- The Kioxia GP Series SSD gives GPUs faster memory access beyond HBM limits
- Storage Class Memory bridges the performance gap between DRAM and conventional NAND flash storage
- XL-FLASH prioritizes low latency and millions of random IOPS over sequential speed
Kioxia has introduced a new type of solid-state drive designed to act as a direct memory expansion for GPUs.
Announced at Nvidia GTC 2026, the new Kioxia GP series is not a replacement for existing storage, but rather an additional level in the memory hierarchy.
Its primary role is to provide a larger pool of quickly accessible data to GPUs, effectively acting as an overflow for the expensive and capacity-limited High Bandwidth Memory.
The article continues below
Memory-hungry AI models are driving the change
The drive utilizes Storage Class Memory (SCM), a technology category that lies in the performance gap between traditional NAND flash and system DRAM.
This concept was popularized years ago by Intel’s now-defunct Optane technology, which aimed to bridge the same gap but failed.
Kioxia’s version, branded as XL-FLASH, prioritizes low latency and high input/output operations per per second rather than raw sequential throughput, allowing finer data access of only 512 bytes.
This development is a direct response to a fundamental problem in the current AI infrastructure, the GPU memory is simply not large enough for the models it is asked to run.
As AI models scale against trillions of parameters and context windows expand to millions of tokens, the demand for memory to store objects like the KV cache has exceeded the physical limits of HBM.
Nvidia’s Storage-Next initiative, which this SSD supports, was created to address this very bottleneck by encouraging storage vendors to build drives that GPUs can link directly to.
“Kioxia fully supports the NVIDIA Storage-Next initiative and will provide purpose-built SSDs to effectively address the need for GPU-accessible memory,” said Makoto Hamada, Senior Director of SSD Division, Kioxia Corporation.
While the GP series aims for millions of IOPS to deliver data to GPUs, the wider industry is chasing even more ambitious performance targets.
The concept of reaching 100 million IOPS is a well-known industry dream that may require inventing entirely new classes of memory.
Other companies are also targeting the niche market left behind by Optane.
For example, the InnoGrit N3X SSD uses Kioxia’s XL-Flash in SLC mode to deliver extreme endurance.
This drive can reportedly withstand up to 50 full drive writes per day for five consecutive years.
Kioxia itself has previously signaled its intention to push for even higher performance numbers.
The company is targeting 10 million IOPS using SLC NAND, suggesting the GP series is just one step in a longer race to eliminate storage latency for processors.
It remains to be seen whether the Kioxia GP Series SSD will succeed where previous SCM attempts stumbled.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews and opinions in your feeds. Be sure to click the Follow button!
And of course you can too follow TechRadar on TikTok for news, reviews, video unboxings, and get regular updates from us on WhatsApp also.
