The recent RAM market crisis is no longer a topic for tech enthusiasts only. It has reached a level where every person using any electronic device is going to be affected in the upcoming months, and it probably is not going to stop in the visible future. Analysts are divided in their assumptions; some are saying it will take until 2028, while others say it will take until 2030 to become slightly normal. But anyone with a basic understanding of how tech businesses work is bound to be skeptical about that too. So, let's try to understand what the hell is actually going on here.
What is Causing the Global RAM Shortage?
The electronic chip needed for RAM is called DRAM (Dynamic Random Access Memory) which is also the building chip for some other stuff like HBM (High Bandwidth Memory), VRAM (Video Random Access memory) and many other components that work with fast memory. The reason I mentioned HBM and VRAM is that these two are related to the current crisis the most.
Guess what they are used for. It is again AI, AI, AI, AI. The thing tech businessmen have gone insane about is that mentioning it once at a time feels like a sin. Anyways, AI needs a lot of fast-working memory. The billions or trillions of parameters we see tech giants keep bragging about with every launch of a new AI model need fast memory for that processing. Otherwise, you cannot work with a massive amount of data. And AI is mostly about how fast you can process big datasets.
The DRAM chips are obviously made from semiconductor wafers, like almost every electronic chip. But the amount of wafer needed to make HBM is 3 to 4 times higher than that of consumer RAM. I mean, this much wafer gives you the same gigabytes of memory. And AI data centers crave this HBM because, although it takes a significantly larger amount of physical components, it provides a lot more efficiency than consumer RAM. And who doesn't know that B2B deals get you more money than B2C? Therefore, manufacturers of DRAM flushed out all their manufacturing of DRAM for consumer RAM and shifted to producing HBM as much as possible.
Now the other thing is GPUs. AI datacenters need a lot of GPUs for processing. GPUs need a lot of DRAM chips too.
The Big Three Monopolizing the DRAM Market
Samsung, SK Hynix, and Micron Technology are the big three in DRAM manufacturing, and they control 90% of the total DRAM production in the world. SK Hynix was the frontrunner in HBM production for the AI boom, and the remaining two caught up by mid-2024 with the latest HBM4 generation. All three have shifted their production to HBM from consumer RAM as it generates more revenue. They also signed deals with GPU companies like NVIDIA for GDDR memory units as they perfected the technology of DRAM for that.
The Big Three are now only providing DRAM for consumer RAM to long-term big business partners. They are allotting the rest of their capacity to HBM and GPUs. Supply has decreased so much relative to demand that RAM prices have increased about 130% within this short time period.
A question comes to mind: is it going to be fixed in the near future? No one can say for sure. Analysts predicting the crisis to subside within 2028 to 2030 are still likely optimistic in their projections. Because the only promising DRAM manufacturer coming up right now is CXMT (ChangXin Memory Technologies), a Chinese company.
Can China's CXMT Save Budget PC Builders?
Considering CXMT as the ray of hope for tackling the RAM crisis is a long shot. Their growth in scaling cannot meet the demand scarcity in the near future as they only hold 7.67% to 8% of the global market currently. Also, come on, as they gain scale, they are definitely going to sell at the current high price. Since the supply is already so low, they are likely to be the only big player in consumer RAM in the near future. There is no reason for them to sell chips at competitive prices if there is no competition. This is how business works. You take the lion's share of profit whenever possible. Besides, CXMT is currently focusing on consumer RAM rather than HBM because they are not yet able to sell to data centers. Big tech giants that make GPUs and data centers still consider them 3-4 years behind the Big Three in making HBM and VRAM for GPUs. Though it might be a geopolitical thing to push Chinese brands downward or maybe to keep the tech monopoly of giants in place, that is a different question. Also, CXMT will start shifting DRAM production to HBM and VRAM the moment they get approved for that—for the same reason the Big Three have shifted, choking the consumer market.
Will RAM Prices Ever Go Down?
The current RAM scarcity is going to make budget computers and mobile phones almost extinct in upcoming years. Especially for computers, any laptop or desktop under $500 will be unlikely to survive.
The AI boom is not likely to stagnate in upcoming years; we can see that the more the public loses trust in AI, the more tech executives bet on the AI gamble. And there isn't really any proper plan for how to generate revenue from AI. They have started this berserk race, and now they can't just pull out. The more time goes by, the more their financial losses multiply if they pull out.
To sum up, in upcoming years, RAM prices don't seem likely to go down unless the AI bubble starts popping. If that happens, no one can say what will occur because the whole tech world and economy will fall into chaos. And there is no predicting chaos; you just see where it goes. But what we can predict until then is that tech is going to be unaffordable for middle-class to lower-class people. Students previously could manage to get a good enough device to learn engineering and specialized work, even if not high-end ones. But that is not going to be possible onwards. Only lightweight devices will be affordable with the highest effort, and those are quite useless for heavy works other than basic use. Keep in mind that my objective isn't to demolish your hope; rather, I am just making you aware of the ground reality we're being dragged into by money-hungry tech giant gamblers.