Kioxia has nothing to lose, so it wants to swap expensive DDR for the world's fastest SSD ever built to save the world from AI RAMpocalypse
PCIe 6.0 SSDs and CXL flash could ease AI's memory crunch
- Kioxia is pushing flash memory into DRAM territory
- AI workloads are forcing data centers to rethink memory expansion methods
- Kioxia XL1 could add capacity without requiring equivalent increases in DRAM
Kioxia is developing flash memory products designed to help data centers reduce their dependence on costly DRAM as AI workloads demand increasing amounts of memory.
The Japanese memory maker plans to combine faster NAND technology with new interfaces built for demanding AI systems that require large volumes of accessible data.
Its approach includes PCIe 6.0 SSDs and CXL memory expansion technology, though the company has not claimed these products will fully replace DRAM.
Kioxia's SSD lineup targets high-performance AI infrastructure
Kioxia is now showcasing its GP1 Series PCIe 6.0 NVMe SSDs, engineered for extremely high input and output operations within AI infrastructure environments.
These drives support direct communication between GPUs and storage, potentially reducing data movement through conventional system pathways during intensive computational workloads.
PCIe 6.0 could draw particular attention because it offers substantially greater interface bandwidth than earlier generations used in many data center SSDs.
The company is also exhibiting its CM10 Series enterprise SSDs and NX1 Series data center drives, which feature liquid cooling support.
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
The CM10 Series relies on Kioxia's tenth-generation BiCS FLASH technology, while the NX1 Series uses an E1.S design with PCIe 5.0 connectivity.
These products specifically target data center environments where strong performance gains and improved thermal management increasingly shape storage architecture decisions today.
Kioxia believes shifting certain workloads onto flash storage could ease the capacity shortages and rising costs currently straining global memory supply chains.
The XL1 series aims to bridge DRAM and storage
According to Nikkei Asia, Kioxia has plans to use fast NAND as a direct alternative to conventional DRAM.
Kioxia's most notable development is the XL1 Series, a CXL-compatible memory expansion module built on its XL FLASH technology for lower latency.
The module intends to use a CXL interface with memory-access speed carried across a PCIe link.
CXL allows memory devices to connect more closely with processors, creating additional capacity without requiring every byte to remain in DRAM.
Kioxia says that combining high-speed NAND with a CXL controller module reduces data-read latency to less than one-tenth that of conventional NAND products.
The company positions XL1 to occupy the performance and cost space between conventional DRAM and ordinary SSD storage devices.
Kioxia states that replacing part of a system's DRAM with CXL modules can double memory capacity while improving performance by 30%.
That distinction matters because NAND flash remains inherently slower than DRAM, meaning wholesale replacement would create performance penalties for latency-sensitive applications.
Evaluation units are scheduled for delivery to technology partners during August 2026, giving potential customers a chance to assess the technology.
Kioxia plans to discuss these technologies at FMS 2026 in Santa Clara, California, through presentations, technical sessions and live demonstrations.
Its overall strategy focuses on supplementing existing DRAM rather than eliminating it from AI server memory configurations.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.
Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master's and a PhD in sciences, which provided him with a solid foundation in analytical thinking.
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.