Thesis
Value accrues in layers. We invest where our experience is.
- 06Applications and vertical AI
- 05Agents, orchestration, workflow
- 04Data, retrieval, developer tooling
- 03Models and model tooling
- 02Infrastructure and inference
- 01Compute and hardware — watched, not underwritten
We follow a layered reading of the AI stack. We study the whole value chain, then concentrate in the software layers where we have built, shipped, and hired — layers two through six.
Silicon and hardware we watch rather than underwrite: longer cycles, heavier capital, richer early pricing.
The question we ask
What breaks if a foundation model ships this natively? If nothing does, we keep listening.
Where the early check goes
Five of the six layers are in scope for an early check. Each one is a different business with different economics, and we underwrite them differently.
Applications and vertical AI
Layer 06 of the stack.
Agents, orchestration, workflow
Layer 05 of the stack.
Data, retrieval, developer tooling
Layer 04 of the stack.
Models and model tooling
Layer 03 of the stack.
Infrastructure and inference
Layer 02 of the stack.