Rumor
Samsung and Nvidia co-develop ferroelectric NAND using AI simulations 10,000x faster than traditional TCAD
Friday, March 13, 2026 at 01:29 AM
Samsung and Nvidia are reportedly collaborating on ferroelectric NAND (FeNAND) development using AI-driven simulation techniques. This new approach is claimed to be 10,000 times faster than traditional TCAD (Technology Computer-Aided Design) methods and could potentially lead to a 96% reduction in power consumption for storage components in AI infrastructure.
Context
On March 12, 2026, Samsung Electronics and Nvidia announced a strategic partnership to co-develop ferroelectric NAND (Fe-NAND), a next-generation memory technology aimed at resolving the dual crises of chip shortages and surging data center energy demands. This collaboration is highly unusual as it involves Nvidia directly participating in memory device research. The partnership leverages a Physics-Informed Neural Operator (PINO) AI simulation framework that operates 10,000x faster than traditional TCAD tools, reducing simulation times from days to seconds.
This breakthrough is expected to enable the realization of 1,000-layer NAND stacks, far surpassing current vertical scaling limits. The technology offers a massive 96% reduction in power consumption compared to conventional NAND, addressing the critical power constraints of modern AI infrastructure. Nvidia plans to integrate this technology as Inference Context Memory Storage (ICMS) in its upcoming Vera Rubin architecture, a move estimated to consume nearly 9.3% of global NAND supply.
Sources (10)
Samsung Electronics partners with Nvidia to build world ...Samsung, Nvidia Partner on Ferroelectric NAND for 1,000-Layer ...Samsung touts 96% lower-power NAND design — researchers investigate design based on ferroelectric transistors | Tom's Hardware[2603.06881] Physics-informed AI Accelerated Retention Analysis of Ferroelectric Vertical NAND: From Day-Scale TCAD to Second-Scale Surrogate ModelSamsung Electronics Accelerates Development of 'Dream ...Ferroelectric NAND for efficient hardware bayesian neural networks | Nature CommunicationsAnsys, NVIDIA Partner for CAE - Engineering.com
Middle Interlayer Engineered Ferroelectric NAND Flash Overcoming Reliability and Stability Bottlenecks for Next‐Generation High‐Density Storage Systems - PMC
Related Companies
Nvidia
NVDA
Samsung Electronics
005930