Micron Technology reported record earnings on Wednesday as demand for its high-bandwidth memory chips outpaced supply by an estimated 50% to 67%. The memory specialist’s fiscal Q3 results showed gross margins at 81%, more than double the prior year’s 39%.

The rally reflects a fundamental shift in AI infrastructure. Every large language model, every training cluster, every GPU array needs Micron’s specialized memory. Micron has said its HBM capacity is completely sold out through 2026. Customers are paying premium prices and waiting months for delivery.

This supply crunch extends across the entire memory market. SK Hynix and NVIDIA announced a multiyear partnership to develop next-generation memory specifically designed for AI factories. Samsung and others are ramping production. Yet demand keeps outrunning supply.

Micron trades at $1,051 per share, up 244% in 2026 alone. The company reached a $1 trillion market capitalization, joining only two other memory manufacturers in that exclusive club. For the broader AI infrastructure buildout, Micron is no longer a supplier—it’s a bottleneck.

The shortage has forced data center operators to make difficult choices. Some are delaying projects. Others are signing long-term contracts at inflated prices. Bloomberg reports that major cloud providers are negotiating exclusive supply agreements directly with manufacturers. Competition for scarce memory is remaking the entire supply chain.