About Inference Stack
Inference Stack is a structured map of AI inference infrastructure and the data-center product portfolios behind it — NVIDIA, AMD, Google, Amazon, and Intel — built for engineers deciding what to run on and agents querying it directly. It's a dataset with a browsable UI on top, not the other way around: an interactive stack browser tracing app archetype through framework, orchestration, silicon, and cloud instance; a multi-vendor product catalog; and reference architectures grounded in real vendor blueprints and production case studies.
Data is maintained by hand, not scraped continuously — this is a fast-moving market. The stack browser, product catalog, and reference architectures were all last refreshed 2026-08. Every entry carries its own source, and roadmap-stage hardware (not yet shipping) is explicitly flagged as such rather than mixed in with shipping products. Spot-check anything you rely on against the linked source before committing spend or design decisions to it.
Every page here is backed by a plain JSON API, and /llms.txt documents it for LLM consumers. Treat id fields as stable keys — they're meant to be joined across responses and linked to directly.