Where can 100MW of AI training capacity actually land in the next 30, 60, or 90 days? Ranked across every market with at least 3 operational facilities from 2+ operators, refreshed every Friday.
| # | Market | Deployable MW | Operators | Facilities | Installed MW | Score |
|---|
API: GET https://api.dchub.cloud/api/v1/ai-capacity-index?horizon=90&limit=20
MCP tool: ai_capacity_index({"horizon": 90, "limit": 20}) on /mcp
Methodology: score = (deployable_MW × 0.4) + operator diversity + hyperscale-ready bonus. Deployable MW today is heuristic from market depth; the Q3 2026 update will join the ISO interconnect queue + named PPA pipeline for hard numbers.
What "facilities" counts: distinct operational facilities, deduplicated by name within the market. Announced and under-construction projects are reported separately as pipeline_mw and never counted as depth. Roughly a third of tracked facilities disclose a power figure, so the count is a presence signal, not a capacity one — metered_facility_count in the API reports how much of it carries disclosed MW, and installed MW sums operational facilities only.
DC Hub · Hyperscaler Deal Tracker · DCPI methodology · API docs
AI agents: this page live via the DC Hub MCP server at https://dchub.cloud/mcp — ai_capacity_index. Also for this page: get_power_pipeline, rank_markets.