In the quiet Washington neighborhood suburb of Ashburn, Virginia, lies a large but unassuming building. There is a small pond with a fountain in front of you, and the only sounds you can hear are the ducks paddling in the pond and the quiet noise of the highway in the distance.
But inside is a huge data center that many consider to be part of the backbone of future technology development. The servers are kept neatly organized and secure behind locked cages with rows of loud blaring sounds and constant surveillance.
Artificial intelligence is driving unprecedented demand for data centers around the world, and Virginia is believed to have at least 650 data centers, leading some technology insiders to call the state the data center capital of the world.
“Some say 80 percent of the Internet traffic in the United States ends up here in Northern Virginia,” said John Lin, Equinix’s chief business officer. The National During a rare tour of one of the company’s factories in Ashburn.
Equinix is one of the world’s largest data center companies, operating 273 data centers in 37 countries, including France, Oman and the UAE.
AI may be the latest technology buzzword, but Equinix believes that the overall trend of cloud computing and digitalization will continue to strengthen this sector.
“The mainstream use of computer data continues to drive demand, and it is a steady development,” Lin said. He added that AI is creating more demand around the world, but nuance is important when discussing it.
“New AI challenges are demanding more equipment with much larger production capacity, creating new pockets of demand around the world.”
He said that with the growth in AI and the surge in investment, some companies are choosing to build hyperscale data centers “focused on supporting one customer’s needs and growth.”
Equinix has on-site backup generators to ensure you always have a reliable power source in the event of a power outage. Various mechanical systems are also in place to prevent servers from overheating.
“Over the past five years or so, we have been supporting liquid cooling more frequently,” Lin said, noting that water is reused in the process of cooling this technology.
“From a sustainability perspective, we want to be as efficient as possible and be good stewards of all the infrastructure and energy we consume.”
grid pain
There is some controversy there because not all data center operators take the same approach.
To realize the vast potential of AI as a force to transform humanity, data centers will require more powerful graphics processing units (GPUs) and central processing units (CPUs), especially compared to cloud computing and traditional data centers of the past few decades.
These GPUs and CPUs require large amounts of power and are already straining the power grid.
Companies like Microsoft are partnering with energy companies to increase power generation, even going so far as to bring previously decommissioned nuclear reactors back into service to power energy-hungry data centers.
In the United States, soaring electricity costs are driving the construction of AI infrastructure and what some are calling “zombie reactor” projects.
Announcements by two U.S. senators late last year provided a glimpse into the continuing debate over electricity rates and AI data centers, which is expected to gain further momentum next year.
Democrats Elizabeth Warren and Richard Blumenthal said they had launched an investigation into the proliferation of data centers, warning that they could be driving up electricity prices across the country.
“Utility companies have spent billions of dollars upgrading their power grids to accommodate the unprecedented energy demands of data centers, including building expensive transmission lines and new power plants,” they wrote, arguing that these grid upgrades would eventually be passed on to consumers.
A recent report from the nonpartisan Pew Research Center notes that Americans have been paying an increasingly high price in recent years to keep the lights on.
The report cites a Carnegie Mellon University study that predicts that data centers could contribute to “an 8% increase in average U.S. electricity bills by 2030, potentially exceeding 25% in the highest demand markets in central and northern Virginia.”
Rising energy costs, combined with the rising cost of living that is already stressing many American households, are taking a toll.
Critics such as Warren and Blumenthal say big tech companies often don’t pay their fair share for additional infrastructure.
“The contracts between data centers and utility companies that underlie these arrangements are in most cases confidential,” they wrote in a recent letter to the chief executive of Amazon, which builds and operates its own data centers.
In the year ahead, policymakers are likely to take a closer look at how best to address these criticisms, but the classic “not in my backyard” factor remains a force to be reckoned with.
Various cities and municipalities are already discussing data center installations in public meetings, and residents are demanding answers about their electricity usage.
At a conference hosted by the Congressional Internet Party Academy, Leslie Abrahams, senior energy fellow at the Center for Strategic and International Studies, said one in six American households are already behind on their electricity bills.
He said the U.S. power grid has extra capacity during off-peak hours, but as more data centers become available, “that excess capacity starts to get squeezed.”
“So we’re talking about the need to add new infrastructure and new capacity,” she said.
According to Columbus dispatchAmazon secretly received a reduction in electricity rates from American Power while the data center was being built.
Amazon later said the deal was no longer necessary. Columbus dispatch reported, but the story sparked widespread concern and backlash from consumers who have been hit by soaring utility bills.
The United States is not alone in trying to deal with the growing pains of building data centers.
Data centers have proliferated in Ireland in recent years, with these sites collectively consuming more electricity than the country’s homes, according to media reports.
Grid gain?
But supporters of expanding AI infrastructure say it’s unfair to blame data centers and point out that electricity bills could be even higher if America’s big tech companies don’t partner with power companies to upgrade the power grid.
Equinix’s Lin said the company has long held the position that consumers should not be burdened by the proliferation of data centers.
“We’re spending money on infrastructure investments with our utility partners, and we’re trying to help in any way we can with that,” he said.
“The kind of trajectory that we’re seeing right now in terms of AI and accelerating growth is huge and I think it’s going to naturally flatten out a little bit eventually because the power demand that’s currently projected is so high and there’s utility concerns and all of that is going to end up holding back some of the growth.”
However, Lin said some people’s concerns center around a number of misconceptions about AI and technology as a whole.
“People think it’s just being used to make things like pictures of cats, but the actual use case is much more complex and could have immense benefits for society,” he said.
AI aside, he reiterated that consumers are streaming more videos, shows and movies than ever before, and are also relying more on the cloud to store their data.
“These are the things that are actually contributing to GDP, and these are the things that are actually producing economic outcomes,” he said, pointing to the employment opportunities that data centers provide with healthy wages.
“It’s a very competitive market with very specific skills needed to do this. The property taxes that we actually give back to the community are a big driver of why Loudoun County, Virginia, has some of the best schools in the country.”
Lin is not alone in his enthusiasm for the benefits data centers bring. Many elected officials across the country have focused on this area and have sought to embrace it.
Levi Patterson, Nvidia’s director of energy, science, and AI infrastructure policy, told the Congressional Internet Caucus Academy that electricity prices don’t necessarily reflect basic economic rules.
“The supply and demand curve that says more demand means more prices is misleading because that’s not how the grid works,” he said, hinting that the power data centers consume could ultimately help stabilize electricity prices.
“If there is more demand to share these costs with the grid, it will actually drive down prices, because we are all sharing these large fixed costs.” He added that big tech companies’ desire to help reduce the cost of transmission and distribution across the grid could also help.
Some analysts and experts say AI is already being used to improve the efficiency of electrical grids, and those behind the effort are just beginning to scratch the surface of the possibilities.
