All projects
ProjectData-vizReact 19MapLibreCloudflare

Topologies

An interactive atlas of global infrastructure across 21 lenses, from submarine cables to chip fabs to AI data centers, with guided stories, country pages, path tracing and what-if simulation.

Topologies cover
Explore it

Topologies is an interactive atlas of the hidden infrastructure that runs the modern world. Pick one of 21 lenses, such as submarine cables, internet exchanges, BGP routing, AI data centers, chip fabs, power plants, critical minerals, shipping lanes or satellites, and see it drawn on a world map or as a force-directed network. Click anything for its data sheet, trace a path between two points, and follow how one layer depends on the next.

Explore it live → worldlenses.janneparkkila.com

The Submarine Cables lens: a world map of landing stations and cable routes, with the lens list on the left. (opens full size)
The Submarine Cables lens: a world map of landing stations and cable routes, with the lens list on the left.

Why I built it

What looks like a borderless global internet is really a stack of small-N systems: 13 logical DNS roots, a handful of chip fabs, a short list of atomic clocks that define what time it is. Switching lenses makes that concentration visible, and shows how a failure, sanction or conflict at one layer ripples through the rest. I wanted a single tool that makes those dependencies tangible for students, researchers and the merely curious.

Start here

You don't need to know what to look for. Five guided stories walk across the lenses in a few steps each:

The story “If Taiwan goes dark” at step 2 of 4. TSMC's Hsinchu fab is removed, and the scenario panel reports 3 nodes disconnected and 30 affected across lenses. (opens full size)
The story “If Taiwan goes dark” at step 2 of 4. TSMC's Hsinchu fab is removed, and the scenario panel reports 3 nodes disconnected and 30 affected across lenses.

Or start from a place. Every country has a page showing what the atlas records there and what it connects to, like Finland or the Netherlands.

The Finland page: 10 sites across 7 of 21 lenses, most connected to Germany, Sweden and the United States. (opens full size)
The Finland page: 10 sites across 7 of 21 lenses, most connected to Germany, Sweden and the United States.

What it does

  • Two views, one language. Every lens renders on a MapLibre world map and as a d3-force network.
  • Path tracing. Find the best route between two nodes with Dijkstra, plus alternatives from Yen's k-shortest paths. Each lens offers the metrics its data supports, such as hops, latency, bandwidth, reliability, distance or throughput.
  • Cross-lens dependencies. Click a datacenter and see the power plants, exchanges and cables around it, then follow the chain.
  • What-if simulation. Remove a node and see what disconnects in that lens and what cascades across the others.
  • Time, filters, stats, heatmaps, compare. Scrub a lens through the years, filter by any field, see distributions, shade the map by a metric, or put two lenses side by side.
  • Works on a phone. The map takes the whole screen, and the panels become bottom sheets.
  • Dark and light. A brass-on-graphite “cartographic instrument” theme, with a daylight variant.
A traced path from Marseille to Singapore in brass across the Submarine Cables lens, with the route's hops listed on the right. (opens full size)
A traced path from Marseille to Singapore in brass across the Submarine Cables lens, with the route's hops listed on the right.
The what-if simulation on ASML in Veldhoven in the Chip Production lens, with the scenario panel listing affected nodes across lenses. (opens full size)
The what-if simulation on ASML in Veldhoven in the Chip Production lens, with the scenario panel listing affected nodes across lenses.
The Chip Production lens as a force-directed network, coloured by supply chain role. (opens full size)
The Chip Production lens as a force-directed network, coloured by supply chain role.
Compare mode with Energy Production and AI Build-Out side by side, both showing North America. (opens full size)
Compare mode with Energy Production and AI Build-Out side by side, both showing North America.
The Energy Production lens in the light theme. (opens full size)
The Energy Production lens in the light theme.
Three phone screens: the AI Build-Out map, the details sheet for Colossus 2, and the Finland country page. (opens full size)
Three phone screens: the AI Build-Out map, the details sheet for Colossus 2, and the Finland country page.

The AI Build-Out

The newest lens tracks 96 of the largest AI data centers, using estimates from Epoch AI built from satellite imagery, permits and public documents. 76 have power online and 20 don't yet. The data sheets show IT power, the total with cooling and other overhead, estimated compute and capital cost, and where each site's timeline is heading.

The AI Build-Out lens centered on North America, with circles sized by IT power clustered across the United States. (opens full size)
The AI Build-Out lens centered on North America, with circles sized by IT power clustered across the United States.

Honest data

Every node and figure carries a confidence level: verified (public records or operator data), estimated (derived or interpolated) or mocked (a structural placeholder, never a fake statistic). Of the 3,811 live nodes across the lenses, 3,267 are verified and 544 are estimated.

Data comes from public sources, including PeeringDB, TOP500, SatNOGS, OurAirports, OpenFlights, Epoch AI, the WRI Global Power Plant Database, USGS, the IEA, root-servers.org and TeleGeography's public cable map. Each stays under its own terms, and all are credited on the in-app Licenses & Disclaimer page.

How it stays current

A GitHub Action re-runs the importers every month, validates the result, and opens a pull request. Nothing goes live until I've merged it, and a run that would remove more than a quarter of a source's records fails instead of proposing it. The app's What Changed feed is generated by diffing the data, so every line reflects a real change.

The What Changed feed, listing a new Google site in AI Build-Out and the larger changes from the previous refresh. (opens full size)
The What Changed feed, listing a new Google site in AI Build-Out and the larger changes from the previous refresh.

Not everything refreshes. Supercomputers, internet exchanges, data centers, AI sites and ground stations come from live sources. Lenses like DNS roots and chip bottlenecks are curated by hand, and change when I edit them.

Read about what the first refreshes turned up →

Built with

React 19 · TypeScript · MapLibre GL · d3-force · Vite · Bun. Static JSON with prerendered pages for every lens, story and country, deployed to Cloudflare Pages through GitHub Actions, with cookie-free analytics from Cloudflare.

Curious about the story behind it? Read the original write-up →

Design and build: Janne Parkkila. Free to use in teaching, demos and research; attribution appreciated. Map screenshots: basemap © CARTO, map data © OpenStreetMap contributors.