AI Demand to Cause Severe HBM Memory Shortage in Next Two Years
RihardJarc · x · 2026-08-09
Investor Rihard Jarc provides a deep analysis of the memory supply and demand cycle driven by AI, particularly for HBM. He argues that if AI demand continues its steep growth, there will not be enough memory supply coming online for the next two years. The cycle could only be broken by a major shift in inference software architecture (he doubts Kimi K3's hybrid mechanism is it) or scaling laws hitting a wall.
He notes that memory companies are using LTAs to cap gross margins and avoid killing value generation downstream. Unlike fixed tech gadgets, memory is now used to produce direct intelligence, acting more like a commodity where falling prices increase demand. Furthermore, the future adoption of consumer AI devices and humanoids could trigger a new memory bull cycle.
More from Venture
- AI Personalized Story App Simmy Hits $1M+ Run Rate in 90 Days — DhruvBhalla_ · 2026-08-10
- Heavy AI Costs Force Canva to Cut 2026 Revenue Forecast to 20% — SumitGup · 2026-08-09
- From Open Source to 180-Member Community: AI Solopreneur Playbook — sujingshen · 2026-08-09
- User to Launch Solana Memecoin Promoted by Seedance Videos on Claude's Advice — beechinour · 2026-08-09
- Local Professional Services Quietly Out-Earning Indie Hackers with Just 6 Pages — gaganghotra_ · 2026-08-09
- SEO Wins Are Amplified in AI Search for Mid-Market Brands — gaganghotra_ · 2026-08-09