NVIDIA Rubin CPX
A context-processing accelerator: GDDR7 instead of HBM, built for prefill.
You're looking at a chip record: what this processor is, its key specs, who designs, manufactures and supplies its parts, and where it gets used.
Announced September 2025, Rubin CPX splits inference in two: a cheaper GDDR7-based chip handles the compute-heavy prefill over million-token contexts while HBM-rich Rubin GPUs handle bandwidth-heavy token generation. It is the clearest sign that inference hardware is specializing by phase. Targeted for end of 2026.
Specifications & key facts
Each figure is labeled by basis: vendor spec, vendor claim, reported, estimate, or official. "Not publicly disclosed" means exactly that — we never invent values.
Overview
Memory
Compute
How this connects to the rest of the stack
This record's links at a glance — click a group to jump to it.
Designed by
Competes with
Related
Sources
NVIDIA Rubin CPX announcement ↗
Record reviewed as of 2026-01-15 — see Methodology.