AI Infrastructure Software
GPU runtimes, training frameworks, orchestration, and inference serving.
Stack position
Where AI Infrastructure Software sits in the AI stack
L10 · GPU Cloud
Accelerator rental: neoclouds, sovereign AI compute, and enterprise private GPU clusters.
L11 · AI Infrastructure Software
GPU runtimes, training frameworks, orchestration, and inference serving.
L12 · Foundation Model
Frontier and open models across text, vision, audio, and reasoning; revenue is token-metered.
L13 · AI Data
Pipelines, embeddings, vector stores, retrieval, labeling, and synthetic data.
Icons mark layer roles — not company logos. Position only; not a price chart.
Why this layer matters
The layer that turns silicon into something a developer can use: GPU runtimes, training frameworks, distributed execution, containers, orchestration, and inference serving. Hardware advantage is defended here as much as in the fab.
What is the bottleneck?
Switching cost is the moat. A faster accelerator that the existing toolchain does not target is not immediately usable.
A structural observation about supplier concentration and lead times — not a view on any security.
Who sits in this layer?
No public companies are listed here because the pure plays in this layer are still private. Public exposure to it runs through the diversified companies that fund and distribute them, which this map already places in their own layers.
Coverage boundary
Coverage boundary: US-listed tickers or ADRs only (same rule as TSM / ASML). Important suppliers without a US listing are omitted even when they matter to the stack. That is a coverage rule, not a quality ranking.
No hand-listed notable omissions for this layer yet. The US-list / ADR rule still applies.
Checked 2026-10-09. Not exhaustive.
Same track
Descriptive positioning only. The positioning label says whether a company's business is exposed to a layer, never whether to buy it. Layers with no exposure are not recommendations to add any.