ChipsJuly 28, 2026via Financial Times Technology

Maybe the chips are down because memory’s too expensive

Why it matters

Memory pricing is constraining AI chip deployment economics. As GPU costs plateau, DRAM expenses are emerging as the underestimated cost driver limiting large-scale model inference and training infrastructure buildouts.

Key signals

  • DRAM pricing pressure identified as infrastructure constraint
  • Memory costs competing with GPU economics for AI capex budgets
  • July 2026 analysis of chip supply chain cost structure
  • Inference/training deployment economics impacted by memory component pricing

The hook

DRAM costs just became the silent bottleneck killing AI infrastructure ROI.

DRAM, DRAM, DRAM / Did I ruin your plans, plans, plans?

The week's key stories, every Friday.

For practitioners and enthusiasts — free, in your inbox.

Free forever. No spam.