On July 29, Kioxia Corporation (Kioxia) announced the shipment of commercial samples of UFS 5.0 embedded flash memory, just five months after delivering functional evaluation and interoperability test samples in February this year.
This progress indicates that the productization process of this standard has entered the final validation stage prior to mass production.
Samsung Electronics officially released its UFS 5.0 solution in June and plans to commence mass production in the fourth quarter. Earlier, in August 2025, Micron Technology (Micron) announced the cessation of its mobile NAND flash memory R&D, shifting its focus to enterprise-grade AI storage. As the new-generation embedded storage standard released by JEDEC in February 2026, UFS 5.0 is accelerating its transition from technical specifications to actual commercialization. The different choices made by the three major manufacturers are also outlining a new landscape in the mobile storage market.
01. Standard Upgrade: Doubled Performance, Optimized Energy Efficiency, and Enhanced Reliability
If the UFS 4.0 era solved the problem of being "fast enough," then UFS 5.0 aims to answer the question of "how to maintain speed while accommodating all scenarios."
The advent of edge AI has transformed storage from a passive data repository into an active participant that must respond in real time to model loading, high-frequency fragmented read/write operations, and continuous high-volume data transmission.
Driven by this shift in demand, the formulation of the UFS 5.0 standard did not bet everything on sequential speed. Instead, it advanced synchronously across multiple dimensions, including bandwidth, random response, power consumption control, physical footprint, and data security. The goal is to enhance peak performance without compromising stability and battery life in actual usage.
The sequential read speed of UFS 5.0 reaches up to 10.8GB/s, and the sequential write speed reaches up to 9.5GB/s, more than doubling that of UFS 4.1 (read speed of approximately 4.2GB/s, write speed of approximately 2.8GB/s). Random read performance is improved by approximately 5 times, and random write performance by approximately 2 times. The latter is even more directly significant for edge AI, as the parameter loading of local large models and the real-time inference of multimodal assistants essentially consist of frequent invocations of massive small data blocks, rather than mere sequential transfer of large files.
In terms of the underlying architecture, the physical layer adopts MIPI M-PHY v6.0, supporting HS-Gear6 mode and PAM4 signal encoding, with a single-lane rate of 46.6Gbps and a theoretical dual-lane bandwidth of up to 10.8GB/s. The link layer is upgraded to UniPro v3.0 to reduce transmission latency. Regarding energy efficiency optimization, technologies such as clock gating, multi-voltage power supply, and independent power rails (VCCQ_M) are utilized. Under the same read performance, energy efficiency is improved by approximately 40% compared to UFS 4.1. For battery-sensitive devices like smartphones and XR headsets, this means that heat generation and battery life pressure under high-load scenarios are expected to be alleviated to some extent.
The package size is reduced to7.5mm×13mm×0.9mm, shrinking the volume by approximately 16.7%. With a maximum capacity supporting 1TB, it provides more flexibility for internal device layout. Inside flagship smartphones, the battle for every square millimeter of space directly relates to the trade-offs between battery capacity and camera modules.
In terms of security and reliability, Inline Hashing technology is added for real-time data integrity verification, and Link Equalization technology compensates for signal attenuation during high-speed transmission, reducing the risk of data packet loss.
02. Manufacturer Dynamics: Samsung and Kioxia Roll Out Solutions, Micron Shifts Strategy
Within half a year after the standard's release, the implementation pace of leading manufacturers has gradually become clear. On June 23, 2026, Samsung Electronics officially launched its UFS 5.0 storage solution based on 9th-generation V-NAND technology. The sequential read and write speeds reach 10.8GB/s and 9.5GB/s, respectively, with improvements in random performance and energy efficiency over the previous generation. The package size is 7.5mm×13mm×0.9mm, with a maximum capacity of 1TB. Mass production is scheduled for the fourth quarter of 2026. The initial applications target flagship smartphones, with plans to expand to XR headsets and AI wearable devices.
Kioxia, on the other hand, introduced commercial samples on July 29. Its solution adopts a self-developed controller and 8th-generation BiCS Flash 3D flash memory, with sequential read speeds up to 10GB/s and sequential write speeds up to 9GB/s. It offers 512GB and 1TB capacity options, and the package size is also 7.5mm×13mm. It specifically emphasizes an enhanced thermal management design to ensure stability during high-speed read/write operations. Kioxia expects to commence mass production by the end of 2026, covering diverse edge AI devices such as AI smartphones, AI tablets, XR devices, and robots.
The two manufacturers above have different focuses in their specifications: Samsung has a slight lead in peak speed, while Kioxia places more emphasis on thermal management and practical application adaptation.
Micron, meanwhile, announced in August 2025 the complete cessation of R&D for mobile NAND flash memory (including UFS 5.0), shifting its strategic focus to enterprise-grade AI storage (such as HBM and enterprise SSDs). Therefore, market competition in the UFS 5.0 era will primarily unfold between Samsung and Kioxia.
03. Mass Production Plans and Application Prospects
According to public information, the mass production timelines for both Samsung and Kioxia are concentrated in the fourth quarter and the end of 2026. By then, UFS 5.0 is expected to be gradually integrated into terminals such as flagship smartphones, XR headsets, AI wearable devices, as well as certain robots and smart vehicles.
The industry believes that the demand for storage bandwidth and random response capabilities from edge AI applications (such as local large model loading and real-time multimodal interaction) is one of the main driving forces behind the implementation of UFS 5.0.
However, the actual commercialization pace still depends on the integration progress of terminal manufacturers, cost control, and overall system adaptation. For instance, collaborative optimization at the levels of main control chips, interface protocols, and operating systems takes time. Market penetration is expected to gradually expand only after 2027.
Overall, UFS 5.0 represents the comprehensive evolution direction of mobile storage in terms of performance, power consumption, and space utilization. Its market performance remains to be seen based on subsequent mass production deliveries and actual feedback from terminal products. However, it is certain that the compelling demand of edge AI for storage performance is accelerating the popularization process of this technology.
#UFS #Kioxia #Micron #Samsung