Skip to content
Slow Variables

Ask the data

Answers come from the same store as the site; every number is checked against the record it cites.

The stack

Seven layers, twenty-one sub-layers, and the companies the tracker follows in each. Membership is many-to-many and dated; an entity is verified once a filing or a dataset row attaches to it. Only the compute layer has segment filings today, so most sub-layers are lists of names waiting for a series.

1. Compute & physical · Semis, hyperscalers and neoclouds, data centres and power, memory/networking.

2. Model · Frontier labs, neolabs, open-weight and Chinese labs.

3. Training input · Human data, RL environments, post-training platforms, synthetic data, licensing, evals.

4. Serving & orchestration · Inference serving, routing and gateways, agent infrastructure.

  • 14. Inference serving
    10 entities · 4 verified

    Anyscale, Baseten, DeepInfra, Fireworks, Hyperbolic, Modal, Novita, Replicate, …

  • 15. Routing
    9 entities · 2 verified

    Helicone, LiteLLM, Martian, Not Diamond, OpenRouter, Portkey, Requesty, TrueFoundry, …

  • 16. Agent infrastructure
    11 entities · 1 verified

    Arcade, Browserbase, Composio, CrewAI, Daytona, E2B, LangChain, Letta, …

5. Deployment & application · Deployment services, coding agents, vertical and consumer apps, rollups.

6. Adopters · Enterprises and individuals using the products; indicator-only layer.

Indicator-only layer: no companies, only measures of adopters, workers and consumers.

7. Labour & consumers · Wages, employment and consumer surplus — the welfare residual; indicator-only layer.

Indicator-only layer: no companies, only measures of adopters, workers and consumers.