JEDEC previews LPDDR6, proving that datacenters have stolen the mobile memory standard

JEDEC is shifting the focus of LPDDR6 from mobile platforms to datacenters – Thanks AI

Last year, JEDEC unveiled its JESD209-6 LPDDR6 memory standard, a new high-bandwidth memory type designed for phones, laptops, and other mobile platforms. Now, JEDEC has previewed new updates to the standard, shifting its focus from mobile platforms to “datacenter and accelerated computing workloads”. In other words, the mobile-focused LPDDR6 standard is being altered to better meet the needs of AI hyperscalers.

While this is not necessarily a bad thing, it showcases how much the memory market has shifted over the past year. LPDDR5X SOCAMM2 modules are now popular for AI datacenter use. As a result, the LPDDR6 standard is being updated to favour these products. Datacenter users are being given more options, enabling the creation of higher-capacity memory modules.

JEDEC plans to allow 512GB-density LPDDR6 SOCAMM2 modules to be produced. This is twice as large as today’s largest LPDDR5X SOCAMM2 modules.

The way I see it, JEDEC is formalising some of the customisations the datacenter industry wants. Options that enable higher-capacity memory modules and other forms of tweaking. These changes do not change LPDDR6’s overall capabilities, as its per-chip bandwidth and hardware-level features appear unchanged. The changes below to LPDDR6 appear to be adjustments to available configurations, giving datacenter users more flexibility. If these changes weren’t made now, they would be more awkward to add to the standard later.

LPDDR6 features

Planned features for the upcoming LPDDR6 update include:

• Narrower per-die interface (x6) enables higher capacities: With the move to a non-binary interface width – from x16 to x24, the inclusion of x12 and an additional x6 sub-channel mode, allows more die per package and higher memory capacities per component and per channel, a critical enabler for AI-scale memory footprints.
• Flexible metadata carve‑out intended to minimize impact to peak data throughput, giving data center customers the option to balance user capacity and metadata needs according to their specific reliability requirements.
• 512 GB density on the horizon: LPDDR6 is expected to unlock densities beyond the current LPDDR5/5X maximum, a capability designed to address the ever-growing memory capacity requirements of AI training and inference workloads.
• LPDDR6 SOCAMM2 module standard in development: JEDEC is actively working on an LPDDR6-based SOCAMM2 module standard, which is being designed to carry the compact, serviceable module form factor forward and offer a clear upgrade path from today’s LPDDR5X SOCAMM2 modules.

JEDEC LPDDR6 PIM memory is also in development

JEDEC has also confirmed that an LPDDR6 Processing-in-Memory (PIM) standard is in development. This standard should help increase the performance and energy efficiency of LPDDR6 modules. Adding processing capabilities to the modules themselves  reduces the number of transfers memory needs to make. This enables more efficient bandwidth and higher module efficiency.

LPDDR6 PIM is seen as an important development for AI inferencing, which is limited by memory bandwidth. An official LPDDR6 SOCAMM2 standard is also in the works.

LPDDR memory’s use in datacenters is growing. AMD has confirmed that its EPYC Verano CPUs will support SOCAMM2 memory modules in 2027. Nvidia already supports SOCAMM and SOCAMM2 memory modules with its products.

At CES 2026, Samsung showcased 10.7 Gbps LPDDR6 memory modules. Currently, the LPDDR6 standard has not been finalised. If no major changes are required soon, the finalised standard may launch later this year.

You can join the discussion on JEDEC’s LPDDR6 memory plans on the OC3D Forums.