Samsung announces microSD cards for products with on-device AI capabilities

With our two new microSD cards, Samsung said, we has provided effective solutions to address the growing demands of mobile computing and on-device AI

Samsung has announced that it has started sampling 256GB SD Express microSD cards. Additionally, the company said, it has started the mass production of its 1TB UHS-1 microSD card. Samsung said that the Express microSD card will launch later this year, and the 1TB UHS-1 cards will be available within the third quarter of 2024.

“With our two new microSD cards, Samsung has provided effective solutions to address the growing demands of mobile computing and on-device AI,” said Hangu Sohn, Vice President of the Memory Brand Product Biz Team at Samsung Electronics.

Samsung said that the 256GB SD Express microSD card boasts read speed of up to 800 MB per second and write speeds of up to 958 MB per second. The read speed indicates the time it takes to open a file that is stored on a card, while the write speed denotes the time taken to save a file on the card.

For comparison, Samsung said, its SD Express microSD card is 1.4 times faster than SATA SSDs (Solid State Drive) that reaches the sequential read speed up to 560 MB per second. The upcoming SD Express cards will be four times faster compared to traditional UHS-1 memory cards, which reach read speeds up to 200 MB per second. For performance and reliability, Samsung said that it has equipped its 256GB SD Express microSD card with Dynamic Thermal Guard (DTG) technology that helps the card maintain optimum temperature even during long usage sessions.

With its next-generation microSD card lineup, Samsung aims to offer storage solutions for products capable of running on-device AI. As AI features increasingly operate without reliance on the cloud, storage size and speed will become crucial specifications. Devices equipped with faster and larger storage capacity will be more capable of executing data-intensive AI tasks, like running AI image generation models directly on the device.