NVIDIA, Samsung Develop Ferroelectric NAND to Boost AI Accelerator Performance and Cut Power Use

Rapidly rising computing demand at AI data centers has made memory supplies and power consumption bottlenecks to expansion. NVIDIA and Samsung Electronics are therefore collaborating on ferroelectric NAND, whose high-speed, low-power characteristics could improve data-transfer efficiency and strengthen memory stacking in AI accelerators. The technology could also reduce reliance on the existing NAND supply chain.

As of July 19, 2026, reports said ferroelectric NAND was expected to be incorporated into NVIDIA's next-generation Vera Rubin AI accelerators and could cut power consumption by as much as 96%. AI models have also made technical analysis nearly 10,000 times faster. NVIDIA and Samsung Electronics, however, have yet to disclose the investment amount, a formal mass-production date or full product specifications.