Memory supply is under strain.

RAM prices have skyrocketed because running and training AI models require far more memory than traditional computing. Training, inference clusters, and GPU servers consume enormous amounts of DRAM and high-bandwidth memory, and suppliers are pivoting to meet that demand.

It's similar to what happened with GPUs during crypto-mining peaks. When software requires more hardware than the market can supply, prices adjust accordingly.

Micron, one of the largest semiconductor companies in the space, confirmed that it will retire Crucial, its consumer-memory brand, and focus on high-bandwidth memory used in the enterprise.

Reports suggest Nvidia may stop bundling VRAM with some GPUs, reflecting the same supply pressure, though the company has not confirmed this. Both changes demonstrate how the memory market is shifting toward enterprise AI rather than consumer hardware.

The result is quite the contrast from just a year ago, when RAM was inexpensive and widely available. As production pivots to AI, the cost of building or upgrading systems rises for everyone else, from PC builders to small companies running their own servers.