Building NAND chips from ferroelectric transistors (FeFETs) could dramatically cut power consumption, according to new research from the Samsung Advanced Institute of Technology (SAIT). The company says it has finally solved the long-standing problems of capacity scaling and memory windows that undermined earlier efforts. With the AI boom driving the construction of enormous data centers, demand for memory keeps climbing, and with it the already high power draw of today's NAND chips. Samsung's researchers are proposing a new architecture that could sharply reduce that energy cost. Mobile devices could also benefit if the findings are widely adopted.
FeFETs: Eliminating Pass Voltage and Cutting Power by 96%
Traditionally, NAND chips scale by stacking layers on top of one another. The problem is that to make full use of the memory capacity, power has to be pushed through those layers in sequence. This voltage (known as Pass Voltage) rises with every layer added, so total power consumption climbs along with it. Earlier attempts to limit power draw shrank the memory window. Even other proposals based on ferroelectric materials could not resolve the trade-off between capacity scaling and energy efficiency.
Samsung's approach combines oxide semiconductors with ferroelectric structures to form ferroelectric field-effect transistors (FeFETs), which removes those limitations. FeFETs eliminate Pass Voltage entirely, cutting power consumption by up to 96% compared with conventional NAND chips. At the same time, they sustain a multi-level cell density of up to five bits per cell, on par with or better than today's most advanced memory. Samsung proposes scaling the FeFET design by stacking transistors with short 25 nm channels.
The findings, produced by more than 30 researchers from Samsung's institute of technology and its semiconductor R&D center, were recently published in the journal Nature. The work demonstrates innovation in both materials and memory chip architecture in the semiconductor industry.
Implications for AI and the Future of Battery-Free Devices
AI processors and data centers require so much memory that they have disrupted the NAND industry, with shortages expected to last at least through 2026. Cutting memory power by nearly 100% could significantly reduce the total energy consumed by AI hardware. The arrival of FeFETs promises to address the energy challenge of building enormous data centers.
Samsung is not alone in this field. Other companies are also exploring ferroelectric paths to persistent, non-volatile memory, the kind of memory that retains data after power is lost. Whether this approach will prevail over competing ones remains to be seen, but the potential impact could be huge for everything from large-scale computing to everyday consumer technology.
The researchers position the technology as a potential route to autonomous, maintenance-free sensor systems that would draw energy directly from their surroundings. This development opens the door to mobile devices that could significantly reduce their dependence on batteries, a major step forward for the consumer electronics industry.
