The Next Key Player in the AI Boom Isn't GPUs — A New Chain of Price Increases Begins with "Memory Shortages"

AI's Next Target: Not GPUs, but Memory

The discussion around generative AI tends to focus heavily on GPUs. However, the real bottleneck emerging in the field is "memory/storage," which includes DRAM (main memory), NAND (flash), and even HDDs.

According to a WSJ report highlighted by Seeking Alpha, the global rush to build AI infrastructure is tightening the supply of memory chips and storage devices, while manufacturers are not in a hurry to ramp up production. The key question is, "Why aren't they increasing production despite profiting from the shortage?"

What’s Happening: Demand is Shifting in One Direction

AI data centers don't just need GPUs. The process of learning and inference involves reading and writing massive amounts of data and storing model weights and intermediate results, which makes server-side DRAM capacity and bandwidth crucial. SSDs (NAND) and HDDs are also necessary for storage, distribution, and backup.
Reuters reports that the surge in AI demand has elevated memory shortages to a "new supply chain crisis," potentially doubling prices in some areas and drastically thinning inventories. Additionally, as production shifts towards high-value AI memory (like HBM), the supply of general-purpose DRAM and flash is reduced, impacting other applications.

This "concentration of demand in the same direction" is problematic. Previously, if smartphones were weak, PCs were strong, and vice versa, creating a "cancellation" effect. However, AI pushes up DRAM, NAND, and HDD demand simultaneously through data center investments. Demand comes in waves, leaving few escape routes for supply.

Why Production Isn’t Increasing: Memory Industry’s "Trauma" and Investment Time Lag

So, why aren't manufacturers stepping on the gas? The WSJ's main point is clear: the memory industry has repeatedly experienced "increased production → oversupply → price collapse → massive losses." Even if factories and equipment are expanded during boom times, by the time operations ramp up, the economy has often reversed, leading to "overproduction" that chokes the industry—a nightmare they don't want to repeat.

Another factor is time. Semiconductor production capacity cannot increase overnight. It takes years for plant construction, equipment installation, startup, and yield improvement. Therefore, manufacturers are more inclined to "maintain supply discipline and sustain high profitability" rather than overinvest in response to short-term shortages.

The WSJ also suggests that weak "long-term supply commitments (long-term purchase agreements)" are a condition for increasing production. Essentially, "if there's a guaranteed purchase agreement, we'll produce. But if it's just spot demand, we won't." This negotiation game delays supply ramp-up, prolonging shortages.

The "Ripple Effect" Likely in 2026: PCs, Smartphones, and Peripherals Will Have a Subtle Impact

Shortages and price increases first become apparent in enterprise servers and then spread to consumer markets. The Boston Globe reports that the AI boom-induced memory shortage could lead to price increases for PCs and smartphones by 2026.

The concern here is that price hikes won't come "all at once," but rather "gradually and repeatedly." Device manufacturers dislike passing on costs, so

• it's harder to increase memory capacity at the same price (effectively a price increase)

• price increases may occur depending on the price range

• supply of some models may dwindle
  These effects are more subtle than the clear "out of stock frenzy" seen during GPU shortages, leading to situations where prices are higher and choices are fewer before you realize it.
  
  
  

How Do Investors View This? Expectations and Caution of a "Super Cycle"

With tight supply and rising prices, memory manufacturers are strong in the short term. Reuters reports that memory-related stocks have been bought amid global supply shortages, and the market is aware of a potential "super cycle."

However, the term "super cycle" itself can be overvalued in situations where supply constraints create the supply-demand dynamics, leading to a state where investors have mixed feelings (both bullish and cautious).

SNS Reactions: Divided into Four Discussion Points

On social media and forums, this topic generally breaks down into the following four patterns.

1) "Please, no more expensive PCs because of AI" group (consumer perspective)
In Reddit's tech communities, there are posts about already feeling the rise in RAM prices and sarcastic comments about the impact extending to cars and appliances.

2) "Memory is ultimately cyclical. Not increasing production is also 'learning'" group (industry perspective)
The cautious stance of manufacturers is seen as rational. Those familiar with past crashes tend to view this decision as "discipline."

3) "How true is this? Is it just hype?" group (skeptical/oppositional)
Similarly on Reddit, there are doubts about whether AI alone can cause a global shortage and whether prices are being "manufactured" beyond demand.

4) Investors' "wait and see" and profit-taking caution (market perspective)
On Japanese forums (Yahoo! Finance), there are cautious comments about "waiting and seeing for a while," based on news quotes.

Interestingly, the same news can simultaneously create a "tailwind for related stocks" and a "headwind for living costs." The divide between those who profit (memory/storage manufacturers) and those who suffer (device manufacturers and consumers) makes timelines prone to a "mix of pros and cons."

Conclusion: The "Real Shortage" of the AI Boom Spreads to the Periphery

The crux of this story lies not in the memory shortage itself, but in the "structure where production doesn't immediately increase despite shortages." Manufacturers have learned from past failures and avoid big bets without long-term contracts. It takes time to enhance facilities, making it difficult for supply to keep up with the speed of AI investment.

2026 might be a year where "subtle price increases," often overlooked when focusing only on GPUs, spread to PCs, smartphones, and storage devices, affecting everyday areas.