Written by Tom of Moore’s Law Is Dead

Special Assistance by KarbinCry & kari-no-sugata

At the beginning of November, I ordered a 32GB DDR5 kit for pairing with a Minisforum BD790i X3D motherboard, and three weeks later those very same sticks of DDR5 are now listed for a staggering $330– a 156% increase in price from less than a month ago! At this rate, it seems likely that by Christmas, that DDR5 kit alone could be worth more than the entire Zen 4 X3D platform I planned to pair it with! How could this happen, and more specifically – how could this happen THIS quickly? Well, buckle up! I am about to tell you the story of Sam Altman’s Dirty DRAM Deal, or: How the AI bubble, panic, and unpreparedness stole Christmas...

But before I dive in, let me make it clear that my RAM kit’s 156% jump in price isn’t a fluke or some extreme example of what's going on right now. Nope, and in fact, I'd like to provide two more examples of how how impossible it is becoming to get ahold of RAM - these were provided by a couple of our sources within the industry:

1. One source that works at a US Retailer, stated that a RAM Manufacturer called them in order to inquire if they might buy RAM from them to stock up for their other customers. This would be like Corsair asking a Best Buy if they had any RAM around.

2. Another source that works at a Prebuilt PC company, was recently given an estimate for when they would receive RAM orders if they placed them now…and they were told December…of 2026!

So what happened?  Well, it all comes down to three perfectly synergistic events:

1. OpenAI executed two unprecedented RAM deals that took everyone by surprise.

2. The secrecy and size of the deals triggered full-scale panic buying from everyone else.

3. The market had almost zero safety stock left due to tariffs, worry about decreasing RAM prices over the summer, and stalled equipment transfers.

Below, we’re going to walk through each of these factors — and then I’m going to warn you about which hardware categories will be hit the hardest, which products are already being cancelled, and what you should buy right now before the shelves turn into a repeat of 2021–2022...because this is doomed to turn into much more than just RAM scarcity...

On October 1st OpenAI signed two simultaneous deals with Samsung and SK Hynix for 40% of the worlds DRAM supply.  Now, did OpenAI’s competition suspect some big RAM deals could be signed in late 2025? Yes. Ok, but did they think it would be deals this huge and with multiple companies? NO!  In fact, if you go back and read reporting on Sam Altman’s now infamous trip to South Korea on October 1st, even just mere hours before the massive deals with Samsung and SK Hynix were simultaneously signed – most reporting simply mentioned vague reports about Sam talking to Samsung, SK Hynix, TSMC, and Foxconn. But the reporting at the time was soft, almost dismissive — “exploring ties,” “seeking cooperation,” “probing for partnerships.” Nobody hinted that OpenAI was about to swallow up to 40% of global DRAM output – even on morning before it happened! Nobody saw this coming - this is clear in the lack of reporting about the deals before they were announced, and every MLID Source who works in DRAM manufacturing and distribution insist this took everyone in the industry by surprise.

To be clear - the shock wasn’t that OpenAI made a big deal, no, it was that they made two massive deals this big, at the same time, with Samsung and SK Hynix simultaneously! In fact, according to our sources - both companies had no idea how big each other's deal was, nor how close to simultaneous they were. And this secrecy mattered. It mattered a lot.

Had Samsung known SK Hynix was about to commit a similar chunk of supply — or vice-versa — the pricing and terms would have likely been different. It’s entirely conceivable they wouldn’t have both agreed to supply such a substantial part of global supply if they had known more...but at the end of the day - OpenAI did succeed in keeping the circles tight, locking down the NDAs, and leveraging the fact that these companies assumed the other wasn’t giving up this much wafer volume simultaneously…in order to make a surgical strike on the global RAM supply chain…and it's worked so far...

Imagine you're running a hyper scaler, or maybe you’re a major OEM, or perhaps pretend that you are simply one of OpenAI’s chief competitors: On October 1st of 2025, you would have woken up to the news that OpenAI had just cornered the memory market more aggressively than any company in the last decade, and you hadn't heard even a murmur that this was coming beforehand! Well, you would probably make some follow-up calls to colleagues in the industry, and then also quickly hear rumors that it wasn't just you - also the two largest suppliers didn’t even see each other’s simultaneous cooperation with OpenAI coming either! You wouldn't go: “Well, that’s an interesting coincidence”, no, you would say: “WHAT ELSE IS GOING ON THAT WE DON’T KNOW ABOUT?”

Again – it’s not the size of the deals that's solely the issue here, no, it’s also the unexpectedness and brazenness of them. On October 1st silicon valley executives and procurement managers panicked over concerns like these:

• What other deals don’t we know about? Is this just the first of many?

• None of our DRAM suppliers warned us ahead of time! We have to assume they also won't in the future, and that it’s possible all of global DRAM could be bought up without us getting a single warning!

• We know OpenAI’s competitors are already panic-buying!  If we don’t move now, we might be locked out of the market until 2028!

OpenAI’s competitors, OEMs, and cloud providers scrambled to secure whatever inventory remained out of self-defense, and self-defense in a world that was entirely defenseless due to the accelerant I’ll now explain in Part III...

Normally, the DRAM market has buffers: warehouses of emergency stock, excess wafer starts, older DRAM manufacturing machinery being sold off to budget brands while the big brands upgrade their production lines…but not in 2025, in 2025 those would-be buffers were depleted for three separate reasons:

1. Tariff Chaos. Companies had deliberately reduced how much DRAM they ordered for their safety stock over the summer of 2025 because tariffs were changing almost weekly. Every RAM purchase risked being made at the wrong moment – and so fewer purchases were made.

2. Prices had been falling all summer. Because of the hesitancy to purchase as much safety stock as usual, RAM prices were also genuinely falling over time.  And, obviously when memory is getting cheaper month over month, the last thing you’d feel is pressured to buy a commodity that could be cheaper the next month…so everyone waited.

3. Secondary RAM Manufacturing Had Stalled. Budget brands normally buy older DRAM fabrication equipment from mega-producers like Samsung when Samsung upgrades their DRAM lines to the latest and greatest equipment.  This allows the DRAM market to expand more than it would otherwise because it makes any upgrading of the fanciest production lines to still be additive change to the market. However, Korean memory firms have been terrified that reselling old equipment to China-adjacent OEMs might trigger U.S. retaliation…and so those machines have been sitting idle in warehouses since early spring.

Yep, there was no cushion. OpenAI hit the market at the exact moment it was least prepared.  

And now time for the biggest twist of all, a twist that’s actually public information, and therefore should be getting discussed by far more people in this writer's opinion: OpenAI isn’t even bothering to buy finished memory modules! No, their deals are unprecedentedly only for raw wafers — uncut, unfinished, and not even allocated to a specific DRAM standard yet. It’s not even clear if they have decided yet on how or when they will finish them into RAM sticks or HBM!  Right now it seems like these wafers will just be stockpiled in warehouses – like a kid who hides the toybox because they’re afraid nobody wants to play with them, and thus selfishly feels nobody but them should get the toys!

And let’s just say it: Here is the uncomfortable truth Sam Altman is always loath to admit in interviews: OpenAI is worried about losing its lead. The last 18 months have seen competitors catching up fast — Anthropic, Meta, xAI, and specifically Google’s Gemini 3 has gotten a ton of praise just in the past week. Everyone’s chasing training capacity. Everyone needs memory. DRAM is the lifeblood of scaling inference and training throughput. Cutting supply to your rivals is not a conspiracy theory. It’s a business tactic as old as business itself.  And so, when you consider how secretive OpenAI was about their deals with Samsung and SK Hynix, but additionally how unready they were to immediately utilize their warehouses of DRAM wafers – it sure seems like a primary goal of these deals was to deprive the market, and not just an attempt to protect OpenAI's own supply…

Alright, now that we are done explaining the how, let’s get to the “now what?” – because even if the RAM shortage miraculously improves immediately behind the scenes – even if the AI Bubble instantly popped or 10 companies started tooling up for more DRAM capacity this second (and many are, to be fair), at a minimum the next six to nine months are already screwed.  See above: DRAM manufactures are quoting 13-Month lead times for DDR5!  This is not a temporary blip. This could be a once-in-a-generation shock. So what gets hit first? What gets hit hardest? Well, below is an E through S-Tier ranking of which products are "the most screwed":

• S-Tier (Already Screwed – Too Late to Buy) -

• A-Tier (Almost Screwed – Don’t Wait to Buy!!!)

  • SSDs. These tends to follow DRAM pricing with a lag.

  • Small Prebuilt PC Companies. That lack large buffers of inventory.

  • RADEON GPUs. AMD doesn’t bundle RAM in their BOM kits to AIBs the way Nvidia does. In fact, the RX 9070 GRE 16GB this channel leaked months ago is almost certainly cancelled according to our sources.

  • XBOX. Microsoft didn’t plan. Prices may rise and/or supply may dwindle in 2026.

• B-Tier (Eventually Screwed – Don’t wait much longer to buy!)

  • Nvidia GPUs. Nvidia maintains large memory inventories for its board partners, giving them a buffer. But high-capacity GPUs (like a hypothetical 24GB 5080 SUPER) are on ice for now because those stores were never sufficiently built up. In fact, Nvidia is quietly telling partners that their SUPER refresh “might” launch Q3 2026 – although most partners think it’s just a placeholder for when Nvidia expects new capacity to come online, and thus SUPER may never launch.

• C-Tier (Think about buying soon)

  • Laptops and phones. These companies negotiate immense long-term contracts, so they’re not hit immediately. But once their stockpiles run dry, watch out!

• D-Tier (Consider buying soon, but there’s no rush)

  • PlayStation. Sony planned better than almost anyone else. They bought aggressively during the summer price trough, which is why they can afford a Black Friday discount while everyone else is raising prices.

• E-Tier (Prices might actually drop)

• ???-Tier —Steam Machine. Valve keeps things quiet, but the big unknown is whether they pre-bought RAM months ago before announcing their much-hyped Steam Machine. If they did already stockpile an ample supply of DDR5 - then Steam Machine should launch fine, but supply could dry up temporarily at some point while they wait for prices to drop. However, if they didn’t plan ahead - expect a high launch price and very little resupply…it might even need to be cancelled or there might need to be a variant offered without RAM included (BYO RAM Edition!).

And that’s it! This last bit was the most important part of the article in this writer's opinion – an attempt at helping you avoid getting burned. Well, actually, there is one other important reason for this article’s existence I'll tack onto the end – a hope that other people start digging into what’s going on at OpenAI.  I mean seriously – do we even have a single reliable audit of their financials to back up them outrageously spending this much money…for this?  Heck, I’ve even heard from numerous sources that OpenAI is “buying up the manufacturing equipment as well” – and without mountains of concrete proof, and/or more input from additional sources on what that really means…I don’t feel I can touch that hot potato without getting burned…but I hope someone else will…