From the many movies and television shows I’ve seen that tackle the subject of AI, the lesson is often a cautionary tale, warning us humans about the hubris it takes to play God and so on. But now that humanity is well and truly in its AI era, I’m finding myself annoyed that more of those films failed to focus on the economic fallout of AI instead, because it’s driving me insane. 

The demand from AI data centres has created a shortage of global memory capacity, with prices for dynamic random access memory estimated to rise more than 400% from 2024 to 2026, according to J.P. Morgan. These price surges have naturally created massive concerns for the millions of businesses around the world that are now (already) dependent on AI. 

What Caused The Memory Shortage? 

The memory industry is well known for both feasts and famines, but the current shortage is far more concerning than others we’ve seen in years gone by. We are now in the midst of the worst memory shortage to ever occur. In history. Period.

As you can see by the title of the article, AI is the primary driver of this, or more specifically, the data centres required to continue to grow and develop the AI slice of the tech industry. The core issue is that data centres aren’t just hungry for memory; they’re ravenous. Each one requires large amounts of memory to process its gigantic data sets, and a different type of memory at that. 

The demand from these very hungry data centres is influencing major chip makers to intentionally shift their factory capacity away from standard consumer memory and towards high-margin High-Bandwidth Memory (HBM) and specialised DRAM, which are both needed for data centres, where each server also requires more memory, too. While an ordinary server requires 32GB to 64GB of memory, AI servers need 128GB, and in some instances, even more. 

It’s also worth bearing in mind that RAM isn’t exactly easy to manufacture. The industry itself has barely had time to recover from the chip shortage of 2020-2023. Though trade routes are no longer affected by the pandemic, the process of manufacturing chips remains just as complicated, requiring resources from around the world and necessitating precision. Without ultrapure water from Xylem to clean the chips and advanced robotics to put them together, products can easily end up contaminated or non-functional – even with the smallest errors. 

Are Actions Being Taken By Memory Suppliers? 

Of course not.

The memory component industry is booming right now, due to the stratospheric demand and ballooning prices. So why on Earth would they want to solve this problem? There is simply no incentive for them to increase the supply; in fact, there is very much an incentive to prolong this shortage for as long as possible. If the shortage ended, the price would drop, and they’d be worse off. 

What Does It Mean For The Consumers? 

Consumers are already feeling the effects of this memory crisis, with some DRAM kits surging by over 400% in the past year. Gamers in particular are being affected, with graphics cards up around 30% on the whole and consoles, like the PS5 and Xbox Series X, actually increasing by 50-60% since launch, depending on the product. So much for waiting for the prices to come down. 

What’s Next?

In the next couple of years, I wouldn’t expect much. Shortages are projected to remain just as severe, or sink to new lows, as AI demand continues to dominate factory output. Further into the future, our only hope is that the likes of Samsung, SK Kynix, and Micron continue to invest in new factories that drive aggressive industry expansion as a whole, leading to supply overtaking demand and forcing prices back down. Eventually. I hope.