Fast, Sub-Volt HZO Ferroelectric Capacitors For DRAM-Like Memory (SNU)

This is not an academic milestone—it is a potential inflection point in the memory architecture roadmap. Cracking the Kay-Dunn limit in HZO ferroelectric capacitors pushes FeFET/FeTRAM speed and endurance to DRAM-adjacent levels for the first time. Sub-volt operation structurally erodes the refresh-power advantage underpinning 3D DRAM economics. Upstream, Hf/Zr target suppliers (JX Nippon, KME) face a spec upgrade to ferroelectric-grade purity—a replay of the 2019 photoresist chokepoint, now in sputtering targets. Supply risk: Korea's vertical material-device-system loop is hardening, yet ~60% of Hf refining sits in China and Zr ore in Australia and South Africa. Target-material fragility is less visible than photoresist but equally systemic. Market dynamics: Micron's HBM and 3D-DRAM capex now carries route-disruption risk. Intel's Optane failure was precisely the FeFET speed gap; this breakthrough may force Samsung and Micron to re-price 2027+ roadmaps. CXMT has a 12-month window to validate equivalent ferroelectric stacks or cede the next non-volatile generation. 12–24 month tail: SNU IP will likely flow into SK Hynix or Samsung pilot lines via licensing within 18 months. The real long-tail effect sits in edge AI inference—when non-volatile read/write approaches DRAM latency, the compute-in-memory thesis gets repriced, and HBM's monopoly premium faces its first structural crack.