If you ever fly to Washington, DC, look out the window when you land at Dulles Airport – and you might just get a glimpse of the biggest economic story of the moment.
Below you’ll see scattered throughout the area’s fields and woods a collection of vast warehouses that, to the untrained eye, might look like supermarkets or distribution centers. But no: these are actually data centers – the largest concentration of data centers in the world.
Because this area surrounding Dulles Airport has more such buildings, housing computer servers that perform the calculations to train and run artificial intelligence (AI), than anywhere else. And with AI accounting for the vast majority of economic growth in the United States so far this year, that makes this topic a huge deal.
At ground level, you can see the features as you drive through what’s called “data center driveway.” There are huge power lines everywhere – a reminder that running these power plants is an incredibly energy-intensive task.
This small area, Loudoun County, alone consumes about 4.9 gigawatts of electricity, more than the entire consumption of Denmark. This figure has already tripled in the last six years and is expected to increase further in the coming years.
Inside “Data Center Alley”
We know this because we gained rare access to the heart of “data center aisle,” to two locations managed by Digital Realty, one of the largest data center companies in the world. It runs servers that power almost every major AI and cloud service in the world. If you send a request to one of these models or search engines, there is a good chance that you have unknowingly used their machines yourself.
Inside a site managed by Digital Realty
Their Digital Dulles site, currently under construction, is expected to consume up to a gigawatt of electricity in total, with six substations to help provide that energy. In fact, it consumes approximately the same amount of energy as a large nuclear power plant.
Walking around the site, a series of large warehouses, some already equipped with rows and rows of backup generators, there to ensure that the silicon chips whirring inside never lose power, is a gripping experience – a reminder of the physical underpinnings of the AI era. Although this technology seems lightweight, it has enormous physical demands. This involves building these huge concrete buildings, each of which requires huge amounts of electricity and water to keep the servers cool.
Sky’s Ed Conway at the data center
Learn more about Ed Conway:
Blow to UK’s bid to become a mining superpower
There’s more to your AirPods than meets the eye
We gained access inside one of the company’s existing server centers – behind several security cordons, in rooms accessible only with fingerprint identification. And there we saw the infrastructure needed to make these AI chips work. We saw an Nvidia DGX H100 running away, into a server rack capable of sucking up more power than a small village. We saw the cooling pipes that run in and out of the building, as well as those that supply coolant to the GPUs (graphics processing units) themselves.
Such evidence highlights that, to the extent that AI has intellectual capabilities, it is not delivered from scratch, but via very physical equipment and infrastructure. And the availability of this infrastructure is one of the main limiting factors for this economic boom in the years to come.
According to economist Jason Furman, once AI and related technologies are subtracted, the U.S. economy saw virtually no growth in the first half of this year. So much depends on that. But some wonder whether the United States will be able to build power plants quickly enough to fuel this boom.
Please use Chrome browser for a more accessible video player
Is Trump’s AI plan a “tech brother” manifesto?
For years, American electricity consumption has remained more or less stable. This has changed rapidly over the past two years. Today, AI companies have made big promises about future computing power, but that depends on their ability to connect those chips to the network.
Last week, the International Monetary Fund’s chief economist, Pierre-Olivier Gourinchas, warned that AI could indeed be a financial bubble.
He said: “There are echoes in today’s technology investment boom of the dot-com boom of the late 1990s. It was the Internet then…it’s AI now. We are seeing rising valuations, booming investment and strong consumption driven by strong capital gains. The risk is that with stronger investment and consumption, tighter monetary policy will be needed to contain price pressures. This is what happened in the late 1990s.”
“What’s terrifying is…”
For those in the AI world, this also seems like uncharted territory.
Helen Toner, executive director of the Center for Security and Emerging Technologies at Georgetown, and former member of the OpenAI board of directors, said: “The terrifying thing is that no one knows how far AI is going to go, and no one really knows what economic growth will result from it.
“The trend is certainly that the AI systems that we’re developing are becoming more and more sophisticated over time, and I don’t see any signs of that trend stopping. I think they’ll continue to progress. But the question is, how much productivity growth is that going to create? How will that compare to the absolutely mind-boggling investments that are being made today?”
Whether it’s a new industrial revolution or a bubble – or both – there’s no denying that AI is a massive economic story with enormous implications.
For energy. For the materials. For jobs. We just don’t know yet how big it is.