Collaboration CEA Leti FMC to Advance FeRAM Cache+ Chiplet for AI Systems

FeRAM transmission electron microscopy image showing memory device/array integration by CEA-Leti on top of 22-nm FD-SOI CMOS (source: CEA-Leti)

Two leading Ferro4EdgeAI partners, CEA Leti and FMC, have announced a strategic collaboration to develop FMC’s Cache+ chiplet, a breakthrough in high-density, low-power memory for next-generation AI systems. This innovative technology leverages 1T-1C FeRAMs based on HfZrO₂, offering a compelling alternative to standard SRAMs.

CACHE+ reimagines conventional SRAM, commonly used as high-speed cache throughout computing systems, by delivering ten times the density of traditional SRAM while reducing standby power by a factor of ten. Beyond efficiency gains, CACHE+ adds persistence, ensuring data retention even in power-off scenarios.

 

Designed to integrate seamlessly into chiplet architectures for ultra-high-performance AI inference systems, CACHE+ empowers system designers to rethink computing architectures from the ground up. These advancements are particularly crucial for AI data centers, where power consumption is a growing concern, and efficient memory solutions are key to sustainable scaling.

 

Learn more about this collaboration and the CACHE+ technology: FMC & CEA Leti join forces

Previous
Previous

Ferro4EdgeAI Highlighted in EE Times Europe: NV Solutions for Edge AI

Next
Next

Insights from SAFS 2025: Advancing Hf₀.₆Zr₀.₄O₂ Ferroelectric Research