Revisiting CXL Progress – a Long-Awaited Next-Gen Memory Interconnect Architectural Shift w/ Inflection Ahead (2027+?)

Research evidence excerpt

Revisiting CXL Progress – a Long-Awaited Next-Gen Memory Interconnect Architectural Shift w/ Inflection Ahead (2027+?)

Equity Research

Industry Update — May 19, 2026

Semiconductors

Revisiting CXL Progress – a Long-Awaited

Next-Gen Memory Interconnect Architectural

Shift w/ Inflection Ahead (2027+?)

Our Call Aaron Rakers, CFA

We think CXL-based memory pooling could become an increasingly visible / viable Equity Analyst | Wells Fargo Securities, LLC

Aaron.Rakers@wellsfargo.com | 314-875-2508

architectural evolution. This report revisits CXL developments to assess whether more

Jake Wilhelm, CFA, CPAsignificant CXL adoption could finally become a reality (2027 inflection?).

Associate Equity Analyst | Wells Fargo Securities, LLC

Jake.Wilhelm@wellsfargo.com | 314-875-2502

Conclusion: With continued ecosystem growth and standards maturation (CXL 4.0 Michael Tsvetanov, CFA

released in Nov ‘25), the expansion of LLM context windows (KVCache = significant Associate Equity Analyst | Wells Fargo Securities, LLC

memory driver), and the overall backdrop of memory pricing + supply, we think Michael.Tsvetanov@wellsfargo.com | 314-875-2558

hyperscalers are actively evaluating the adoption of CXL-based DRAM pooling. We see Richard Strifler, CFA

MRVL best positioned w/in our coverage (w/ CXL controllers & switches; XConn acq.) and Associate Equity Analyst | Wells Fargo Securities, LLC

reaffirm our view that CXL adoption should not be viewed as bearish for memory long- Richard.Strifler@wellsfargo.com | 314-875-3234

term.

What Has Held CXL Back? CXL adoption has been slower than expected due to several

independent factors: 1. AI shifted the focus / investment away from trad'l CPU-centric

compute (this is changing w/ disaggregated inference). 2. DDR5 DIMM capacity has