Download the data

838 sites in 63 countries. Download a spreadsheet, map file or database.

Updated 2026-10-27. Status observations: 2024-08-06 to 2026-10-27. Each project keeps its last reported status and date.

Excel workbook Sites CSV GeoJSON SQLite

Browse all projects and multi-site programs. Project details preserve investment, hardware, capacity, organizations and timelines with their sources. The workbook includes separate sheets for these figures and events.

Files #

All files (zip, 17.3 MB)

GitHub Pages serves .ndjson, .sqlite and SHA256SUMS as application/octet-stream and .zip as application/x-zip-compressed; identify formats by file extension or by media_type here, not by Content-Type.

Pinned URLs #

The current downloads update as records are added or corrected. For reproducible analysis, use the archived source files from 2026-10-27. Their bytes are fixed: the generated formats are frozen GitHub Release files. Check files against the SHA256SUMS listed there.

Use the data #

curl #

curl -LO https://kiankyars.github.io/datacenter-atlas/downloads/latest/sites.csv
curl -LO https://kiankyars.github.io/datacenter-atlas/downloads/latest/datacenter-atlas-latest.zip
curl -LO https://kiankyars.github.io/datacenter-atlas/downloads/latest/SHA256SUMS
sha256sum --check --ignore-missing SHA256SUMS

Python (pandas) #

import pandas as pd

base = "https://kiankyars.github.io/datacenter-atlas/downloads/latest/"
# keep_default_na=False keeps codes such as "NA" (Namibia); only empty cells are missing.
options = {"keep_default_na": False, "na_values": [""]}
sites = pd.read_csv(base + "sites.csv", **options)
projects = pd.read_csv(base + "projects.csv", **options)
evidence = pd.read_csv(base + "evidence.csv", **options)
print(projects.groupby("last_observed_physical_status").size())

Python (standard library) #

import csv
import io
import json
import urllib.request

with urllib.request.urlopen("https://kiankyars.github.io/datacenter-atlas/downloads/latest/projects.csv") as response:
    projects = list(csv.DictReader(io.TextIOWrapper(response, encoding="utf-8", newline="")))
with urllib.request.urlopen("https://kiankyars.github.io/datacenter-atlas/api/v1/sites.json") as response:
    sites = json.load(response)
print(len(projects), "projects at", len(sites), "sites")

Python (GeoPandas) #

import geopandas as gpd

sites = gpd.read_file("https://kiankyars.github.io/datacenter-atlas/api/v1/sites.geojson")
print(sites[["name", "country", "geometry_type", "url"]].head())

R (readr and sf) #

library(readr)
library(sf)

base <- "https://kiankyars.github.io/datacenter-atlas/downloads/latest/"
# na = "" keeps codes such as "NA" (Namibia); only empty cells are missing.
sites <- read_csv(paste0(base, "sites.csv"), na = "")
projects <- read_csv(paste0(base, "projects.csv"), na = "")
evidence <- read_csv(paste0(base, "evidence.csv"), na = "")
geometry <- st_read("https://kiankyars.github.io/datacenter-atlas/api/v1/sites.geojson")

DuckDB #

SELECT * FROM read_csv_auto('https://kiankyars.github.io/datacenter-atlas/downloads/latest/sites.csv');

SELECT last_observed_physical_status, count(*) AS projects
FROM read_csv_auto('https://kiankyars.github.io/datacenter-atlas/downloads/latest/projects.csv')
GROUP BY 1 ORDER BY 2 DESC;

SQLite command line #

curl -LO https://kiankyars.github.io/datacenter-atlas/downloads/latest/atlas.sqlite
sqlite3 atlas.sqlite "SELECT name, country, newest_status_as_of FROM site_overview ORDER BY newest_status_as_of DESC LIMIT 10;"

JavaScript (fetch) #

const response = await fetch("https://kiankyars.github.io/datacenter-atlas/api/v1/sites.json");
const sites = await response.json();
console.log(sites.length, sites[0].name, sites[0].statuses);

Excel and Google Sheets #

Excel: open https://kiankyars.github.io/datacenter-atlas/downloads/latest/atlas.xlsx (one sheet per table, plus a README sheet).
Excel with a CSV: Data > From Text/CSV, File origin 65001: Unicode (UTF-8).
Google Sheets: =IMPORTDATA("https://kiankyars.github.io/datacenter-atlas/downloads/latest/sites.csv")

QGIS and Google Earth #

In QGIS choose Layer → Add Layer → Add Vector Layer and paste https://kiankyars.github.io/datacenter-atlas/api/v1/sites.geojson (its features carry each site's page URL, API URL and geometry licence). In Google Earth open sites.kml. To browse the SQLite database in the browser, open it in Datasette Lite.

Dictionary and API #

• Field dictionary: every column, type and coded value.
• JSON API: per-site documents, GeoJSON, aggregates, an OpenAPI description and JSON Schemas.
• Data history: additions, corrections and archived files.
• llms.txt: a plain-text guide for language models and agents.

Licence and citation #

Creative Commons Attribution 4.0 International for Atlas-authored content. Geometries derived from OpenStreetMap are © OpenStreetMap contributors and available under the Open Data Commons Open Database License 1.0. Other third-party facts, titles, URLs and geometries keep their source terms, recorded per evidence row (each site record lists its geometry_licenses) and in the attribution file. Details on the about page.

Kyars, K. (2026). Data Center Atlas (2026-10-27) [Data set]. Source files

Copy citation

@misc{kyars2026datacenteratlas,
  author = "Kyars, Kian",
  title  = "Data Center Atlas",
  year   = "2026",
  note   = "Data as published 2026-10-27",
  url    = "https://kiankyars.github.io/datacenter-atlas/downloads/2026-10-27-research-hundredsecond/"
}

Also as BibTeX file, RIS (EndNote, Zotero, Mendeley) and JSON with CSL-JSON and biblatex.

Data notices #