DRAM prices have risen 300-1000% since early 2025 due to AI datacenter demand cannibalizing consumer supply

Samsung, SK Hynix, and Micron have shifted DRAM fab capacity toward HBM (High Bandwidth Memory) for AI accelerators, causing a severe supply squeeze on standard DDR5 and DDR4 modules. A 32GB DDR5-6000 kit that cost $95 in mid-2025 hit $400 by December 2025, and 32GB DDR4 kits doubled from $60-$90 to $150-$180. So what? Small and mid-sized hosting companies, independent PC builders, and IT departments at budget-constrained organizations cannot afford to provision or upgrade servers and workstations. So what? They delay hardware refreshes, which means running older, slower machines longer, reducing employee productivity and increasing maintenance costs. So what? Startups and bootstrapped companies that depend on physical infrastructure (game studios, VFX houses, on-prem ML teams) face ballooning capex that eats into runway. So what? Some are forced to cut headcount or delay product launches because hardware budgets consumed funds earmarked for engineering salaries. So what? The entire mid-market hardware ecosystem stagnates while hyperscalers with billion-dollar budgets absorb all available supply, widening the resource gap between big tech and everyone else. This persists structurally because memory manufacturing is an oligopoly (three companies control 95%+ of DRAM production), capital expenditure cycles for new fabs take 2-3 years, and the financial incentive to serve AI customers at premium margins far exceeds serving consumer markets at commodity margins.