Kioxia Corporation and Sandisk Corporation have introduced their next-generation Quad-Level Cell (QLC) 3D flash memory technology, achieving what the companies describe as the industry’s highest bit density for QLC NAND. The new technology delivers up to a 60% increase in bit density compared with their eighth-generation flash memory and exceeds 37 Gb/mm², establishing a new benchmark for density, performance and power efficiency.

The announcement comes as artificial intelligence, hyperscale data centers and cloud computing continue to drive rapid growth in data creation and storage requirements. Consequently, storage vendors are under increasing pressure to deliver higher-capacity solutions that also improve energy efficiency and performance.

Advanced CBA Architecture Drives Density Gains

At the core of the new technology is the companies’ CMOS directly Bonded to Array (CBA) architecture. The approach involves manufacturing CMOS logic and memory arrays on separate wafers before bonding them together using high-precision wafer-to-wafer alignment.

As a result, Kioxia and Sandisk have been able to enhance density scaling, improve manufacturing efficiency and increase performance. The new solution utilizes a 332-layer architecture and an optimized floorplan design to achieve its density improvements.

In addition, the technology becomes the first QLC 3D flash memory to reach a 4.8 Gb/s interface, enabled by Toggle DDR6.0 and the Separate Command Address (SCA) protocol. These advancements are designed to unlock faster data transfer speeds while supporting increasingly demanding enterprise and AI workloads.

Improved Power Efficiency Addresses AI Challenges

Beyond performance gains, the companies emphasized significant improvements in energy efficiency. Interface enhancements and the integration of Power-Isolated Low-Tapped Termination (PI-LTT) technology improve I/O data-out transfer power efficiency.

These enhancements are particularly important as data center operators seek to reduce power consumption and address cooling requirements associated with large-scale AI deployments.

“As AI continues to underpin the advancement of society, its applications are expanding from generative AI to agent-based and physical AI, while the use of data is becoming increasingly diverse and sophisticated,” said Hideshi Miyajima, Chief Technology Officer at Kioxia.

He added that QLC technology enables efficient storage of rapidly growing data volumes and supports greater scalability and performance for AI systems.

Focus Turns Toward Next-Generation Infrastructure

The companies believe the new memory technology will help support future infrastructure requirements as AI training, inference and hyperscale computing workloads expand.

“As AI training, inference, and hyperscale datacenter workloads drive unprecedented growth in data generation, the storage industry faces increasing pressure to deliver greater capacity, higher performance, and improved energy efficiency,” said Alper Ilkbahar, Chief Technology Officer at Sandisk.

According to Sandisk, the 10th-generation QLC 3D flash memory simultaneously advances density, bandwidth and energy efficiency, creating a scalable foundation for next-generation infrastructure applications.

10 August 2026