By James Pethokoukis and Mark Mills

From the Industrial Revolutions to the post-war boom, general-purpose technologies like the steam engine and electrification have always been behind periods of rapid economic growth. Looking ahead, some economists predict that artificial intelligence could be the next general-purpose technology to usher in a new economic boom. What has received comparatively little attention is the rapid expansion of cloud computing over the past decade. What, exactly, is the cloud? And could cloud computing spark the next economic and technological acceleration? To answer those questions, I’m joined by Mark Mills.

Mark is a senior fellow at the Manhattan Institute and a faculty fellow at Northwestern University’s McCormick School of Engineering and Applied Science. His latest book is The Cloud Revolution: How the Convergence of New Technologies Will Unleash the Next Economic Boom and A Roaring 2020s.

What follows is a lightly edited transcript of our conversation. You can download the episode here, and don’t forget to subscribe to my podcast on iTunes or Stitcher. Tell your friends, leave a review.

Pethokoukis: One of the big issues here in Washington has been infrastructure. The president signed a big infrastructure bill, and when people hear infrastructure, they think roads and bridges — that sort of thing. But your book is about infrastructure and a kind of infrastructure that may be more important than some of the things that are going to be funded by this new trillion-dollar bill.

Mills: Yeah, I think so. I mean, we’ve been building roads since Roman times famously, so we’re all going to need roads for a while, I think, and bridges. The biggest new infrastructure that humanity has ever built is being built by free markets (and some non-free markets, but mostly free markets) today, which is, loosely, the cloud. And most people don’t really know what that means, but it is not an ethereal thing; it’s a physical infrastructure. By all measures — dollars, physical equipment, square feet of buildings — it’s the biggest infrastructure humanity has ever built. It’s really quite remarkable.

In the book, you call it the cloud. A somewhat less glamorous name that it’s also been referred to as, or at least it used to be referred to as, is a “warehouse-scale computer” — you quote a Google engineer who says that. It’s somewhat less prosaic than the cloud. Is that a good way of thinking about it? It’s just like a big computer or a big warehouse full of so many computers that it, itself, is a mega computer?

Yeah, well, that’s a part of the cloud. I would say 90 percent of people in America today use the cloud close to daily. If you do finance on your phone or a computer, if you order with DoorDash, Airbnb, or you use Google Maps, all these functions use a device that you hold, whether a smartphone or a computer. They connect through wired and wireless networks to massive, (as you put it and Google called it) warehouse-scale computers, data centers, to do processing.

They don’t just store information and calculate. They do inference things, not just calculations. They’re trying to disintermediate all kinds of complex relationships for you to find the food you want or the house or the ride. So all those features together, all that physical infrastructure combined, is the cloud. But at the middle of the cloud — sort of its beating heart, if you’d like — are these extraordinarily big warehouse-scale, Walmart-sized buildings that are jam-packed full of computers and servers and storage that don’t just store information. They do things that we talk to Alexa or Siri about.

We can communicate with it, of course, but it’s more than communication. It’s all the things that we take for granted. Zooming or asking for opinions, something like “If I like this, I would like that; if you like A, you might like B.” What Amazon does when giving you suggestions about books. These are all inferential exercises that look at all kinds of data. Most of it we think is private, some of it is not so private. (That’s a whole separate issue.) But it’s also what’s making supply chains more efficient shortly. It’s hard to do supply chains, like it’s hard to do health.

So the things that the cloud has made better in our lives, by and large most people — not everybody — think that the way we get our news is easier and better. It has had disruptions. The way we get entertainment is easier and better with disruptions. The way we do our finance is clearly easier and better than having to trudge down to a bank, and doing Zelle or Venmo takes out a lot of friction. But those are the worlds of information, bits worlds, that are much easier to automate. That’s 20 percent of GDP. People have been babbling about the next Industrial Revolution for decades.

It’s really hard to disintermediate the physical stuff — transportation, manufacturing, supply chains, growing and building things — because it’s the world of atoms. They move with the laws of physics. They have inertia. They have consequences if you get it wrong. Blue screen death, jitter — it’s inconvenient when you’re watching Netflix streaming. If that happens when the self-driving car is driving you or to a load of goods from a port, it’s not so good. It could be real death. So it’s slower moving, harder to do, but that’s what’s happening now. My book is really about the intersection that the engineers and scientists call a “cyber-physical revolution.” The cyber world finally merges with the physical world to add efficiencies.

I think when people think about it, they think about data centers, buildings. But as you nicely put it in the book, it’s not just that: “Arrays of communications hardware propel bytes along ‘highways’ constituting not only roughly three billion miles of glass cables, much of it buried, but also the equivalent of another 100 billion miles (that’s 1,000 times the distance to the sun) of invisible connections forged by 4 million cell towers.” So even though we’re talking about the cloud and bits, this is a real, massive, physical thing that we’ve created almost, in a way, very quietly. (We may see some cell towers.) This is a massive human endeavor that to many people seems like it just kind of happened. When did it start happening?

We began with the first computer rooms in 1944 and ’45, right?

Right.

Colossus was the first computer. I give credit to my homeland: the British beat the Americans in that case. ENIAC came second, and then in the early days of computing, we talked about computer rooms and the computer centers, and then we called them data centers.

You had to use terminals. You didn’t have a computer. There would be a kind of a dumb terminal that you would go to, which was the computer. I think that’s how Bill Gates and Paul Allen started.

They all begin with a keyboard. You type in a dumb terminal. We think of smartphones as smart terminals — that’s essentially what they are, but your smart terminal in your hand or pocket has 10,000 times the computing power of an IBM mainframe of 1980. So it’s a pretty smart terminal, but it’s dumb by data centers standards. It began a long time ago, but the cloud as we imagine it — which is different than a data center, you’re right — you were forced, you and I, if we wanted to use a computer, to go to the computer, to present ourselves to the beast. Maybe we would type code on a Fortran card that it would read. You went to it. Obviously, what’s happened now is the computing function is coming to us, to the edges, to human beings, to machines, to parts of our body.

That’s really consequential. It’s very different. The cloudification, if you like, of computing, really began in earnest about 20 years ago and took off about 10 years ago. And I’ll give Jeff Bezos credit for figuring that out. Not only that, what the cloud could do was different than “You’ve got mail” AOL email, circa 1990. Yeah, email is a big deal, but email is just a communications device. It’s a way of sending mail. We know that what you use a smartphone for: from facial recognition to turn it on, to not just literally talk to the cloud or somebody else, but get information, instruction, advice, talk to a doctor. Increasingly, that smart terminal can be a diagnostic device.

Probably most people have seen the advertisements for the EKG that you can connect to your smartphone and give your doctor medically useful information. That kind of capability is not just about end-use devices but analyzing the data with supercomputers in the cloud, what we used to call supercomputers (now so super it defies description). And it’s happened, to your point, sort of quietly. People have taken this velocity for granted. They’re both amazed by it, like, “Wow, look at my smartphone,” and the same time inured by just the common distribution of it, like, “Well, it’s whatever. I got a smartphone.” Well, computing power has increased a billionfold in 50 years, in computations per second of the best of computers. A billionfold. No other product has improved its performance that much in all of history. That has to be consequential, but what’s even more consequential is you move computing into not just calculations, but inference in information storage.

Inference is obviously different than computation. I’m sort of estimating, guessing. “It’s probably the best route for you to take.” “This might be the best medication for you to take.” I mean, like a doctor. Doctors don’t know, so AI in the cloud could advise a doctor. It doesn’t replace a doctor, but that function, if you think in economic terms, has a cost function, right? It’s “What does it cost me to get that stuff?” Well, we can measure that, too. So over the same 50 years, if we measure computations per second per dollar, how many computations per second do I get to buy for one buck? How has that changed in 50 years? Well, Moore’s law gave us a billion times more computation horsepower in 50 years. What the cloud has done is increased our access, in economic terms, a hundred billionfold.

You get a hundred billion times more computational power per dollar now than 50 years ago. I mean, we’re both underestimating in the sense of what that means in many forecasting domains and, in a way, overestimating in some areas where people are being a little goofy about AI — like somehow, it’s going to replace people. It’s “Artificial intelligence: end of jobs.” So we have this dichotomy, almost schizophrenia about it. But this is sort of typical history. We were schizophrenic about nuclear energy. We were schizophrenic about automation when it started. We were schizophrenic about the Industrial Revolution, about trains, about cars, about the telephone. I mean, the list goes on. People have schizophrenia of these things.

And just sort of one more basic question: Who owns this? You mentioned Amazon.

That’s right.

This giant network machine that does all these calculations, the inferences — who’s running this thing?

Well, if you’re in North Korea, we know the answer.

Right, right.

If you’re in China, we know the answer. In the free world, there’s no one organization that runs “the cloud.” Amazon has the biggest cloud presence in the world and is still the leader, but others are catching up. There are features to the cloud that Amazon doesn’t run or own, like the networks themselves, for example. Some data center operators, Facebook and Amazon, do lay their own fiber cable, typically to ensure they have access. It’s like private highways for their users, and often it’s to ensure security. I mean, one of the things we all care about is the security of our information. So there’s no owner. There are dominant players, and what’s happening right now as the infrastructure expands is that there are a lot of new players growing very rapidly that own different features of it.

The analogy doesn’t really work because information is so different, phenomenologically, physically, but it would be like asking, “Who owns transportation?” Well, there are airplanes and ships and there are cars. There are highways. There’s a lot of stuff in it. There are businesses that own a lot more of the infrastructure, and governments have a big say in where roads go. They have a big say in where cell towers go and where fiber cables get laid, right? But we do have the capacity to make the rules of the road, which is the regulatory space, and make them onerous, Chinese-like or North Korean-like? One hopes we keep making them lean towards freer markets. The idea that we should control this beast, the cloud, by overregulating it sort of assumes it’s mature, that we don’t know what the next evolution will be.

And I think, to put it obviously, we’re at the end of the beginning of the evolution of the cloud. We’re not at the end phase. I mean, let me do a calibration in dollar terms because money matters. At the macro level, the world is now spending more on capital equipment, so hardware. I’ll talk about building up cloud infrastructure, which is again, everything I described. It’s not just data centers. We’re spending more on that now than we are spending on oil and gas infrastructure hardware. Think about it. Or put it in utility terms: We’re spending far more building out clouds that consume electricity than all the world’s utilities are spending annually to produce electricity and distribute it, so it’s already become a big consumer of capital.

The warehouse-scale data centers are hard to imagine and visualize if nobody’s ever been in one. If you’re a science fiction fan and you picture the Borg ship in Star Trek, they kind of look like a Borg ship. Not many people, a lot of blinking lights and computers running hard. So one square meter of a typical cloud data center has about a thousand times more compute horsepower than the whole world had in the early ‘80s, and we’re building out data centers at the moment at the rate of about 10 million square feet a year. And data centers interestingly cost about the same to build as a skyscraper like the Empire State Building or the World Trade Center. Dollar per square foot, it’s about the same.

They rent out for about 500 percent more per square foot unsurprisingly, because of the density of information compared to the density of human beings in a skyscraper. We’re building data centers at about $300–400 billion worth a year. We’re not building $300 billion worth of skyscrapers a year. We’re building a 10th of that, so all the metrics that we think about in the economy — square footage, dollars, rents, and let’s talk about power: A data center uses a hundred times more power per square foot than a skyscraper. And we’ve already built out in square footage 300 percent more data centers than scrapers, and we’re building them out at about 400 percent the rate.

So it’s a big infrastructure, and it’s not something that’s being pushed on people. It’s being built in reaction to our consumption of the products it offers, and the products it offers are . . . well, we know what they are. It’s not just Zooming. It’s not just Netflix and streaming video. There isn’t one of us who wouldn’t be happier with an even better medical advice interface. Not replacing our doctor, but making it easier for my doctor to know something about me. It would be easier to collect my personal medical information, easier to store it securely, and then to have supercomputers look at my blood chemistry, my EKG over the two weeks before I go to the doctor to give advice to the doctor. The first thing that happens when we all go to the doctor, what do they say to you if you’re not feeling well? “Oh, what did you eat two days ago?” Or all sorts of questions that you have no idea about. And here’s a doctor, he gets a snapshot in time.

Economists like to use the phrase “general-purpose technologies,” and those are technologies that enable other technologies and become sort of embedded in those technologies. I mean, electrification is a general-purpose technology. Maybe the internal combustion engine is, too. People will talk about AI as a general-purpose technology, but your book isn’t called “AI.” It’s called The Cloud Revolution, so is the cloud a general-purpose technology? And if it is, how does it enable or work its way into the other technologies that you think are going to speed up the economy and productivity growth? All the stuff we love to talk about in this podcast.

I like economists. I make fun of them a lot because it’s easy to do. It’s kind of like making fun of lawyers because there’s a lot of both of them, but I do quote some of my favorite economists in my book. A good economist is a really magical thing in my opinion. They understand how markets and economies work. So first, to be clear, the electric motor is a general-purpose technology. Electricity is how you power it, so I’m saying that for deliberate reasons. An internal combustion engine is a general-purpose technology. You can use it to fly airplanes, drive ships, cars, make electricity — so that’s a general-purpose technology. Microprocessors are general-purpose technologies in the same way. They’re the building blocks of the cloud, your smartphone, end-use devices, so that’s the general-purpose technology.

The cloud is more like a utility infrastructure. The way electricity is, the way water is, with a distinction that’s not trivial. It enables other general-purpose capabilities quite unlike electrification, because it invades all the others. Electricity has invaded a lot and we keep electrifying things, but it’s not that easy to electrify everything. It just isn’t. Some things are better done mechanically or with combustion. There’s almost nothing that we do for which more information is not useful. In fact, I would word it differently. In everything we do, everything in life, more information and knowledge has utility. So when I take a general-purpose engine like a microprocessor and distribute it to the scale we distribute it at, we really change the world in a way that no other general-purpose technology has.

Now, one economist in particular likes to label AI — which is a different class of computing chip, computing software, because it infers, estimates, instead of calculates — as a general-purpose technology. I don’t call my book “AI” because AI is one of the tools in the cloud, both on the edge that is in your device — because there’s an AI-driven engine that does facial recognition. It’s a very simple one, but that’s what that is. And there’s AI in the cloud. The AI in the cloud is what we use to do drug discovery now. We put a lot of data about molecular behavior into a supercomputer and run an AI program, not to calculate the best combination of molecules for therapy. We put AI in the cloud in order to estimate what would make sense as a better drug or better therapeutic, and that’s what human beings do. We estimate. We don’t really calculate.

Some things should be calculated; some things should be estimated. And there are differences, so they’re complementary. The idea that AI is general-purpose technology is true, but you need more than just AI to do the kinds of things we’re talking about. Let’s use healthcare as an example. I want very specific data about my body’s state, but I don’t want to estimate it. I want to know my temperature, EKG, EEG, blood sugar — there are hundreds of things I’d like to know in real time if I could collect it. You would do that with sensors which are made possible by new classes of materials, which I write about, and we can have those sensors communicate in real time to a device like a smartphone now, which was unprecedented in history. I don’t have to wire my body.

Engineers are now talking in terms of a body-centric network, so you have your own body internet that collects information about you, which you can securely keep and protect, and share with your doctor when you choose to. Pretty useful, but then some features of that are digital, “I want to compute,” some features are inferential. So it’s the fusion of those functions, combined with the fusion of large to small computing — in effect computers I can swallow, with supercomputers the size of a warehouse that can make sense of the complex data.

So it feels like a complicated story, but if I tried to tell the story of electrification in 1902, you begin to describe, “Well, I could make an air conditioner.” People would look at you doe eyed, like, “What do you mean, an air conditioner?” Hero of Alexandria imagined cooling things, but no one could imagine an air conditioner until you had electric motors and compressors. You couldn’t imagine all the things we’ve done with electricity since, including lasers and radar and radio communications. Those are all entrepreneurs doing things with the new phenomenology. We’re at that stage with the general-purpose technologies that are part of the future of the cloud.

I think a skeptic might say, “Well, the cloud’s been here. We’ve had the cloud, and no one is saying it’s not super useful, but do all these uses really add up to enough?” I’ve certainly written about weak productivity growth. How innovative is the economy? Does the cloud, and the technological spheres in which it gets woven into, add up to ultimately a much faster economy, a much more productive economy, and an economy that will raise living standards faster than they have been raised?

Yeah. Well, that’s the part that matters, right? I mean, the other stuff is interesting. We can see movies more easily, so if you’re an economist measuring the productivity of movie distribution, it’s off the charts better. Whatever economic measure we want to use, dollars spent per minute of TV. And by minute — you count the minutes it takes not just to watch it, but to get to it, right? So it’s an incredibly productive way to watch movies, but so what, right? I mean, it’s fun. Netflix is making a lot of money, but does that move the meter in the productivity of the economy? And the answer is, at the high level of economics, no, because the bits part of our economy (news, entertainment and finance) are important, but collectively, they are roughly 20 percent of GDP.

So we have improved, with information, the information-centric parts of our economy. No one would dispute that the ATM in your handheld is better than walking to an ATM. Zelle and Venmo are faster. They’re productive, but they’ve accelerated productivity in the easiest part of the economy. Your point is well taken because most economists have been saying correctly that productivity growth has been lagging in industrial markets, mining, manufacturing, and especially in healthcare. A group of researchers turned Moore’s law upside down, called it Eroom’s law, and pointed out that we’ve been increasing spending and getting poorer outputs, fewer outputs, which is the inverse of Moore’s law, in healthcare. So at best, it’s been flat for two decades.

If you want more healthcare, you spend more money and hire more people. That’s the model. That’s a terrible model that’s unsustainable. So I spend time in my book specifically mapping out that domain about “Where are we going to get productivity in the things we care about?” Manufacturing, transportation, healthcare, especially. I look at entertainment and education. We need more productivity in education. I think that’s coming as well. Well, the way I map it out is not to speculate “Wouldn’t it be nice if . . .” but rather looking at what’s already happening that’s pre-commercial in each space, and then one can easily speculate. Again, let’s do healthcare.

Telemedicine is a much better way to do pre-screening rather than going to the doctor for everything that ails somebody. You still want to go to the doctor? Great. Maybe one of the three times you should have shown up. The other two, we do by telemedicine. That got unlocked by the great acceleration known as the great lockdowns, right? But that’s consequential. I’ll use an easy to imagine example. One of the biggest problems in hospitals is staffing, and the nurses spend a lot of their time doing physical lifting, for obvious reasons. It’s no longer unimaginable that we can make physical lifting a lot easier using anthropomorphic robots. It seems kind of like Terminator or science fiction. It’s not anymore. The crossover comes when the economics are better. When it’s, let’s say roughly equal, or even slightly more expensive, to have robot assistance for the physical task a nurse does in a hospital, rather than hiring another nurse because nurses have much more utility and value than physically lifting something up.

But you have to have robots, which get vilified and they’re fun in science fiction. I love robots in science fiction, but robots have been overpromised for a long time because they’re damned hard to make useful in the environments in which humans operate, which is they need to be anthropomorphic to be useful. They’re great in a factory where there are no humans and it’s a repetitive task, but these general-purpose tasks which humans perform need a general-purpose technology like a robot to emulate them. I mean, anybody that’s interested in this space has probably seen lots of videos on YouTube of robots that are pretty impressive. They’re not CGI.

They’re jumping, right? They’re jumping.

They’re jumping, backflipping, carrying things. As amazing as they may seem, I would put them technologically with what was, you probably know your history, the Duryea Wagon, which was one of the first automobile manufacturers in the late 19th century in the United States. Rich people had cars for 20 years, 30 years before the automobile got cheap. The Model T moment, which just predated 1920, was when automobile ownership took off. We finally made them cheap enough and generally useful and reliable enough so people could buy them. It’s interesting that in inflation-adjusted dollars, in real dollar terms, the Duryea Wagon cost about the same as this Boston Dynamics Spot Mini, which they’re selling today on a lease basis to do patrols for safety and industrial sites, and I’m sure almost everybody’s seen Spot Mini from Boston Dynamics. If they haven’t, they do great stunts with it, dancing, but it’s a real robot that’s deployed with very specific tasks so far. It’s not general purpose yet, but neither was the first car, and it did take a couple of decades to get the cars that were really useful, about 18 years. We are sort of halfway along that path. We’ll have a Model T moment for general-purpose robots before long because they already exist. We don’t have to guess. It’s not science fiction.

One of the things most resistant to improvement is education, training. They make that more productive. We’ve actually had the inverse of productivity, right? We don’t want fewer teachers per student, right? You want more teachers. You want the ratio to go the opposite direction of productive, right? Everybody does. Every parent. You don’t want to be in a class of 500 people. You want a two-to-one ratio if you can. One-to-one. We all obviously know that the internet makes that much easier. Physical teaching still matters, but tutoring is very powerful on this medium, the Zooming medium, and getting better all the time. And it’s getting easier to connect people with skills to people who need help, literally globally, easily, because the networks are so fast and the latencies are so low. So if one looks across the physical domains of making stuff, what you see is that in every one of them, engineers are building pre-commercial or just-commercial capabilities that are as consequential to productivity as Airbnb and Uber are to those information-mediated domains.

They are slower in being implemented because the regulations are properly more challenging. You don’t want to kill people, just to be callous. There’s a reason we have an FDA. We have safety regulations on cars and airplanes, and we can ironically accelerate that regulatory process soon because computers are on the cusp of being good enough to do not only in-silico drug discovery — instead of in-your-body drug discovery, we do preliminary discovery in computer models of cells, then we still have to test people — but we can get to the testing faster because we eliminate dangerous things easier in silico. We can test machines in silico, too, before we have to test them in real life, which is what Boeing and Sikorsky do now.

Right, well, that sort of brings us full circle to what we began with. I was talking about something that happened in Washington with policy: the passage of this infrastructure bill. From a policy standpoint, in the West, the cloud is not run by the government. There’s an aspect of freedom there. Is there a role for government either in facilitating or just not screwing it up? What are the public policy issues that you care about if you care about the cloud being useful and becoming more useful?

If you accept the possibility that we’re on the cusp of a pretty big technological boom that will drive productivity broadly and deeply in the economy, much like what happened starting around 1920, they have very similar characteristics in terms of their qualitative and quantitative architectures. Then the question would be, how do we make sure it happens and not screw it up, right? Because, I say in my book, I only deal with politics in the preface and the epilogue because I really want to build the case that something big is happening, which is the case for why we want to make sure we get it right politically in policy fields. So I know, because I was a Cold Warrior back in the day, that we can Sovietize an economy because the Soviets did it.

The great expansion of wealth that the West and the United States in particular enjoyed from 1920 to 2000 — the Soviets didn’t enjoy that. They Sovietized their economy. They’re not stupid people; they just picked the bad political system, so I worry about that to your point. I worry that we’ll go down a prescriptive path that politicians and policymakers endemically desire. They want to fix something, because there are problems; because we’re people, we create problems. Even the best meaning person creates problems. Forget malfeasance and maliciousness. People create problems because we’re human and the future is hard to predict except sort of directionally. So I do worry about that, and I do worry we’re preoccupied with the wrong problems. We’re worried about energy: That’s not a problem in my view. We’re worried about the wrong aspects. We’re worried about climate change. Sure, the climate is changing. Money, wealth, and technology deal with those problems as they happen over time.

We need money, wealth, and technology, and it’s a holy trinity — those things are related. What could kill this prosperity boom? The same thing. Punitive regulations, punitive taxation, or to put it differently, the question one would ask, not knowing what the efflorescence of technologies will allow innovators to do — I can’t know that — is: “How do I want to make sure it happens in the first place?” I want a system that rewards risk taking. I want a system that makes it possible to have access to capital, risk capital. I want a system that rewards winners when they actually win: low taxes, because it’s indisputable that overall wealth has risen but inequalities keep existing as overall wealth rises. If we focus on wealth redistribution as a policy matter instead of wealth creation, to just reduce it to simple terms, that’s destructive.

It only makes sense if we have a static economy and we want to redistribute what is — and you know better than most that in people’s heads it is — a fixed pie. It’s a pot of gold somebody found, and there’s no more gold, so we just have to make sure it’s fairly distributed. For most of human history up until the Industrial Revolution, wealth was created by taking other people’s wealth. I mean, the world had a lot of trouble becoming more productive each year. The last great productivity boom before the Industrial Revolution was in the Middle Ages. They had a productivity boom of unprecedented proportions in human history. It was based on the camshaft and gear and water wheels, and information and knowledge. The same metrics are happening today. It was a big deal. It really caused a lot of wealth growth in Europe at that time. (It was tied to warmer weather, by the way.) That was the “Little Optimum,” but that’s a whole separate discussion.

So a long answer to what worries me is, we will be preoccupied with fixing bridges the old way, by building only roads instead of making sure the world’s biggest infrastructure gets built, because the people that are building it aren’t building it with direct subsidies and mandates. They’re building it because the market wants these capabilities. If you asked a business 10 years ago, would they want to store their records or do their analysis in the cloud? Their typical reaction would be, “Look, it’s not secure. I’m worried about that. I want to have the computers on my premises. I’ll have security here. I’ll hire a security guy.” So the question you would ask somebody then, and it was an obvious one: If you run a hospital or you make tires or you produce aspirin or whatever, do you actually think your people are smarter and better computer operators than those guys at Amazon and Google? And do you actually think you’re better at security than they are? I mean, just honestly. Do you think you can hire the most talented guy? I don’t think so.

Now, you could have been reluctant in terms of its cost or its feature set or its latency, but what’s happened is businesses have discovered two things. It’s profoundly cheaper to put this in the cloud, because the utility operation of the cloud is constantly refreshing and updating the capacities and making them cheaper every day. You’re stuck with equipment you bought, first of all, and the evolutionary pace of change is so fast.

They’re much better at security than you are. Not perfect, as we all know, but they’re a lot better, and they’re much smarter at giving you more capabilities. So you could focus on what you know how to do, and we make the cloud easier and easier to use. So what has happened in the last decade is — we can see it in the data, the share of computation and the share of information storage, both those activities — the proportion that was on the end-in devices, in premises, flipped from 80-20. It’s flipping the other way, across the Rubicon in 2020. It’s a total percentage of information processing, devices versus cloud, and it’s accelerating. That’s consequential, I mean, but it’s a leading indicator.

It doesn’t change things overnight. So the first thing that happens when you give a business, any business, new capabilities that require thinking about things differently, organizing slightly differently, you don’t just do that overnight. Everybody takes weeks, sometimes months, sometimes years, because making the wrong decisions again has consequences. Real, rapid, negative bottom lines to the dollar or safety consequences. But we can see in all the data that this shift has happened in terms of where businesses want to be, and back to the point I started with: Microsoft decided to be a sort of business cloud-centered company. They had another advantage too, which is not nothing. Imagine for a moment that you do a business that competes with Amazon’s commercial business. Do you really want to put your data on the Amazon cloud?

Right.

So if you’re a retailer, maybe you think, “No, they have a firewall, and I believe their firewall is a real firewall.” I mean, the two businesses are profoundly firewalled. You could talk to people who work for Amazon off the record. But even so, psychologically, Microsoft doesn’t sell books, right? They don’t sell pharmaceuticals, but they provide the cloud service. If I were guessing today, given the appetites for “break up Big Tech,” and given the natural division of services, it would not be surprising to see Amazon become multiple companies. You could imagine it having a retail business that was legally separate from its cloud service business, and they may do that themselves at some point the way GE broke itself up because Amazon is evolving into a conglomerate. Conglomerates work for only so long, then they stop working.

So if we look back a decade from now, is your best guess that we will have indisputably seen a productivity acceleration where a key technology has been the cloud?

Yes. I mean, the short answer is: If you’re betting today based on everything I’ve learned and seen, the patterns I’m seeing, absent bizarre exogenous events — we Sovietize our economy, we have a nuclear war, I mean, gross stuff . . . (Even the pandemic won’t stop the generation of a one percentage point average higher GDP growth rate in America, which is not a crazy number. It’s way below the long-run history of productivity booms. Just that one percentage point more per year for that decade would generate tens of trillions of dollars of cumulative net new wealth in America. Money doesn’t solve all problems, but it goes a long way to ameliorating some of the most serious challenges we have.) I think we’ll look back and say the same kind of thing that people were starting to say when the internet took off early on. You and I both remember a lot of people were skeptical. “Internet, email? Big deal. Computers . . .” And most people in the general public and punditry did not anticipate the scale of the expansion of the internet, which is a component of the cloud. It’s not the cloud, and it took everybody by surprise. In hindsight, “Wow, it’s a big deal. It generated whole new industries, trillion-dollar industries that didn’t exist before.” We’re going to see that times 10 in the next decade. I’ll end with Peter Drucker’s line, who was one of the great analysts of business and of human nature. He said that he only predicted what’s already happened in forecasting, and by that, he meant he looked at patterns of things that were already underway that were firm and had high inertia and said, “That will continue.” And I think the pattern for a boom is already happening.

My guest has been Mark Mills, author of The Cloud Revolution: How the Convergence of New Technologies Will Unleash the Next Economic Boom and A Roaring 2020s. Mark, thanks for coming on the podcast.

Thanks, Jim, for having me.

James Pethokoukis is the Dewitt Wallace Fellow at the American Enterprise Institute, where he writes and edits the AEIdeas blog and hosts a weekly podcast, “Political Economy with James Pethokoukis.” Mark Mills is a senior fellow at the Manhattan Institute and the author of The Cloud Revolution: How the Convergence of New Technologies Will Unleash the Next Economic Boom and A Roaring 2020s.

The post Will cloud computing create a New Roaring ’20s? My long-read Q&A with Mark Mills appeared first on American Enterprise Institute – AEI.