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

Astral Abyss said:
Ah geez, you're right. Thank you.
Click to expand...

"CXMT Trails By Just 2 Years On DRAM And 3 Years On HBM, According To Korea Semiconductor Industry Association​

CXMT is also slated to receive some domestically produced DUV lithography units from Shanghai Aishegna and Yuliangsheng in the ongoing year. These DUV makers, however, continue to face a few bottlenecks, especially related to projection lenses and light sources.

Do note that CXMT is aggressively expanding its capacity right now, to the tune of 300,000 wafers per month - via three 300mm DRAM fabs with a capacity of ~100,000 wafers per month each - by the end of this year. It will also have an HBM-focused capacity of around 50,000 wafers per month by the end of 2026.

What's more, the memory maker is currently building two new fabs in Shanghai and Hefei, which would increase its production capacity to 600,000 wafers per month. In fact, CXMT is now on course to overtake Micron in volume production by 2030."

https://wccftech.com/chinas-ymtc-is...-to-korea-semiconductor-industry-association/
 
Marees said:
this is storage right? not memory ??

previously CXMT has mentioned wanting to take over the storage market
Click to expand...
1789915976154.png
 

“Acer Expects PC Prices to Rise by Up to 20% in Q4, Peak Around Mid-2027​

by DrUsmanSaleem Today, 16:03 Discuss (0 Comments)
Acer Chairman and CEO Jason Chen says PC prices could rise by another 5% to 20% in the fourth quarter of 2026, according to Taiwan's Economic Daily News. Chen reportedly said that DDR4 and DDR5 memoryare no longer in short supply and that CPU supply is mostly fine, with only some low-end and mid-range chips still tight. Prices haven't followed this change, though. Memory and SSD costs are still high, so systems with more RAM or storage will probably see bigger increases. Chen expects the average selling price of PCs to peak around mid-2027, with a possible drop in the second half of that year.

Even if component costs fall earlier, it takes several months for cheaper parts to reach finished PCs. Chen also reportedly said suppliers keep pushing through price increases, some of which he considers hard to justify. He also pointed to expanding Chinese memory production, and by the sound of it, he doesn't think the shortage can run until 2030, as many in the industry have claimed. That capacity is still ramping up, though. Chinese giant CXMT is reportedly preparing to enter the NAND flash marketwith an R&D line, and YMTC is expanding with new fabs.“
 
erek said:

“Acer Expects PC Prices to Rise by Up to 20% in Q4, Peak Around Mid-2027​

by DrUsmanSaleem Today, 16:03 Discuss (0 Comments)
Acer Chairman and CEO Jason Chen says PC prices could rise by another 5% to 20% in the fourth quarter of 2026, according to Taiwan's Economic Daily News. Chen reportedly said that DDR4 and DDR5 memoryare no longer in short supply and that CPU supply is mostly fine, with only some low-end and mid-range chips still tight. Prices haven't followed this change, though. Memory and SSD costs are still high, so systems with more RAM or storage will probably see bigger increases. Chen expects the average selling price of PCs to peak around mid-2027, with a possible drop in the second half of that year.

Even if component costs fall earlier, it takes several months for cheaper parts to reach finished PCs. Chen also reportedly said suppliers keep pushing through price increases, some of which he considers hard to justify. He also pointed to expanding Chinese memory production, and by the sound of it, he doesn't think the shortage can run until 2030, as many in the industry have claimed. That capacity is still ramping up, though. Chinese giant CXMT is reportedly preparing to enter the NAND flash marketwith an R&D line, and YMTC is expanding with new fabs.“
Click to expand...
memory prices should stabilize sometime in 2028
& then open ai could go bankrupt in 2029
so memory prices to crash by 2030
 
“Assuming 100% yield...One DRAM 1b wafer is $46000One DRAM 1b wafer is about 3.5 TiB. So about $12 per GiB or $1.50 per Gib.So 256 GiB is ~$3000.That's just contract price, not module cost + upsale.Spot pricing for DDR5 is ~$30/GB at peaki.e one wafer packaged is $115000”

1789987320260.png


https://x.com/IanCutress/status/2101851979144159246
 
It won't stop until a year or two after the AI bubble pops. If it doesn't, every new batch of graphics cards, memory, and storage will be first snatched up by the AI datacenters. Successful IPO's will just make it worse.

I live in the middle of nowhere, and listened in on my wife's corporate call since it was on speaker. They MUST start incorporating AI. That is the mandate. AI is here to stay folks, even if the bubble bursts.
 
Marees said:
prices should start decreasing after Open AI / Anthropic IPOs
Click to expand...
If openAI/anthropic get an giant influx of money, I am not sure why it would decrease dram price...

Marees said:
the IPOs are a year away
Click to expand...
https://www.wsj.com/tech/ai/anthropic-shifts-planned-ipo-to-november-8874dffc

Talks/rumours are november 2026 for Anthropic, it was mid-october not so long ago so who knows.... Q3 2026 look like the first above 1 billion figure in operating profit for them (and true GAAP one), if they launch a strong new model could be a good time to do it, 2 positive quarters in a row...
 
LukeTbk said:
Q3 2026 look like the first above 1 billion figure in operating profit for them
Click to expand...
"profit"

inference makes money

but not training.

training is on top of stolen content
but despite stealing that content, training still makes loss

wait & watch what happens after the IPO
 
Marees said:
"profit"

inference makes money

but not training.
Click to expand...
Pre-ipo is always a bit fuzzy of course, but that 1 billion is after training cost, inference has been profitable for a long time (Anthropic api is around 80%+ gross margin), what new is so much money that it pay for the current training of the next one, interest of the debt of training the previous one and still leave positive cash flow.

Marees said:
training is on top of stolen content
Click to expand...
Less and less, Anthropic has a giant amount of content from user that accept to pay less in exchange of giving content to training (as the userbase grew that became more and more the main source of novel data), synthetic data, giant licensing deal for data and more and more those system are trained on objective/goals throught reinforcement learning, not raw data.
 
erek said:
i think they already missed the peak opportunity to IPO

shoulda been much earlier

they could’ve had some of the biggest blockbusters of all time prior to the heavy public scrutiny
Click to expand...
They got a little too greedy, thinking the AI hype would get bigger. It did... just not the way they thought it would.
 
erek said:
is it possible the IPOs will get bigger and bigger no matter the delay?

because everyone is greedy?
Click to expand...
Well, SpaceX was first, and they're still down ~30.5% since their peak of $201.80 back in mid-June. I don't see OpenAI or Anthropic reaching their goals for valuation unless SpaceX gets back up near their peak or dropping considerably. Coming in 2nd while SpaceX is back up to ~$200, or down below $90 is important, because both SpaceX and whoever's next is going to cause the value to rebound, and with competition, the one who comes in 3rd place isn't going to get valued as high as either of them, unless one of them drops considerably. If anyone wants to invest in the AI bubble, investing in SpaceX while it dips below $85 might be a good short-term bet.
 
Yeah, it’s crazy how quickly the memory prices have jumped, hopefully supply catches up sooner rather than later.
 
El_Capitan said:
Well, SpaceX was first, and they're still down ~30.5% since their peak of $201.80 back in mid-June. I don't see OpenAI or Anthropic reaching their goals for valuation unless SpaceX gets back up near their peak or dropping considerably. Coming in 2nd while SpaceX is back up to ~$200, or down below $90 is important, because both SpaceX and whoever's next is going to cause the value to rebound, and with competition, the one who comes in 3rd place isn't going to get valued as high as either of them, unless one of them drops considerably. If anyone wants to invest in the AI bubble, investing in SpaceX while it dips below $85 might be a good short-term bet.
Click to expand...

It's still +14% above its IPO price of $135
 
kram182 said:
It's still +14% above its IPO price of $135
Click to expand...
It's going to fluctuate. Lowest was $108.27 in 08/04, which was -35.36%. It's going to be pretty volatile for a few years. If Elon was smart, he'd split the stock if it got up to $200 again before the next AI IPO launches, so it forces the next AI IPO to issue more stock at launch, or not get as much market cap. An investor having to decide on whether to buy SpaceX at $50, or Anthropic at $150, I think more people would choose SpaceX at $50, especially if they recently split.
 
El_Capitan said:
It's going to fluctuate. Lowest was $108.27 in 08/04, which was -35.36%. It's going to be pretty volatile for a few years. If Elon was smart, he'd split the stock if it got up to $200 again before the next AI IPO launches, so it forces the next AI IPO to issue more stock at launch, or not get as much market cap. An investor having to decide on whether to buy SpaceX at $50, or Anthropic at $150, I think more people would choose SpaceX at $50, especially if they recently split.
Click to expand...

None of that is happening nor would be the right or smart thing to do 1) a company wouldn't split this shortly after IPO 2) a company wouldn't dilute/issue more shares just to make/at IPO (especially in response to another company's split) - that would all actually scare away investors

The only thing on the horizon besides IPOs for Anthropic/OpenAI - is Tesla and SpaceX merging (SpaceX buying out Tesla)
 
erek said:
is it possible the IPOs will get bigger and bigger no matter the delay?

because everyone is greedy?
Click to expand...
The a bit the opposite or like everything it is always greed vs greed (classic is regular customer being the most greedy people pretty much all the time).

The bigger the number for the IPO the less greedy the new investor were (they want the number to be has small as possible to own as much as they can of the company with their money they put in).

El_Capitan said:
If Elon was smart, he'd split the stock if it got up to $200 again before the next AI IPO launches, so it forces the next AI IPO to issue more stock at launch, or not get as much market cap. An investor having to decide on whether to buy SpaceX at $50, or Anthropic at $150, I think more people would choose SpaceX at $50, especially if they recently split.
Click to expand...
Not sure (specially now) why splitting stock would change anything, the price of a stock is nearly irrelevant in the digital world, specially when it is that low, no one look at the price of the stock but at the company evaluation.

kram182 said:
2) a company wouldn't dilute/issue more shares just to make/at IPO (especially in response to another company's split) - that would all actually scare away investors
Click to expand...
splitting and dilution are 2 fully different thing, splitting change nothing, it just double everyone (if it 2:1) number of shares, they not own a smaller portion of the company after the split.
 
LukeTbk said:
splitting and dilution are 2 fully different thing, splitting change nothing, it just double everyone (if it 2:1) number of shares, they not own a smaller portion of the company after the split.
Click to expand...

in the sense/scenario he was talking about they are the exact same thing - increasing number of shares, not just to give to existing shareholders, but to bring in new buyers/investors from/at IPO
 
kram182 said:
None of that is happening nor would be the right or smart thing to do 1) a company wouldn't split this shortly after IPO 2) a company wouldn't dilute/issue more shares just to make/at IPO (especially in response to another company's split) - that would all actually scare away investors

The only thing on the horizon besides IPOs for Anthropic/OpenAI - is Tesla and SpaceX merging (SpaceX buying out Tesla)
Click to expand...
A company splits in order to get more investors. It doesn't matter how long after IPO, especially in this crazy market. What AI companies need/want is capital. SpaceX held back to only issue as little as they could since they were first to market. They make more money by issuing as little as possible, because if they need/want more capital, they can issue more or split. I think there was a clause to limit how much employees at SpaceX could sell at a given time.
 

“DRAM Module Revenue Surged 59% YoY in 2025 as Crowding-Out Effect Constrained Chip Supply and Drove Prices Higher​

PRESS RELEASE by TheLostSwede Today, 04:22 Discuss (0 Comments)
According to TrendForce statistics, the global DRAM module market generated $21.2 billion in revenue in 2025, up a strong 59% YoY. Rising prices and concentrated shipments from DRAM module makers in the second half of the year helped drive growth. This marked a significant acceleration from the 7% growth recorded in 2024.

DRAM market conditions changed dramatically between the first and second halves of 2025. In early 2025, some module makers increased their inventories of DDR4 chips in Q2, spurred by tariffs and the end-of-life (EOL) plans of DRAM suppliers for DDR4. They also supplied PC OEMs and other consumer electronics customers that had pulled forward procurement in response to tariffs.“
 

Market Structure Transformation​

  1. Memory industry is abandoning its traditional 3-5 year boom-bust cycle in favor of long-term agreements (LTAs) where suppliers like Micron, Samsung, Kioxia, and Western Digital pre-secure 70% of capacity 5 years out with cloud service providers, fundamentally removing cyclical pricing that previously benefited consumers during oversupply periods.
  2. Cloud service providers' memory spending will dominate the market with DRAM and NAND accounting for 47% of total capex in 2026 ($433 billion) and exploding to 68% in 2027 ($940 billion) across the nine largest CSPs, creating structural supply constraints for consumer markets.

Price Impact and Supply Reallocation​

  1. Consumer memory prices experienced unprecedented increases from September 2025 to September 2026: 32GB DDR5 kits up 363% to $567, 64GB DDR5 kits up 483% to $1,300-$1,400, 2TB NVMe SSDs up 137-139% ($200 increase), and 2TB SATA SSDs up 183% to $318, with RAM seeing steeper percentage increases than SSDs.
  2. Retail SSD market has almost disappeared as DRAM supply is reallocated to server applications (CPU, HBM for graphics cards/accelerators) and NAND is diverted to servers and data centers, with demand growing faster than supply despite manufacturing capacity increases.

Geopolitical Dynamics​

  1. Chinese manufacturers CXMT and YMTC achieved rapid market penetration with CXMT holding 10% global DRAM market share (up 6 percentage points year-over-year from 4% in Q2 2025) and YMTC breaking into top three global NAND manufacturers by shipments, surpassing Kioxia, Micron, and SanDisk.
  2. Chinese suppliers face government guidance to support local industries and cannot allocate all supply to data centers like foreign competitors, while US politicians including Chuck Schumer pressure companies like Apple to avoid Chinese memory despite Micron's US production only projected to reach 10% by 2035 (up from 2% in late 2023), with most going to HBM, not consumer products.

Strategic Market Consequences​

  1. Amazon's AWS CEO stated in Q1 2026 earnings that memory price and supply changes are pushing companies with on-premises infrastructure into cloud services, with Apple characterizing the situation as a 100-year flood on memory pricing with exponential increases and Xbox expecting another doubling of console storage and memory prices by fall 2027.
  2. TrendForce predicts NAND supply will exceed demand by 2027 while DRAM shortages worsen, creating a bifurcated market where one memory type may see relief while the other remains constrained, with long-term agreements preventing traditional market corrections that would normally bring prices down during oversupply.
 
Epyon said:
TrendForce predicts NAND supply will exceed demand by 2027 while DRAM shortages worsen, creating a bifurcated market where one memory type may see relief while the other remains constrained, with long-term agreements preventing traditional market corrections that would normally bring prices down during oversupply.
Click to expand...
price of storage & slower memory to reduce by next year end

but faster/advanced memory prices could stay high thru 2028
 
Do these higher RAM prices affect older DDR4 modules etc for servers? My work has decommissioned 7 of their older servers that had a bunch of memory in them. Just wondering if its worth selling used or not?
 
Scoobydo2 said:
Do these higher RAM prices affect older DDR4 modules etc for servers? My work has decommissioned 7 of their older servers that had a bunch of memory in them. Just wondering if its worth selling used or not?
Click to expand...
yes new or used DDR4/ECC DRR4 is also higher than in the past.

https://www.ebay.com/sch/i.html?_nkw=ECC+DDR4&_sacat=0&_from=R40&LH_Sold=1&rt=nc&LH_Complete=1

nice matching klts do seem to fetch big price from time to time (there is a 4x32GB 3200 kit that sold for $700 for example)
 
As an eBay Associate, HardForum may earn from qualifying purchases.
You must log in or register to reply here.
Back
Top