Dictionary entry Trending
CXL memory pooling
also CXL memory pool or Compute Express Link memory pooling
Definition of CXL memory pooling
-
: a server architecture that uses Compute Express Link to combine memory devices into a shared pool that can be allocated among hosts or accelerators, giving AI workloads access to more flexible capacity than the memory attached to one machine or GPU
- The inference cluster used CXL memory pooling to hold a KV cache that no longer fit in local DRAM.
- CXL memory pooling reduced stranded capacity by assigning extra memory to whichever model server needed it.
Two meanings, one word
In everyday English
Pooling means combining resources so they can be shared and allocated where needed.
In AI
A server architecture that uses Compute Express Link to combine memory devices into a shared pool that can be allocated among hosts or accelerators, giving AI workloads…
Origin & history
The Compute Express Link Consortium introduced memory pooling as a feature of the CXL 2.0 specification released in November 2020. The architecture became increasingly relevant to AI infrastructure as large models and long-context inference created demand for capacity beyond local HBM and DRAM.
Editorial references
Sources & further reading
- CXL Consortium Releases Compute Express Link 2.0 Specification↗ Standards Announcement · Compute Express Link Consortium · 2020-11-10
- Evaluating CXL Memory Pooling for Scalable LLM Inference↗ Peer Reviewed Research Paper · IEEE · 2025-12
- OCP Global Summit 2026↗ Official Technical Demonstration · Samsung Semiconductor · 2026-10-15
Test yourself
Which of these is the meaning of CXL memory pooling?
Cite this entry
"CXL memory pooling." AI Dictionary, Dadgogo, https://dadgogo.com/dictionary/cxl-memory-pooling/. Accessed 16 Oct. 2026.