A single data center is impressive on its own, but it is really one node in a planet-spanning machine. Zoom out and the building described in Parts 1 through 3 becomes a small piece of the system that runs the internet and, increasingly, artificial intelligence.
The Global Network
Data centers are valuable only because they are connected. Most are linked by fiber-optic cable, and the long-distance connections between continents run through submarine cables, fiber lines laid across the ocean floor. The numbers are hard to picture: a few hundred of these cables, some thousands of miles long, carry almost all intercontinental internet traffic, and the volume crossing them climbs every year as video and AI traffic grow. When you video call someone overseas, your words travel through one of them. The large technology companies now help fund many of these cables themselves to keep their data centers tied together.
Regions, Zones, and the Cloud
The biggest operators, the cloud providers, organize their data centers into regions, geographic clusters of facilities, each containing several separate sites so that if one fails the others carry on. Within a region, providers run multiple independent sites, sometimes called availability zones, with their own power and networking, so that a fire, flood, or local outage at one does not take down the rest. A company renting cloud computing chooses which regions to run in, balancing cost, speed, and rules about where data is allowed to live. It is why an app can feel fast in one country and slower in another: it depends on how close the nearest region is.
Core and Edge
There is a useful split between two kinds of location. Large core data centers sit where land and power are available, often far from cities, and do the heavy lifting. Smaller edge data centers sit close to population centers and handle work that has to feel instant, like streaming or quick AI responses. The pattern is to do the big, heavy work in a few enormous facilities and push the time-sensitive work outward to many small ones, and as AI spreads into everyday products, more of that lighter work is moving to the edge, closer to where people actually are.
The AI Buildout
For most of the internet's history this system grew steadily. Artificial intelligence broke that rhythm. Training and running AI models demands far more computing, and far more power, than anything before it, which has set off the largest construction boom the industry has ever seen. The money involved is staggering: the largest operators are together spending hundreds of billions of dollars a year on these facilities, a level of capital investment more often associated with national infrastructure than with the technology industry. Global data center capacity is projected to roughly double by the end of the decade, and the biggest operators have stopped designing one building at a time, instead planning entire campuses, some approaching a gigawatt of power, as single integrated machines for AI.
Who Builds and Owns It
The system is built by several kinds of players. The cloud giants build enormous facilities for their own services. Specialized operators run colocation facilities for rent. And independent developers build new sites, often from the ground up, to lease to the companies that need capacity faster than they can build it themselves. The result is an ecosystem built on specialization: companies that need computing capacity can lease or buy it from operators built specifically to design, power, and run these facilities well.
Growing at This Scale
Growth of this pace and size runs into real limits, and each one comes with a way developers are already addressing it. Power is the first, covered in Part 3: developers fund the transmission and substation upgrades their own projects require, which is how the grid expands alongside the demand rather than falling behind it. Water is the second, since cooling systems can use significant volumes of it; operators increasingly design around reclaimed or non-potable water sources and closed-loop systems that recirculate rather than consume, cutting the net draw on local supply. The third is the relationship with the surrounding area itself, since a data center is a sizable industrial neighbor. Developers typically bring meaningful investment with the project, expanded tax revenue, utility upgrades that benefit the wider grid, and local jobs during construction and operation, so the arrangement runs in both directions rather than one.
What It Comes Down To
Step back and the picture is of a single global system: thousands of buildings, joined by cable over land and under sea, organized into regions and pushed out to the edge, now expanding at a furious pace to feed AI. Part 5 looks at where that system is heading next.
This is the fourth article in Data Centers 101, a Blueprint Data Centers series explaining how modern data centers work, from the hardware inside them to the power that runs them. Blueprint is an independent data center platform developing greenfield data centers designed with flexibility to support a range of use cases including high-performance computing, AI and other advanced workloads. Follow Blueprint for more on the infrastructure behind the digital economy.
