 7700k system that originally was destined for FB marketplace because it wasn't "not Win11 compatible" into a garage PC (mainly just browsers/music duties, nothing wild) and threw CachyOS on it.

This is my first true run with Linux tbh, where I'm forcing myself to not revert as soon as I hit snags, and to dive in and learn about a new OS. As a long-time Windows admin (professionally), it's been interesting... I've certainly had to change my thought process and not assume things behave the same way; some of the things on Windows that "just work" didn't in Linux (audio out via optical was a PITA...), and I'm sure some things are easier (I just haven't found them yet :)), but for basic usage; once I got a few of the initial snags sorted, it's been 100% smooth sailing over the past 3-4 months.

kamikazi said:
Maybe that will be the silver lining. A larger linux install base may get the attention of software providers. Of course, it would probably have to increase by a factor of 10 to be noted. I recently switched from Linux Mint on my laptop to CachyOS. I am also about to finally ditch my Windows 10 install and go to a dual-boot Win 11/Cachy OS setup.
Click to expand...
The success with the garage PC above has got me actually considering Linux as my primary DD... not sure I'm QUITE there yet, but honestly its not something I thought I'd ever be typing out lol. Perhaps the dual boot method is the way to go out of the gate like you mentioned. Part of me doesn't want the fallback option though, it'd be too tempting to use it!
 
AlphaQup said:
I'm in that camp! Repurposed my old 7700k system that originally was destined for FB marketplace because it wasn't "not Win11 compatible" into a garage PC (mainly just browsers/music duties, nothing wild) and threw CachyOS on it.

This is my first true run with Linux tbh, where I'm forcing myself to not revert as soon as I hit snags, and to dive in and learn about a new OS. As a long-time Windows admin (professionally), it's been interesting... I've certainly had to change my thought process and not assume things behave the same way; some of the things on Windows that "just work" didn't in Linux (audio out via optical was a PITA...), and I'm sure some things are easier (I just haven't found them yet :)), but for basic usage; once I got a few of the initial snags sorted, it's been 100% smooth sailing over the past 3-4 months.


The success with the garage PC above has got me actually considering Linux as my primary DD... not sure I'm QUITE there yet, but honestly its not something I thought I'd ever be typing out lol. Perhaps the dual boot method is the way to go out of the gate like you mentioned. Part of me doesn't want the fallback option though, it'd be too tempting to use it!
Click to expand...

Just gonna mention that Cachy (Arch) is a very different animal than Ubuntu (and its derivatives), especially true for newbies.
Have fun though!
 
DukenukemX said:
How many presidents does Samsung need?
Click to expand...
Given the size of that company? Probably at least 50? Mind you, the article might be doing a bit of fluff. That could include a ton of divisional vice presidents and they just lumped them all in one bucket.

DukenukemX said:
Gotta wonder if Grok doesn't use a lot of ram or Elon Musk is just fishing for brownie points?
Click to expand...

He's constantly trying to present himself as the tech industry's best buddy and all around "normal computer geek guy" when he hasn't been that in a long time. So, he'll help drive the price up while trying to present himself as a sympathetic actor.
 
erek said:

“Sam Altman Thinks Your PC Is Wasting Precious RAM​

Gilmoure 38 minutes ago		25
"Looks like OpenAI CEO Sam Altman wants to own all the computing resources and rent out terminals to the world as a way of propping up his business plan, now that debt-ridden AI companies are teetering on the edge," writes ancient Slashdot reader Gilmoure. Tom's Hardware reports: In an interview over beers with the YouTube channel Cheeky Pint, Altman said that local inference, and, by extension, people having DRAM in computers at home, is "quite terrible." When discussing local inference on home computers with open-source models, Altman discussed the assumption that people will be able to run local AI on their own computers, the ones they may already own or want to build.

Then he explained why that's a bad idea (at the 29-minute mark): "From a global allocation perspective, actually, it's quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer and being used 5% of the time," Altman said. "You want maximum utilization in the cloud if you're trying to deal with this very supply-constrained environment. And somehow many people's intuition is the opposite of that."

While OpenAI does have some open-source models, its primary business is in closed-source models that run in data centers, so he's really defending the company's business model. OpenAI isn't exactly threatened by some laptops and gaming rigs just yet, but local AI could mean people one day paying less for subscriptions to models like GPT-6 Luna if they have the power to run them at home. Altman's talk suggests that the company would prefer that more RAM go to OpenAI's data centers rather than home computers, which also sparks enthusiast fears of not being able to own your own hardware or software, but rather rent an agent and a computer in the cloud.”
Click to expand...
Yes his business is teetering on the edge of failure because it isn't producing a product which people actually need or want. I'm certainly not going to pay a monthly fee to do normal, everyday stuff on the cloud. That's ludicrous.
 
erek said:

“Sam Altman Thinks Your PC Is Wasting Precious RAM​

Gilmoure 38 minutes ago		25
"Looks like OpenAI CEO Sam Altman wants to own all the computing resources and rent out terminals to the world as a way of propping up his business plan, now that debt-ridden AI companies are teetering on the edge," writes ancient Slashdot reader Gilmoure. Tom's Hardware reports: In an interview over beers with the YouTube channel Cheeky Pint, Altman said that local inference, and, by extension, people having DRAM in computers at home, is "quite terrible." When discussing local inference on home computers with open-source models, Altman discussed the assumption that people will be able to run local AI on their own computers, the ones they may already own or want to build.

Then he explained why that's a bad idea (at the 29-minute mark): "From a global allocation perspective, actually, it's quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer and being used 5% of the time," Altman said. "You want maximum utilization in the cloud if you're trying to deal with this very supply-constrained environment. And somehow many people's intuition is the opposite of that."

While OpenAI does have some open-source models, its primary business is in closed-source models that run in data centers, so he's really defending the company's business model. OpenAI isn't exactly threatened by some laptops and gaming rigs just yet, but local AI could mean people one day paying less for subscriptions to models like GPT-6 Luna if they have the power to run them at home. Altman's talk suggests that the company would prefer that more RAM go to OpenAI's data centers rather than home computers, which also sparks enthusiast fears of not being able to own your own hardware or software, but rather rent an agent and a computer in the cloud.”
Click to expand...
Altman said. "You want maximum utilization in the cloud if you're trying to deal with this very supply-constrained environment. And somehow many people's intuition is the opposite of that."

He's technically not wrong in that statement. The big point is that his goals are different than mine. He wants to scale a business that uses massive amounts or RAM. Any business wants to be efficient in its major costs.

My goal is not international company efficiency scaling. My goal is personal data privacy and ownership. Somehow he fails to recognize that his goal is not my goal.
 
TordanGow said:
Altman said. "You want maximum utilization in the cloud if you're trying to deal with this very supply-constrained environment. And somehow many people's intuition is the opposite of that."

He's technically not wrong in that statement. The big point is that his goals are different than mine. He wants to scale a business that uses massive amounts or RAM. Any business wants to be efficient in its major costs.

My goal is not international company efficiency scaling. My goal is personal data privacy and ownership. Somehow he fails to recognize that his goal is not my goal.
Click to expand...
I guess this comes down to the definition of technically.

In terms of one person or company owning all the RAM and centralizing it to data centers, yes, that would be more efficient. But in terms of what he said, I don't feel it's technically correct as it assumes most people want one goliath centralized AI and people think that decentralizing it is the best. Because, like you said, many peoples' goals are different. I don't trust one entity, be it public or private, to have complete control.

I'd be willing to bet that most people who are against the large data center monopolization are the same. They understand efficiency, but they have different goals than him, and his statement conflates the two.
 
You must log in or register to reply here.
Back
Top