• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Why have RAM prices skyrocketed?

I was looking at NewEgg and as much as I hate this AI Slop Bull Shit making our hobby unobtainium I found this a little funny!

V-COLOR-Manta-XSky-DDR5-32GB--2x16GB--6400MHz-CL32-1-4V-SK-Hynix-IC-RGB-Gaming-D_2026-03-04.png


V-COLOR-Manta-XSky-DDR5-32GB--2x16GB--6400MHz-CL32-1-4V-SK-Hynix-IC-RGB-Gaming-D_2026-03-04 (1).png
 
Chinese speculators are crashing out after RAM prices started to fall.


 
Wait does this mean I can get the new server board I wanted? The RAM price was the only thing holding me back. I recently replaced some sticks of dying DDR3 in my 2013-era server.
 
I don't think people understand what "price fixing" actually is.
That said, if there was COLLUSION to structure future pricing levels, across multiple companies, that is price fixing.

And as he states when mentioning CXMT's entrance into the market, why would they sell it "cheap?" The inventory amounts that are needed to truly bring prices down are going to be huge, unless the purchase orders going into the future, do not get executed, or are reneged on.
 
Samsung chip division's single-year profits beat its past 40 years of profits, combined, due to increased memory and storage prices — Samsung passes Nvidia to become most profitable company in the world, notches 19x quarterly increase in profit
https://www.tomshardware.com/tech-i...o-out-earn-its-entire-40-year-history-in-2026

View attachment 814143
insane...
That quote is disingenuous as fuck - Samsung is a “ram company” as much as apple is a “headphones” company

A better example would be Bosch having a larger market cap than the Volkswagen group which … is the truth.

Still, the AI circlejerk is gonna make all previous financial bubbles look like child’s play.
 
https://wccftech.com/sk-hynix-ceo-w...ear-ever-shortages-set-to-outlast-the-decade/

for those who thinks price can't get any higher, make sure you are not driving when you click the above
I'm hoping we'll get some relief from some of the hyperscalers falling behind, getting beaten by the competition, and getting into financial trouble. I have some firms in mind I'd like to see go down due to other things they do, but this isn't soapbox so I'll keep quiet about that. I'm thinking there will be a weeding out of weaker players. On a good day a crash out or two will scare the rest into slowing down a bit so fab construction and power generation can catch up.
 
I'm hoping we'll get some relief from some of the hyperscalers falling behind, getting beaten by the competition, and getting into financial trouble. I have some firms in mind I'd like to see go down due to other things they do, but this isn't soapbox so I'll keep quiet about that. I'm thinking there will be a weeding out of weaker players. On a good day a crash out or two will scare the rest into slowing down a bit so fab construction and power generation can catch up.
It becomes a question if they can make stuff profitable enough. The consumer market for ML and AI is probably almost non-existent in terms of potential for monetization, but the business market can see quite a large ROI so it is all about getting businesses hooked in a way that they see enough benefit from it ,that they are willing to pay the prices needed for the large players to be profitable. They will most likely have to start reducing their investments into hardware to become profitable as it is likely not sustainable for them to buy everything they can get their hands on over time.

Having more and proper competition is what is missing right now. The CPU market is highly competitive now that AMD is back in the game, which is why prices have almost stood still for a decade. The GPU market is not competitive as Nvidia is happy to charge an arm and a leg for a GPU and AMD decided not to compete on price this cycle. The RAM and NAND market is similar in that they have artificially low supply to hype up demand and send prices through the roof and no one wants to compete on price.

I managed to get everything I needed and most of what I wanted, before the prices skyrocketed so I am good until I want to move my current GPU and CPU to my secondary when Zen 6 and next gen GPUs come out. It's not something I will do unless I feel the prices are decent enough. GPU probably happens either way, unless the next gen is even worse value than the current gen, but not going to massively overpay for VRAM.
 
The consumer market for ML and AI is probably almost non-existent in terms of potential for monetization,
Depends what we mean, Ads-content recommendation engine are still by far the biggest money maker for ML and AI, like it has been for a decade+. Now they added generative ads creations to it, it is about 200 billions a year market.

Self driving is an other one consumer seem ready to pay for and once personnal assistant type get good enough, they will accept ads on them (or pay like for youtube for an better-ad free experience), they will be moneytised, right now it is a bit soft (gemini free tier is seen as an ads for the pay tier-api and google product, but active ads will come), llm user intent are really clear, high, the targetting quite something, ads are sold are really high rate.

Personnal robots that do laundry-grass and so on, will be an other one consumer will be ready to pay for, specially the giant already starting to need help elderly boomer generation. Domain expert they probably be ready in some case, a great tax maker, sport coach, tutors for grand kids, everyone is getting quite intense (or think they are) with their hobby, will pro-summer level getting quite popular and possibly there will be a market for some advanced scam detector.

Consumer market robots could be the biggest market of all time.

Having more and proper competition is what is missing right now.
Depends where in the stack imo, would we have 4 fabless gpu maker it would maybe not change much, Intel fab capacity-yield near the frontier, Micron/CXMT new fabs getting up fast, gaz turbine and their connection being built and so on is more the type of stuff that would help than fabless designer. You always compete in price (that why amd gpu are cheaper per mm of die than Nvidia right now and intel way cheaper), it is just the price that you can set and still sell everything you made at a good pace is really high in an high demand for the supply world, add a surviving 3dfx in the mix..... does not change much the math of how costly putting 16 GB of good GDDR on a gpu is.

The RAM and NAND market is similar in that they have artificially low supply to hype up demand and send prices through the roof and no one wants to compete on price.
artificially low supply ? the supply never been higher on either and it is not that close, there is a giant competitive race to boost supply with probably more than100 billions in just 2026 invested at boosting it (big 3 in the dram alone are in the 60-70 billions capex this year range).

world supply progression look like this:

YearDRAM Total Bits (Exabytes)NAND Total Bits (Exabytes)
2016​
~18 EB~120 EB
2018​
~28 EB~230 EB
2020​
~40 EB~410 EB
2022​
~52 EB~670 EB
2023​
~55 EB~810 EB
2024​
~64 EB~960 EB
2025​
~75 EB~1,140 EB
2026 (Est.)~87 EB~1,330 EB
2027 (Proj.)~102 EB~1,580 EB

Once you consider HBM low yield, the jump in bits become even more impressive, you need a giant amount on memory capacity to ramp it up that fast despite HBM dragging you down like that (probably why NAND grow is a bit higher, no HBM effect there).

This was possible in 2025 to jump so much from 2022 baseline because of the memory crisis, they went overcapacity in those year with a terrible crash, so they could at first simply just use the infracstructure more, by 2026-2027 is when actually reacting to the 2023 revival will start. in early 2023 supply was "artificially" low because of low demand, that has not been the case, everything running to the max for a while now.
 
Depends what we mean, Ads-content recommendation engine are still by far the biggest money maker for ML and AI, like it has been for a decade+. Now they added generative ads creations to it, it is about 200 billions a year market.

Self driving is an other one consumer seem ready to pay for and once personnal assistant type get good enough, they will accept ads on them (or pay like for youtube for an better-ad free experience), they will be moneytised, right now it is a bit soft (gemini free tier is seen as an ads for the pay tier-api and google product, but active ads will come), llm user intent are really clear, high, the targetting quite something, ads are sold are really high rate.

Personnal robots that do laundry-grass and so on, will be an other one consumer will be ready to pay for, specially the giant already starting to need help elderly boomer generation. Domain expert they probably be ready in some case, a great tax maker, sport coach, tutors for grand kids, everyone is getting quite intense (or think they are) with their hobby, will pro-summer level getting quite popular and possibly there will be a market for some advanced scam detector.

Consumer market robots could be the biggest market of all time.
Most consumer stuff like robots, self driving are one time fees or a small subscriptions that pale in comparison to what 1 business user can use in a year. AI assistants for consumer use costs a fraction of what it costs for businesses and a lot of companies are actually losing money on the consumer side. Businesses are willing to pay as long as they have a good forecast for ROI. The general consumers willingness to use thousands of dollars on AI stuff per year is fairly limited and a lot of consumers can't afford it. You generally need 10-100 paying consumer users to reach the same cost as a lot of AI/ML products are to 1 business user and in quite a few cases it is more from the consumer side. The companies that are heavy into AI/ML aren't primarily targeting single consumers, they are targeting businesses that may use those services to produce to provide other products to consumers.

One example is using Agents to do the trivial boiler plate stuff, adding test coverage etc. can free up developers to focus on the important stuff, while only spending a fraction of the time reviewing the trivial code, compared to writing it. It can also easily be a good sparring partner, due to it's ability to read lots of code fast. AI makes quite a few mistakes, so it needs to be supervised carefully with lots of restraints, but it can help productivity and help you with solving complex issues by giving ideas etc. It can easily be a profitable business case to let developers use $10 000 or $20 000 a year on AI/ML code assistance due to massively increased productivity. The same amount of AI credits is a few hundred per year for a consumer, if it is even that expensive. If $10 000 a year per developer increases productivity by 20%, then it is actually huge savings by letting businesses run with less manpower.

Depends where in the stack imo, would we have 4 fabless gpu maker it would maybe not change much, Intel fab capacity-yield near the frontier, Micron/CXMT new fabs getting up fast, gaz turbine and their connection being built and so on is more the type of stuff that would help than fabless designer. You always compete in price (that why amd gpu are cheaper per mm of die than Nvidia right now and intel way cheaper), it is just the price that you can set and still sell everything you made at a good pace is really high in an high demand for the supply world, add a surviving 3dfx in the mix..... does not change much the math of how costly putting 16 GB of good GDDR on a gpu is.
AMD hasn't competed at the high end since the 6xxx series and even then it was the less feature rich product. The current gen of AMD are upper mid-range and lower products, with a bit less features than their Nvidia counterparts. At MSRP I would put the 5070 ti and 9070 xt at about equal value as the 5070 ti stack is capable of doing things the AMD stack is not capable of doing. AMD is getting there on the software side, but they are still quite a bit behind. Putting 16GB of VRAM on a GPU wasn't all that much when these products launched, especially on the more expensive parts. The price to performance increase has been bad since the 30x0 series from Nvidia and the 6xx0 series from AMD. If Intel actually had good drivers and were capable of producing high-end GPUs then the market would actually come down quite a bit. E.g. Intel was putting 16GB of VRAM on lower end GPUs and had decent performance in some games, but drivers were lacking and they never made products to compete in the upper mid-range or high end.

artificially low supply ? the supply never been higher on either and it is not that close, there is a giant competitive race to boost supply with probably more than100 billions in just 2026 invested at boosting it (big 3 in the dram alone are in the 60-70 billions capex this year range).

world supply progression look like this:

YearDRAM Total Bits (Exabytes)NAND Total Bits (Exabytes)
2016​
~18 EB~120 EB
2018​
~28 EB~230 EB
2020​
~40 EB~410 EB
2022​
~52 EB~670 EB
2023​
~55 EB~810 EB
2024​
~64 EB~960 EB
2025​
~75 EB~1,140 EB
2026 (Est.)~87 EB~1,330 EB
2027 (Proj.)~102 EB~1,580 EB

Once you consider HBM low yield, the jump in bits become even more impressive, you need a giant amount on memory capacity to ramp it up that fast despite HBM dragging you down like that (probably why NAND grow is a bit higher, no HBM effect there).

This was possible in 2025 to jump so much from 2022 baseline because of the memory crisis, they went overcapacity in those year with a terrible crash, so they could at first simply just use the infracstructure more, by 2026-2027 is when actually reacting to the 2023 revival will start. in early 2023 supply was "artificially" low because of low demand, that has not been the case, everything running to the max for a while now.
Maybe look at the consumer side? A lot of production of consumer product was shifted into the business products, causing a lack of RAM for the consumer market and massive price increases.
 
Most consumer stuff like robots, self driving are one time fees or a small subscriptions
Yes, but there will be potentially around 8 billions of them. Apple business is a lot consummers bases and they are quite big, it is an exploding world market.

Will have to see how margin and price shake up, tesla self driving is $100 a month and people seem to be ready to pay significantly more for waymo than human drivers service.. on a cheaper system to make margin could be good there has well.

The general consumers willingness to use thousands of dollars on AI stuff per year is fairly limited and a lot of consumers can't afford it.
The median american household pay for cable tv in its peak was over $1000 a year in today dollar, right now for mobile phone and their data service it is $2000 a year. Every people risk to become in some an added service producer (to themselve first and maybe some ohers) and will be ready to pay for it by how much money they save buying those service from outside company, 10 regular user for a business user is probably a good ballpark. Average married still together baby boomer couple in the US as an average net worth of 2.2 millions, median is around 550k, without includign the ~650k of value social security would tend to be worth, they will want to stay in their house and have money for the first wave of expensive robots, which could become quite cheap relative of the yearly work output/yearly payment on them (around very cheap car) for regular people to buy all around the world after that.

Maybe look at the consumer side? A lot of production of consumer product was shifted into the business products, causing a lack of RAM for the consumer market and massive price increases.
That not artificial at all but caused by a real shift in production from a real demand no ?, consumer revenues exploded (near triple 2023 revenues in 2026) so it is hard to say, but it could be that totals bits is a bit down (or more so they just not kept up as they would have with the growing world demand), apparently with all the new players and high price there is more consummer memory bits produced now than ever before, but as the world poverty continue to plummet down fast and device would tend to ask for more memory... production boost slowed because of all the datacenter shift and did not kept up. Consummer grow is in the low 2-3-4% a year instead of being big.
 
Last edited:
It becomes a question if they can make stuff profitable enough. The consumer market for ML and AI is probably almost non-existent in terms of potential for monetization, but the business market can see quite a large ROI so it is all about getting businesses hooked in a way that they see enough benefit from it ,that they are willing to pay the prices needed for the large players to be profitable. They will most likely have to start reducing their investments into hardware to become profitable as it is likely not sustainable for them to buy everything they can get their hands on over time.

Having more and proper competition is what is missing right now. The CPU market is highly competitive now that AMD is back in the game, which is why prices have almost stood still for a decade. The GPU market is not competitive as Nvidia is happy to charge an arm and a leg for a GPU and AMD decided not to compete on price this cycle. The RAM and NAND market is similar in that they have artificially low supply to hype up demand and send prices through the roof and no one wants to compete on price.

I managed to get everything I needed and most of what I wanted, before the prices skyrocketed so I am good until I want to move my current GPU and CPU to my secondary when Zen 6 and next gen GPUs come out. It's not something I will do unless I feel the prices are decent enough. GPU probably happens either way, unless the next gen is even worse value than the current gen, but not going to massively overpay for VRAM.
this whole skyrocket pricing reminds me of the airline industry. There is not enough players. So when 1 co. comes up w/ some way to jack up the price, another co. follows along, and next thing you know, all the prices went up. Everyone is in the con game.

As of June 30, ^SOX is up 101.139% compares to Dec. 31, 2025 close. I haven't seen sector price up that much since the dot com boom. So if you use ^SOX as a guide, then the only way to at least break even from the RAM that you are about to buy, would be buying your SOXX ETF, and take that money to pay for the increase price on video card, NVMe drive, sys. RAM, etc. I want to buy a RTX 5060, at the time, the price is $600, now it's $800. I originally planned to buy 64GB of RAM for my new PC, now I can only afford 32 GB of RAM. I never seen anything like this.

P.S. The last time I heard of the phrase "Exabyte", is when they went belly up where nooone is using tape backup anymore.
 
Back
Top