Why the pressure is severe: AI servers consume unusually large quantities of high-value memory. One industry assessment says AI-server DRAM costs roughly doubled in the first quarter of 2026 and projected much larger increases over the full year; other reporting quotes suppliers expecting no meaningful relief before 2028. These are market forecasts, not guaranteed outcomes.
Enterprise consequences: Higher memory bills can mean more expensive AI clusters, longer lead times, delayed refreshes, reduced memory specifications, and fewer servers available for customers without long-term supply agreements. Suppliers are prioritizing AI-oriented memory capacity, especially HBM and high-density server DRAM.
Consumer-hardware spillover: If DRAM and NAND capacity continues to be redirected toward AI data centers, smartphone, PC, TV, and graphics-card makers may face higher component costs, tighter inventories, or lower-memory base models. Retail-price increases are plausible but not automatic: vendors can absorb costs, cut other features, or accept lower margins. NAND is also exposed, with forecasts of continued contract-price increases rather than an isolated DRAM problem.
Other manufacturers: The evidence reviewed supports broad component-price pressure, but does not establish a comprehensive, verified list of specific IT hardware makers that have already raised prices. Claims of industry-wide price hikes should therefore be treated cautiously.
Inflation implications: Central-bank and international-policy analysis warns that AI can raise near-term inflation through a surge in spending on data centers, chips, power and related scarce inputs, even if AI eventually improves productivity. The ECB notes that expectations of permanently higher productivity can lift demand early in the transition, while the SNB has said AI could push inflation higher in the short term; the eventual net effect remains uncertain.
In short, the near-term effect is likely to be more expensive and less readily available memory-heavy equipment—especially AI servers—while the longer-run outcome depends on how quickly memory capacity expands and whether productivity gains become broad enough to offset the investment and supply shock.