Brandon Donnelly — Daily insights for city builders. Published since 2013 by Toronto-based real estate developer Brandon Donnelly. — Page 123

As you know , Northern Virginia is now referred to as "data center alley." It has, by far, the largest agglomeration of data centers in the world. The latest figures are somewhere around 200 completed facilities and some 49 million square feet, with a lot more in the pipeline.

Here's the global top 10 list via Bloomberg :

And here's a map of existing (blue) and proposed (purple) data centers via Loudoun County, Virginia :

This has been an economic boon for Virginia. It's estimated that the data center industry contributes up to 74,000 jobs and $9.1 billion in GDP to the state each year. But along with these benefits come some trade-offs, one of which has to do with the region's built environment.

Here are two zoom-ins of an area to the west of Dulles International Airport:

These maps raise a question that is only going to become more important as time goes on: What's the best way to insert large insular boxes into the fabric of a city or suburb? Of course, in some ways, this is not a new phenomenon. The suburbs are no stranger to this kind of built form.

But it's unique in that these boxes are not meant to be experienced in real life. They're a physical manifestation of our online activities, juxtaposed against our offline lives. It's two different worlds colliding. And already, it may be more appropriate to ask our question in the opposite direction: What's the best way to plan a city or suburb around data centers?

The Wall Street Journal recently published an article called, " Atlanta's Growth Streak Has Come to an End ." It's behind a paywall, though, so I don't actually know what it says. But Paul Krugman did write about it, here , and I do know that one of the key statistics that you should know is this: For the first time since the data was collected, net domestic migration to Atlanta has turned slightly negative.

Overall, the metro area is still growing because of natural births and international migration, but it's still noteworthy that more Americans are leaving Atlanta than moving there. Because up until recently, Atlanta was a high-growth metro region. It's an important logistics hub and it has had an elastic housing supply model. That is, it used suburban sprawl to keep home prices in check.

But that is starting to change. Housing supply is dropping and traffic congestion has become one of the worst in the US. Paul Krugman hypothesizes that this is an example of "the limits of sprawl." And I would agree with this. Sprawling cities have the advantage of being able to grow quickly when they're relatively small. But eventually, they reach a population and geographic limit where the model starts to fail.

The Atlanta urban region is massive. As defined by the US Census Bureau , it is 6,612.4 km2. The only urban region that is bigger is the one around New York City. Los Angeles — which might come to mind as another large car-oriented metro region — is smaller. It's about 4,239.4 km2, but with ~2.4x the population of Atlanta.

It may also surprise you to learn that Los Angeles is remarkably dense. When looking at the entire built-up urban area, it's the densest in the US at 2,886.6 people per km2; whereas Atlanta is one of the least dense big city regions at 771.3 people per km2. This figure really stands out when you compare it to its peers, which means it's going to be that much harder for it to overcome the limits of sprawl.

Density is the unlock that allows you to get people onto trains.

One traditional metric for measuring the performance of a company is revenue per employee. And in a knowledge-based economy, this makes a lot of sense. Human capital is often the biggest expense. But as we enter the age of AI, this is now being called into question.

Sara Menker has, for example, proposed a new metric: revenue per MWh . (See above comparing Meta, Alphabet, and Microsoft.) This is meant to reflect the fact that, as AI infrastructure scales, it is likely that operating costs in the future will be dominated by electricity consumption, rather than employee count.

Naturally, this should make you wonder about a few things, namely: How will we manage the inequality that might (or will) arise from the decoupling of revenues from employees? And how are we going to sustainability supply this rapidly growing need for more and more electricity?

Albert Wenger argues that the comparable metric for nations will be GDP per GWh . This means that, to win, you're going to want cheap electricity. And as I understand it, the cheapest sources are wind, solar, and hydropower. This bodes well for Canada given that we dominate in the latter.

Cover photo by Thomas Reaubourg on Unsplash