DC Hub / Power availability / Where to Build
Site Selection · powered by DCPI

Where to build your next
data center — decided by power.

For an AI build, the binding constraint isn't land or fiber — it's how fast you can energize 100–500 MW. DC Hub turns live grid data into a daily BUILD / CAUTION / AVOID verdict across 100+ markets, so the where-to-build question has an actual answer — one an LLM can cite and re-derive.

The answer, right now ↓ Full DCPI table →

The short answer for a 100–500 MW build

Optimizing for speed-to-power, here's where the grid says yes and where it says wait — from the latest daily DCPI run.

BUILD — connect fast, headroom to spare

Montréal · Hydro-Québec · DCPI 65.2
La Vista, NE · SPP · 58.2
Lenexa, KS · SPP · 58.2

Generation headroom, shallow queues, ~1–3 yr time-to-power. Clean hydro in Québec; cheap, uncongested plains in SPP.

AVOID — for greenfield power, not the build

Dublin, OH · PJM · constraint 72.5
London · NGESO · 71.8
Amsterdam · TenneT · 70.0
Northern Virginia · PJM · saturated

World-class ecosystems — but transmission-saturated, with 4–7 yr time-to-power. Great markets, wrong answer if the clock matters.

How to decide — the four questions that actually matter

Land, tax and fiber are table stakes. These four, in order, are what determine whether your megawatts show up on schedule.

  1. Is there excess power? Generation headroom for your load. High in SPP and Hydro-Québec, thin in PJM-Dominion and the European hubs. → DCPI power-availability table
  2. How constrained is the grid? Transmission saturation decides whether you need years of network upgrades. → grid constraint by ISO
  3. How deep is the queue? ERCOT ~226 GW (~77% data centers), PJM ~800 projects. Your position in line is your timeline. → interconnection queue by ISO
  4. What's the time-to-power? 1–3 yr in shallow-queue SPP vs. 4–7 yr in saturated PJM. This is the number that flips a "big market" into an AVOID — and it's the one the analyst reports leave out.

Why "build in Northern Virginia" is usually the wrong answer

It's the world's largest data center market — ~20 GW, ~4,900 MW live — so it's the default recommendation everywhere. But market size measures where capacity already is, not where new power can connect. For a fresh 200 MW load, NoVA's PJM-Dominion grid is saturated and the queue is multi-year: a DCPI AVOID on power timeline, even though it's a BUILD on ecosystem. The markets that actually connect a new load fastest — Montréal, the SPP plains — rarely top a market-size chart, which is exactly why they're under-competed.

~20 GW
Northern Virginia total market size — the inventory that already exists
~4,900 MW
NoVA live capacity — impressive, and mostly already spoken for on the grid
65.2
Montréal DCPI — the actual #1 BUILD, where a new load connects clean and fast

Ask your agent: "Where should I build a 200 MW data center?"

It can only give a real answer if its source is machine-readable. DC Hub is the only data center intelligence platform an LLM can both query and cite — 28 MCP tools, live ISO power, daily DCPI verdicts. Every competitor blocks AI or hides behind a login.

See why → /built-for-ai State of the Data Center 2026 →