Particle.news
Download on the App Store

AI Hardware Boom Redraws Economics: Memory Makers Capture Profits as Firms Build 'AI Factories'

Rising demand for always-on agent-style AI is driving large DRAM and flash price gains that are redirecting profits to memory and equipment suppliers.

Overview

  • NVIDIA CEO Jensen Huang said on June 24 that companies are building long-lived AI data centers he called “AI factories,” framing AI infrastructure as a multi-decade project that will sustain massive compute and storage needs.
  • Jefferies forecasted sharp memory price jumps in 2026 with DRAM and flash gains of roughly 40%–50% quarter over quarter in Q3 and further rises in Q4, and it expects price relief only around 2028.
  • Stronger-than-expected guidance from suppliers such as Micron and Qualcomm triggered a rapid market revaluation that lifted semiconductor sector market caps by hundreds of billions of dollars in a single session.
  • Industry analysis shows profits concentrating in the hardware stack—HBM, server DRAM and enterprise flash vendors report outsized margins while model-layer businesses face pressure to cut per-token prices.
  • The push to embodied AI and long-running agents will raise sustained token and storage demand, which means supply limits in high-bandwidth memory and validated server parts could keep costs high and shape investment, talent and policy choices for years.