Specifications
Product
Samsung 16 Gb LPDDR6 SDRAM
Density
16 Gb per die (2 GB)
Per-Pin Rate
12.8 Gb/s
Generation
LPDDR6 (JEDEC)
Successor Sampling
LPDDR6X samples delivered to Qualcomm
Previous Generation
LPDDR5X, up to 10.7 Gb/s
Session
ISSCC 2026 — DRAM, SRAM and Non-Volatile Memories
Same-Cycle Disclosure
36 GB 12-Hi HBM4 at up to 3.3 TB/s per stack
Target Platforms
Flagship mobile, on-device AI, edge AI modules, AI PCs
Channel Architecture
LPDDR6 sub-channel architecture per JEDEC
Supplier Position
One of two suppliers in the first LPDDR6 wave (with SK hynix at 14.4 Gb/s)
Ecosystem
LPDDR-based module formats (SOCAMM2-class memory modules for AI servers)
Package Direction
Stacked die packages for higher module density
Why Two Speed Bins
Platforms trade the top LPDDR6 bin for power, yield and cost
Product Status
Technology paper — availability follows the JEDEC ramp
Availability
Roadmap allocation — quote for planning
Overview
Samsung's ISSCC 2026 LPDDR6 paper details a 16 Gb device running at 12.8 Gb/s per pin — the second of the two qualified first-wave LPDDR6 parts, alongside SK hynix's 14.4 Gb/s design. It was presented in the same session block as Samsung's 36 GB 12-Hi HBM4 paper, which is a useful reminder that the two memory tiers are now developed in parallel for the same AI-era workloads.
12.8 Gb/s versus 14.4 Gb/s is not a defect — it is a different point on the power/yield/cost trade. Mobile and edge platforms frequently specify below the top bin because sustained power matters more than peak transfer rate, and a part that holds its rate at lower voltage is worth more in a thermally constrained device than one that peaks higher. Having both parts qualified gives platform designers a genuine choice rather than a single source.
Samsung has also moved to the next step, sending LPDDR6X samples to Qualcomm — which is where the generation is heading. In the meantime LPDDR6 is the memory behind on-device AI features and the LPDDR-based module formats entering AI servers. QS Compute supplies LPDDR6-generation modules and the boards that use them.
Key Benefits
12.8 Gb/s per pin with the power behaviour that lets a mobile or edge platform sustain it. Second qualified LPDDR6 source removes single-supplier risk in the first generation wave. LPDDR6X already sampling to Qualcomm shows a clear forward roadmap, and the part shares the LPDDR6 sub-channel architecture used by LPDDR-based AI server modules.
Applications
Flagship smartphones and on-device AI inference; edge AI modules and AI PCs where sustained power beats peak bandwidth; and LPDDR-based server module formats for bandwidth-dense AI memory.
Request a Quote — SAMSUNG 16GB LPDDR6 AT 12.8 GB/S — ISSCC 2026 LOW-POWER DRAM
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →