Technology · Dev
Kioxia and SanDisk Roll Out QLC Flash Memory Targeting AI Data Centers
The new 3D flash technology promises higher storage density and power efficiency as hyperscalers race to meet AI workload demands

KEY TAKEAWAYS
- ·Kioxia and SanDisk introduced a new QLC 3D flash memory technology designed to boost storage density, bandwidth, and power efficiency for AI and cloud workloads.
- ·The technology addresses rising energy costs and cooling pressures in Asia-Pacific data center markets including Singapore, Tokyo, Seoul, and Sydney.
- ·Volume production timelines and pricing remain undisclosed, with qualification samples expected to reach hyperscale customers within the next quarter.
A New Memory Architecture for AI Infrastructure
Kioxia and SanDisk announced the launch of their latest QLC 3D flash memory technology, a product aimed squarely at data centers grappling with the storage and energy demands of AI workloads. The companies say the new memory delivers improvements in storage density, bandwidth, and power efficiency, according to Kioxia.
QLC, or quad-level cell, architecture stores four bits per cell, allowing manufacturers to pack more capacity into a single chip than older TLC or MLC designs. That density advantage has made QLC attractive for read-heavy applications, though historically the technology has lagged in write endurance and speed. The new generation aims to narrow that gap while maintaining the cost-per-gigabyte economics that have driven QLC adoption in hyperscale environments.
Why Data Centers Are Watching
The timing reflects broader pressures on cloud and AI infrastructure. Training large language models and serving inference requests generate massive read operations, and enterprises are deploying storage arrays that must balance capacity, throughput, and thermal output. Energy costs now represent a growing share of total cost of ownership for operators in Singapore, Tokyo, and Sydney, where power prices and cooling requirements are particularly acute.
Higher bandwidth in flash memory translates directly to faster data retrieval, a metric that matters when AI models pull training sets or vector databases during inference. Power efficiency, meanwhile, reduces both electricity bills and the thermal load on cooling systems, a consideration that has prompted several hyperscalers to re-evaluate their storage roadmaps over the past eighteen months.
Regional Implications for Asia-Pacific
Asia-Pacific accounts for a significant portion of global data center capacity, with major clusters in Singapore, Tokyo, Seoul, Mumbai, and Sydney. Operators in these markets face land constraints, strict energy regulations, and rising electricity costs. Any technology that improves storage density allows providers to defer or avoid physical expansion, while power-efficient memory can help facilities stay within regulatory caps or avoid costly grid upgrades.
Several Chinese and South Korean memory manufacturers are also developing competing QLC and PLC products, intensifying the race to supply the region's cloud giants. Kioxia, headquartered in Tokyo, and SanDisk, a Western Digital brand with manufacturing ties across Asia, are positioning the new flash as a near-term answer to this competitive and regulatory environment.
What Comes Next
The companies have not disclosed volume production timelines or pricing, though industry observers expect qualification samples to reach select hyperscale customers within the next quarter. Adoption will hinge on real-world endurance testing and integration with existing storage controllers, areas where QLC has historically required more validation than TLC.
For AI developers, the calculus is straightforward: faster, denser, and more power-efficient storage reduces infrastructure overhead and accelerates time to insight. For data center operators, it offers a path to meet capacity growth without proportional increases in energy consumption. Whether the new flash delivers on those promises will become clearer as deployment data emerges from early adopters across the region.
RELATED STORIES
Spot something wrong? Email editor@briefasia.com. We log every correction publicly.



