THE MACHINE ROOM
SECTION 04
ISSUE 001
Chiplet Marketplaces for AI
Projection: Qualcomm proposes disaggregated near-memory inference and rack systems; AMD combines accelerators, open racks, ROCm, and a long hardware roadmap. Those moves could support composable CPU, accelerator, memory, network, and security chiplets. A marketplace requires interoperable packaging and software that survives mixing vendors—neither is yet established.
Why this idea is here
What the evidence establishes.
Qualcomm announces agentic CPUs, near-memory high-bandwidth compute, disaggregated inference, liquid-cooled racks, and claimed improvements in bandwidth per watt; AMD documents the Helios rack-scale platform with Instinct MI450X GPUs, Venice CPUs, advanced networking, and ROCm, on track for second-half 2026 deployments. These are source-backed premises for this projection; they do not by themselves prove broad adoption or the eventual outcome.
Source ledger
Read the sources.
- S01Qualcomm Dragonfly Data Center Roadmap
official hardware announcement / published 2026-06-24 / retrieved 2026-07-09
- S02AMD Helios and Instinct MI450X move toward deployment
official infrastructure announcement / published 2026-05-21 / retrieved 2026-07-10