Investors tracking Micron's AI growth should focus on memory capacity per chip, not just chip sales volume—a trend that could sustain the company's growth even if AI hardware spending slows.

What happened: Each new generation of Nvidia's AI accelerators requires substantially more memory than the previous one. The H100 contained roughly 80 gigabytes of high-bandwidth memory (HBM), the H200 increased that to approximately 141 gigabytes, and the latest Blackwell chips reached roughly 192 gigabytes. This pattern of rising memory demands per chip creates a second growth driver for Micron beyond simple GPU sales volume.

Why it matters: As AI models grow larger and more sophisticated, they need access to more information simultaneously—particularly reasoning models that perform more computational steps before producing an answer, and applications that process text, images, audio, and video at once. Memory now plays a larger role in overall AI performance than it did a few years ago. This means Micron could generate meaningful growth even if AI server growth eventually moderates, because each chip would require more memory content than before.

What to watch: The critical question is whether HBM—the memory specially designed for AI data centers—remains difficult to manufacture and increasingly critical to AI performance. If HBM stays scarce and essential, Micron could deliver stronger returns. If HBM becomes a commodity product, the industry may repeat its historical memory-price-decline cycles.