According to Korean technology media The Bell, Samsung Electronics has delivered LPDDR6X low-power memory prototype samples to Qualcomm. This move has attracted industry attention because related products may be applied to the AI250 inference accelerator server platform Qualcomm plans to launch in 2027.
Last October, Qualcomm, a San Diego-based chip design company, released its AI200 and AI250 chip accelerator cards and rack products. It is worth noting that the company mentioned in its promotional materials that it will use low-power DRAM memory technology, rather than the high-bandwidth memory technology that has become the mainstream choice on cutting-edge artificial intelligence enterprise platforms. Low-power memory is usually used in portable devices equipped with Qualcomm chipsets, but now Qualcomm seems to be exploring the possibility of deploying LPDDR6X modules into server-class devices.
The Bell quoted anonymous industry insiders as saying that the timing of Samsung's move was slightly unusual. Because the LPDDR6 non-X version standard has just received official JEDEC certification in the third quarter of 2025, the LPDDR6X technology, which is still under development, has begun to provide samples to partners. The source believes that Qualcomm’s compact AI200 series product release roadmap has prompted Samsung to supply LPDDR6X modules in advance.
There were rumors last month that Qualcomm's AI200 and AI250 inference accelerators were considering using SOCAMM2 memory solutions. In December last year, Samsung's memory product planning team revealed the future development branches of non-X technology, specifically mentioning that "low-power DRAM technologies such as SOCAMM" are being derived from "AI-specific module solutions based on LPDDR6 architecture."
This cooperation trend shows that in the field of artificial intelligence servers, the memory technology route is showing a diversified trend. Although high-bandwidth memory dominates the field of AI training and inference, low-power memory may find its own market position in specific application scenarios due to its power consumption advantages.