Most infrastructure questions fail on a boundary nobody names: which layer holds the answer, how fresh that layer is, and whether you can query it or only read about it. This page is the inventory — every DC Hub data layer, what it covers, how often it moves, and the three ways to reach it: a page a person can read, a REST endpoint, and an MCP tool an agent can call.
94 live MCP tools sit over these layers. Counts are floors, not estimates — they are the number DC Hub can stand behind today, and they are read from /api/v1/canon/phrases rather than typed into this page by hand.
Every facility DC Hub has discovered and de-duplicated, across 170+ countries: operator, address, status, market, and the tenants where they are known. This is the spine the other layers hang off.
A 0-100 power-availability score per market with a BUILD / CAUTION / AVOID verdict, an excess-power score and a constraint score. The numeric answer to “where can a data center actually get power”, recomputed every day rather than published once a quarter.
Real-time demand, generation mix and reserve margin from the ISOs and system operators themselves — US ISOs plus TVA and BPA, Great Britain, EU bidding zones, Taiwan, Japan, South Korea, Brazil, with partial feeds for Australia and Singapore. This is the layer that goes stale in minutes, which is why it cannot come from a document.
Where a project sits in its ISO's queue and how long that queue is actually taking — the number that decides whether a site is buildable on your timeline, and the one most often quoted from a research note that is two years old.
Mapped grid assets with their locations, so a candidate site can be measured against the nearest point of interconnection rather than described as “near transmission”.
Long-haul and metro fiber routes, subsea cables and their landing stations, and peering presence — plus what it takes to get a lead-in to one specific parcel.
Operating and planned generating units across every status — an inventory, not a plant count — alongside gas pipeline segments and delivered-gas economics for behind-the-meter and bridge generation.
Water stress, flood and wildfire exposure, seismic and storm risk, resolved at a coordinate rather than a county — the checks that disqualify a site late if nobody runs them early.
Sales- and property-tax abatements, enterprise-zone treatment and the permitting and moratorium picture, which increasingly decides a market before power does.
Data-center M&A, platform investments and JVs with their parties and, where disclosed, value — plus who owns and who occupies a given facility today.
Powered land, shell and turnkey capacity to buy or lease, searchable by size and location; operators and brokers list capacity free at /listings. An introduction sends the provider only a company name and a requirement; identities are exchanged only if they accept.
Four DC Hub pages get confused for each other. They do different jobs.
| If you are… | Go to |
|---|---|
| deciding whether any of this is useful to you | the homepage |
| asking what is actually in the dataset | this page |
| plugging DC Hub into Claude, ChatGPT, Cursor or another client | /connect |
| building an agent, or writing its system prompt | /ai-agents |
Three doors onto the same data. None of them needs a signup to start.
| Door | Where | Notes |
|---|---|---|
| MCP | https://dchub.cloud/mcp | 94 tools, Streamable HTTP. Keyless at preview depth; claim_free_key from inside the session raises it. Per-client setup at /connect. |
| REST | /api/v1/… | The free, no-auth endpoints are listed in /llms.txt, CORS open. |
| Pages | linked per layer above | Server-rendered and crawlable, for people and for retrieval crawlers that do not run tools. |
DC Hub-authored data (DCPI and our derived analysis) is CC-BY-4.0: cite it, redistribute it, build on it with attribution. Upstream sources keep their own terms (licence by layer). Attribution is the only ask for the former, and a citation should carry the timestamp the response came back with.