Training data and labelling
The Stack Capacity Index measures how much an entity shapes the world's AI capacity within its layer of the stack. It is built mostly from measured, sourced facts, each scored on the board's own scale; stated estimates of AI exposure, position and momentum, scored on one written rubric, add to them but never outweigh a row's own facts. A figure nobody publishes is left out, never estimated.
AI exposure reads how much of the business serves AI compute demand today. It is a stated estimate.
Scores run 0 to 100. The top 10 companies · updated 21 Oct 2026.
| 01 |  Mercor | Expert data & RLHF | 86.1 | 95.0 |
| 02 | Scale AI | Expert data & RLHF | 75.2 | 95.0 |
| 03 | Surge AI | Expert data & RLHF | 74.3 | 100.0 |
| 04 | Handshake (AI) | Expert data & RLHF | 69.5 | 75.0 |
| 05 | Snorkel AI | Evals & RL environments | 54.5 | 95.0 |
| 06 | micro1 | Expert data & RLHF | 53.8 | 100.0 |
| 07 | Innodata | Crowd & annotation services | 42.2 | 85.0 |
| 08 | Invisible Technologies | Expert data & RLHF | 41.2 | 90.0 |
| 09 |  Appen | Crowd & annotation services | 35.7 | 90.0 |
| 10 | Labelbox | Labelling platforms | 32.9 | 95.0 |
Index scores are model outputs, not predictions. The facts behind each score are sourced, and the estimates are stated with their basis in the full dataset, where figures resting on a secondary source are marked.