Mark Mills joins Brian Anderson to discuss the enormous energy demands of the world’s modern information infrastructure—“the Cloud”—the subject of his new book, Digital Cathedrals.
“Tech companies confront an inconvenient fact,” writes Mills. “The global cloud uses more energy than is produced by all the planet’s wind and solar farms combined.” In fact, digital traffic has become the fastest-growing source of energy use. While nearly every tech company has pledged to transition to renewable energy sources, most data centers are physically connected to the conventional power grid, fueled by hydrocarbons. The modern economy won’t be exclusively powered by renewables any time soon.
Audio Transcript
Brian Anderson: Welcome back to the 10 Blocks podcast. This is Brian Anderson, the editor of City Journal.
Coming up on the show today, I’ll be joined by a frequent City Journal contributor, the futurist Mark Mills, to discuss his new book Digital Cathedrals.
It’s an important contribution about the incredible energy demands of our modern digital infrastructure—the massive warehouse-sized datacenters built over the last few decades—that has allowed billions of people to access the global cloud of information and online services like streaming videos. It’s published by Encounter Books, there’s a link to it in the podcast description, or you can find it on Amazon.
Before we get started with the interview, I want to note that this week marks the fourth anniversary of launching the 10 Blocks podcast, which just keeps on growing. Thank you to all of our listeners for tuning in, and we hope you’ll tell your friends to subscribe to the podcast. And maybe get them a copy of City Journal magazine while you’re at it.
My conversation with Mark Mills begins after this.
Hello again, everyone. This is Brian Anderson, the editor of City Journal. Joining me in the studio today is Mark Mills. Mark is a senior fellow at the Manhattan Institute, writes regularly for City Journal, and he's a partner in an energy tech venture fund. He's had a long career analyzing the energy economy in government and the private sector, and we're happy to have him on the podcast. Mark has an interesting pithy new book published by Encounter called Digital Cathedrals. We'll link to it in the description, and you can find it on Amazon or wherever books are sold. Steve Forbes calls it, “A mini sized book that packs a mega-sized wallop of information” about the digital age. Mark, thanks for joining us.
Mark Mills: Great to be here, Brian.
Brian Anderson: Your book has a striking title. I just mentioned it, Digital Cathedrals.. What you're really talking about is data centers, the physical infrastructure that allows us to shop online, to stream videos, everything else digital, but that term digital cathedral, can you explain what that means to you?
Mark Mills: Well, as a sort of an amateur student of the history of technology, one era of history that I am most fascinated by was the era of the construction of the great cathedrals. The masons of that era were revered. The cathedrals, of course, were built for a theological purpose, but they also served a commercial purpose. They were the epicenters of commerce in the villages, but more importantly, the ability to build those cathedrals, they were the biggest habitable edifices ever built by humanity for thousands of years. And in fact, the tallest cathedral, which was competed in the 14th century, remained the tallest habitable building ever built by humans until the Woolworth Building in Manhattan in the early 1900s. So a 500 year run.
So it epitomized the technology of the era, that one could even build such things, and the reverence. In our era, we're building these massive data centers by what we could call digital masons, that there are thousands of them. There are more of them being built than there were cathedrals. It's not just the technology that allows us to do the modern digital cathedral. For me, it's also, it's kind of the reverence that we hold in our modern era for the companies and the people who do these things. Many people are in awe of that almost in a quasi-theological way.
Brian Anderson: For years, we've heard the phrase data is the new oil bandied about when it comes to digital technology, but data requires energy to transmit, and you calculate that the global internet cloud now uses more energy than is produced by all the planets, wind, and solar firms combined. Yet despite that, nearly every tech firm in the country has now taken the pledge to transition to use 100% renewable energy. How do companies expect to meet those goals and are they even remotely realistic?
Mark Mills: Well, no, they're not realistic, but it's good PR. It may be serious, to give them credit. It's certainly good virtue signaling that we're going to run our digital cathedrals from wind and solar. As a practical matter, all of them know, in fact all of them honestly disclose as much at their websites, that their operations physically operate. They're physically connected to the conventional grid, using whatever electricity the grid provides, which is globally on average roughly three to four percent of global electricity from wind and solar, maybe five next year. So if the world's...
Brian Anderson: Steadily growing.
Mark Mills: It's growing, yeah. So the question would be, we already know that the digital enterprise broadly uses more energy than all the wind and solar farms produce. The question is, how fast are the two growing? Wind and solar are growing very fast, but the demand for digital things and the fuel for them, which is electricity, is growing even faster, which by the way, in my world is a good thing because this reflects the utilization of the most profound and biggest new infrastructure that humanity has ever built.
Brian Anderson: Google, to single them out, for the last two years, and this relates to the last question, they've reported that their company has actually made the energy transition, that they're powering all of their operations with renewables. That doesn't seem to be true. Is it?
Mark Mills: Well, so they're using a, to use a phrase, a Jesuitical definition of powering. Their data centers are directly connected to the grid. So if the data centers are in Iowa, they're using coal and wind. If they're in southeastern United States, using natural gas and some wind. If they're in Texas, they're using gas mostly and then some wind. What they do, and what all companies do, is they buy, to stick with theology, they buy indulgences. It's a legal transaction in which if you purchase the output from a wind farm somewhere on the planet, anywhere on the planet, in fact, you get a renewable credit. You can then legally bolt that credit onto your facility, data center, could be an office building, like the Empire State Building, which did that. The Empire State Building bragged that it was 100% renewable power by buying an equivalent amount of electricity that it consumes on the New York grid from wind farms in Texas. Last I checked, you can't move the electricity from those wind farms to Manhattan. So these are niceties, they feel good in PR, but the practical reality is that these facilities depend on a reliable grid with low cost electricity.
Brian Anderson: You know, every month we're reading about improvements to green energy technology, whether it's improved battery storage, wind is becoming cheaper in certain areas for sure. But you've remained pretty skeptical, I think it's fair to say, about the ability of the current renewable energy sources to fully power the future economy. What are you looking at that environmental activists and green investors are missing?
Mark Mills: Reality.
Brian Anderson: Yes. I guess it just comes down to it, when you talk about Jesuitical distinctions.
Mark Mills: I mean, I'm only being half facetious, so of course wind farms, wind turbines, and solar cells and batteries get better, this is what engineers always do. There's two questions one has to answer in the real world, is how much better can they get, because those are boundaries that are just dictated by physics, and could you use them as good as they are at scales where you replace, which is the contention, replace all the oil, gas, and coal use in the world? The short answer is no. Not only do we not have the scale to do that, the world gets 3% of its energy from wind and solar, 54% of all of its energy from oil and gas. To replace the oil and gas share would require an infrastructure expansion of hardware, concrete, steel, nickel mining, 90 times more than we did to increase the infrastructure of oil and gas over the last half century. This is sort of fantastical, not going to happen. There'll still be a lot more windmills and solar arrays in the world, lots more Teslas, but they're not going to replace all the oil, gas, and coal, and so therefore everything to do with data, which infuses the entire economy increasingly and deliberately, will remain dependent on the mix of energy that we use to provide society.
Brian Anderson: What's your view of the possibilities for battery storage? This is something which is improving. There's been some advances in recent years, but it would seem to be very crucial for electric vehicles, say, to become fully operational. Right?
Mark Mills: Right. It's crucial for electric vehicles. It's a crucial for...
Brian Anderson: Solar power.
Mark Mills: Crucial for the electric grid, too, because when the sun doesn't shine, the wind doesn't blow, you need to have energy. So you have two choices. You can either store it in batteries if you like, or you can build a shadow grid sufficient to provide the power when the wind isn't blowing, the sun's not shining. And the size of the shadow grid would have to be rather large, because it'd be very expensive because you don't use it very much. All you need to know is, for example, last I think two weeks ago in Alberta, because they have a lot of wind farms there, there were four days where it was cloudy in a row with no sun and no wind at all. In the United States for the last 30 years, there have been dozens of times when the entire continent of the United States has been cloud covered continuously for two days, and no wind. The whole country be calmed. That would mean that if you're planning a grid for a few decades, you would have to count on a dozen or more times where you would need to replace the entire grid instantaneously for days. The cost of that is in the trillions of dollars. It just...
Brian Anderson: Boggles the mind.
Mark Mills: So build batteries. The scale of batteries involved also boggles the mind. So if we just picked one day of total storage we'd need for the United States of no wind and sun, we would have to use the physical output of the Tesla Gigafactory, or world's biggest battery factory. We'd have to use a thousand years of production from that battery factory to build enough batteries to be available to store that much energy. I'm not even talking about what that costs. I'm just talking about the scale. But in the physics of storage of energy, the thing to keep in mind is that storing energy, which is critical, storing food's critical, storing water is critical. It costs about 50 cents a barrel of oil to store a barrel of oil for a month. That's a rough average storage cost. To store a barrel of oil equivalent of energy in a battery, the cost is about $100 to $200. This is a big gap. It's not going to be closed with subsidies and wouldn't even be closed with the physics of batteries.
Brian Anderson: One of the next steps in the modern digital transformation of our world and economy is, of course, artificial intelligence, AI as everybody calls it. Now, it's an incredibly power intensive form of computing, as I understand it, and the applications for it are extraordinary and hard to imagine. Where do you see the development of artificial intelligence at right now? Where do you think it's going to be taking us, and what does it mean for energy consumption?
Mark Mills: So artificial intelligence is both underappreciated and over hyped. It's kind of hard to imagine that they're simultaneously possible. The over hyping is partly in the fear of what it will do, but also expecting or promising something that can't happen as fast as claimed. So artificial intelligence, though, is as big a leap in computing as going from mainframes to personal computers. It is possibly a bigger leap than that, because what we're now doing is changing what we use computers for. Computers were intended to compute, hence the word, spreadsheets, calculate things, give a specific answer and outcome. What AI does, what artificial intelligence is intended to do, is not compute, but to estimate, to advise. That's what Siri does, or Google Voice, [inaudible 00:12:05] Alexa. This is a profoundly different use of computing. It turns out in the engineering sense, it's the single most energy intensive use and data intensive use of computing in the history of computing.
So it promises in the social economic sphere some remarkable advances. You know the idea that, for example, that AI will replace pathologists in reading x-ray film, I don't think that's going to happen. What it'll actually do is amplify the efficacy of the pathologists, reduce errors, improve accuracy. That's what AI does broadly. This is a great thing, and that would apply to every feature of life, not just driving vehicles, but automating supply chains and changing how we can do translations and communicate. The applications are in a sense more unlimited than imagining, say it's 1984, who imagined Amazon? Who imagined e-commerce? We thought that only about email, you've got mail of AOL days.
So the transformations will be as significant as that, but because it's so profoundly energy intensive and so profoundly useful, it actually has pretty interesting energy implications. Facebook has already said that their data centers' power use is doubling every year, and they're saying it's almost entirely because not just there's more social media. It's because of AI.
Brian Anderson: Fascinating. Before we wrap up, Mark, why don't you tell our listeners a little bit about what your next project is, what you're working on?
Mark Mills: Well, my next project is a little grander than just Digital Cathedrals and the future of energy. It's the future of technology in general. So the interesting thing if you're a forecaster is the thing that's, I think easiest to forecast, is technology. It's hard to forecast politics because of humans, and economies are a little hard to forecast, but I think technology is relatively easy to forecast in its direction and structure. And it's what I've been doing a lot of my life, so I'm coalescing a lot of the lessons I've learned as a forecaster, not just in energy uses and energy production, not just in computing, but across this domain. And in doing this research over the years, it's led me to the conclusion as sort of the opposite of where the current [inaudible 00:14:26] is, which even though people are unhappy about the disruptions from digital technologies, we have a general theme that innovation, the foundational innovation, has slowed down. A lot of scholars think that.
I think they're profoundly wrong, I think we live at the beginning of an era as profoundly transformative as the so called Industrial Revolution of a century ago, and it means that for people, particularly people who are young today, it's the most exciting time you could be alive.
Brian Anderson: Don't forget to check out Mark Mills' latest book. It's called Digital Cathedrals. It's published by Encounter Books. You can find it, as I mentioned at the top, on Amazon and we'll link to it in the description. His latest piece for City Journal is drawn from the book. It's on our website. It's called Old Energy, New Boom. You can follow Mark on Twitter, @MarkPMills. You can also find City Journal on Twitter, @CityJournal, and on Instagram, @cityjournal_mi. And always, if you like what you've heard on the podcast, give us a rating on iTunes. Thanks for listening, and thanks Mark Mills for joining us.
Mark Mills: Thank you.
Photo by gorodenkoff/iStock