Skip to main content
The Map page on the platform plots the physical GPU supply behind the market: individual “neocloud” datacenter sites, who operates them, what hardware they run, and where each site is in its buildout. It’s the geographic companion to the price indices, letting you see the capacity that price signals are moving against.
The full site payload — operator, power, hardware, timelines — is a Full plan feature. Index and Premium see the public points view (coordinates only). GET /api/neocloud/sites, /api/neocloud/map, and /api/neocloud/sites/aggregate require a sk_live_ key; a sk_prem_ key returns 403. See Access tiers and the API reference.

​
What each site shows

Every point on the map is a datacenter site. Fields below are Full-only unless noted. Index and Premium see coordinates on the public points view.

​
How to read it

  • Capacity, not price. The map answers “where is the compute, who runs it, and when does it come online” — it doesn’t carry rental prices. Pair it with The Price Index to connect supply to rates.
  • Known vs. estimated. GPU counts are split into a confirmed gpuQtyKnown and an gpuQtyEst estimate. Where only an estimate exists, treat the figure as directional.
  • Timelines are targets. Buildout dates (phase1LiveDate, fullCapacityTarget, and so on) are announced or projected milestones and can move.
Because the dataset is curated, coverage is broad but not exhaustive, and individual fields can be blank where a detail hasn’t been reported. Use sourceNotes for context on any specific row.