Marvell Launches Six-Channel CXL 3.2 Memory Expanders for AI Inference
Event summary
- Marvell unveiled second-generation Structera X 3406 and 3506 CXL 3.2 memory expanders, offering six-channel DDR4/DDR5 support and PCIe Gen 6 connectivity.
- The new controllers support up to 9TB of DDR4 or 6TB of DDR5 per controller, a 50% increase in memory channels over the prior generation.
- Structera X 3406/3506 delivers up to 40–50% faster time to first token and 2–3x more effective memory capacity with inline LZ4 compression.
- The products are expected to sample in Q1 2027, extending the existing Structera X 2404 and 2504 architecture already shipping to hyperscalers.
The big picture
Marvell's new CXL 3.2 memory expanders address the growing memory demands of AI inference workloads, particularly for large language models and deep learning recommendation systems. The six-channel architecture and inline compression capabilities position the company as a key player in the evolving data center design landscape, where memory bottlenecks are a critical constraint. The collaboration with hyperscalers and memory vendors underscores the strategic importance of CXL technology in scaling AI infrastructure.
What we're watching
- Adoption Pace
- How quickly hyperscalers and cloud providers integrate the new Structera X controllers into their AI infrastructure.
- Competitive Response
- Whether competitors like Micron or Broadcom introduce comparable CXL 3.2 memory expansion solutions.
- Ecosystem Expansion
- The pace at which Marvell expands its CXL ecosystem partnerships to support broader AI workloads.
Related topics
