Key insights
- The article suggests investors should focus on the amount of memory packed into each AI chip rather than just the number of AI chips sold. This trend of increasing memory per chip per generation represents a significant growth driver for memory manufacturers like Micron, potentially overlooked by a focus solely on GPU sales. This could positively impact semiconductor companies involved in AI infrastructure.
Most investors believe Micron (MU +8.80%) must sell memory into more artificial intelligence (AI) servers to keep growing. That sounds reasonable.
After all, the artificial intelligence boom has sparked a massive wave of spending on data centers. Technology companies continue to build the infrastructure needed to train and run increasingly powerful AI models.
But investors may be focusing on the wrong number. Instead of tracking how many AI chips companies sell, they should focus on how much memory engineers pack into each chip. That distinction may seem small. But it may represent one of the most overlooked trends in the entire AI investment story.
Most investors watch GPU sales. The logic seems straightforward. More AI spending leads to more GPUs, which in turn create greater demand for memory.
This Card Just Received a Rare 5-Star Rating
Our team of credit card pros don’t just recommend this card—they actually use it. Motley Fool Money calls it a top pick for a reason.
But another trend has quietly emerged. Each new generation of AI hardware requires significantly more memory than the previous one.
Consider what has happenedover the past few years:
When our analyst team has a stock tip, it can pay to listen. After all, Stock Advisor’s total average return is 936%* — a market-crushing outperformance compared to 209% for the S&P 500.
They just revealed what they believe are the 10 best stocks for investors to buy right now, available when you join Stock Advisor.
Here, the exact figures matter less than the trend -- that every new generation requires more memory than the one before it. That trend creates a second growth driver for Micron.
A useful way to think about an AI system is to imagine a chef working in a busy kitchen.
The GPU serves as the chef. Memory serves as the countertop. A larger countertop allows the chef to work with more ingredients simultaneously. A smaller countertop forces the chef to stop and reorganize.
AI systems face the same challenge. As AI models grow larger and more sophisticated, they need access to more information simultaneously. Particularly, reasoning models now perform more computational steps before producing an answer. Moreover, applications now process text, images, audio, and video simultaneously.
Each of those trends increases memory requirements. As a result, memory now plays a larger role in overall AI performance than it did just a few years ago.
This is where the investment thesis becomes interesting. Many investors tie Micron's future directly to growth in AI servers. If AI spending slows, Micron slows. Simple.
But reality may not work that way. Imagine a world where AI server growth eventually moderates. Most investors would view that outcome as bad news for companies that supply AI infrastructure.
However, Micron could still generate meaningful growth if each new generation of hardware requires substantially more memory. In other words, Micron does not need AI chip sales to grow at today's pace forever. The company may simply need more memory content per chip to keep rising.
Of course, rising memory content does not guarantee success.
Memory has historically operated as one of the most cyclical industries in technology. When demand rises, manufacturers build capacity. Supply eventually catches up, leading to falling prices and shrinking margins.
The industry has repeated that cycle many times over the decades.
Hence, the key question is whether HBM -- the memory specially designed for AI data centers -- changes the equation.
If HBM remains difficult to manufacture and increasingly critical to AI performance, Micron could deliver stronger returns than in previous memory cycles. If HBM becomes another commodity product, history could repeat itself.
Investors often ask how many AI chips companies like Nvidia will sell next year. Micron investors should ask a different question: How much memory will those chips need?
So far, the answer appears simple. More. And then more. And then more again.
If that trend continues, Micron could capitalize on a powerful growth driver that many investors still overlook, because Micron's future may depend less on the number of AI chips companies sell and more on the amount of memory those chips require.
And that trend is the single most important metric that investors should monitor in the coming quarters.