The scarcest ingredient in artificial intelligence may no longer be the flashy graphics chips that power chatbots and data centers. It's memory.
According to The Motley Fool, Nvidia CEO Jensen Huang told the audience at CES 2026 that memory is now the biggest bottleneck in AI. In plain terms, the industry has plenty of raw computing horsepower, but the high-speed memory that feeds those processors has become the limiting factor.
The Motley Fool reports that this shift has reshaped where investors are placing their bets. Since Huang's remarks, shares of memory makers Micron and SanDisk have outperformed Nvidia's own stock — a notable reversal for a market that has spent the past few years treating Nvidia as the single most important name in the AI trade.
Why the change? As AI models grow larger and more demanding, they need to store and move enormous volumes of information quickly. Processors can only run as fast as the memory supplying them, so when memory becomes the constraint, the companies that make it gain leverage.
That helps explain the market's rotation toward Micron and SanDisk, two firms whose products sit at the center of this squeeze.
A word of caution: the underlying source is an investment-focused article, and past stock performance does not guarantee future results. Huang's assessment reflects one influential voice at a single industry event.
Why it matters: If the AI industry's tightest constraint is shifting from processors to memory, the balance of power — and profit — in the chip world could be moving toward companies that were, until recently, overshadowed by Nvidia.