History shows that DRAM is a brutally cyclical business with boom and bust cycles. It's also heavily capital intensive. You build an expensive fab now, and then demand plummets and you're stuck selling at near cost while carrying enormous fixed assets.

IBM is already not the best at making strategic calls. I can't imagine them being saddled with DRAM business with its repeated busts.

There is graveyard of DRAM companies that never made it.