Rohan Paul@rohanpaul_ai · X美光市值已突破1万亿美元，一年前其市值还仅为700亿美元。这一飞跃的核心驱动力是AI发展对高带宽内存（HBM）的迫切需求。尽管GPU备受关注，但HBM已成为确保GPU持续高效工作的关键部件。随着AI智能体的兴起，无法快速将数据送达芯片已成为新的系统瓶颈，使得内存成为AI增长背后的关键约束。行业的焦点正从单一的逻辑芯片性能，转向整体数据流能否跟上AI推理和智能体工作负载的需求。此外，瑞银将其目标价从535美元大幅上调至1625美元，认为其长期供应协议可能平滑盈利的周期性波动。
Micron， the memory-maker for AI， just became a $1T company because AI's next shortage is memory， not models.
12 months ago， it was worth just $70 billion.
GPUs get most of the attention， but HBM has become the part that keeps them useful.
AI agents are making memory the new bottleneck of AI， because the smartest systems are useless when data cannot reach the chips fast enough.
High-bandwidth memory， or HBM， sits close to the accelerator and feeds it data at extreme speed， which is why it has become one of the quiet constraints behind AI growth.
The old semiconductor story was about who had the best logic chip.