Micron reports data center revenue more than tripled year-over-year in Q2 amid constrained high-bandwidth memory supply for AI chips.
Micron satisfied only 50% to 66% of customer demand for high-bandwidth memory (HBM) used in AI accelerators, driving gross margins up 54 percentage points in Q2. HBM pricing remains several times higher per bit than conventional memory as AI infrastructure demand accelerates.
Data center revenue surged over threefold year-over-year, reflecting structural pricing power in AI workloads. Competitors SK Hynix and Samsung may pressure margins later if they expand capacity aggressively, though current supply constraints persist.
Nvidia and AMD rely on HBM from Micron and other suppliers to power AI chips, with inference and agentic AI systems requiring constant memory capacity. The shift from training to deployment continues to strain semiconductor packaging and power infrastructure.