The Environmental Impacts of Digital Technologies, with George Kamiya

Notable Quotes

• The diverse emissions of digital technologies: “We can think about how energy is used as listeners are listening to this podcast. First, there's energy that's used at the data center. These are large warehouses that basically have a lot of computers that store things like this podcast recording. Then we need to use energy to transmit that data for the recording, from the data center to the listener. It might be going through fiber optic cables, routers, or maybe mobile network towers. Finally, there's electricity that's used by the user. Maybe they're listening to it on their smartphone or a laptop. So there's electricity at each step of that chain, and of course there's energy use to manufacture all those pieces of equipment … It's a big area.” (3:07).
• Impacts of the ICT sector should be interpreted in context: “It's dangerous when we try to focus too much on these new emerging technologies, and it distracts us from the innovation that we really need to [happen in] the big sectors. Not to say that the ICT sector doesn't warrant attention and efforts to reduce emissions, but we really need to cut emissions in power, in industry, in buildings, and in transport … Emissions and energy use of the ICT sector are significant … but we can't let that distract us.” (23:42)
• The future of digital emissions is hard to predict: “Things change really quickly. So, Bitcoin is something that emerged in a couple of years very quickly and suddenly is consuming 0.2–0.3 percent of global electricity use. It's really hard to predict how the energy use of ICT is going to evolve over the next 10 to 20 years … I think companies and governments need to really think about investing in R&D to figure out newer technologies that will get us to even higher efficiencies, as the demand for all these services is just going to continue to go up.” (25:07)

Top of the Stack

• “Factcheck: What is the carbon footprint of streaming video on Netflix?” by George Kamiya
• “Calculate your emissions” data tool, from the International Energy Agency
• “Bitcoin energy use—mined the gap” by George Kamiya
• “Digitalisation and Energy” from the International Energy Agency
• Carbon Brief newsletter;
• The Citizen’s Guide to Climate Success by Mark Jaccard
• “The AI Revolution: The Road of Superintelligence" by Tim Urban
• “Tackling Climate Change with Machine Learning” interactive summary

The Full Transcript

Kristin Hayes: Welcome to Resources Radio, a weekly podcast from Resources for the Future. I'm your host Kristin Hayes. My guest today is George Kamiya, digital energy analyst at the International Energy Agency (IEA) in Paris. George leads the IEA's analysis on the energy use of digital technologies and coordinates cross-agency efforts on tracking clean energy progress, digitalization, and automated and shared mobility. Today, George and I will be talking about the environmental footprint of the many digital technologies that have now become fixtures of many of our lives. We'll focus on energy consumption—George's primary expertise—but also touch base on a few broader issues and impacts. And we'll also talk briefly about some of the technologies that most of us probably don't use in our day-to-day lives like Bitcoin mining, but which experts are keeping an eye on nonetheless. Stay with us.

George, it's a pleasure to have you on Resources Radio. Thank you very much for joining me today.

George Kamiya: Thanks for having me, Kristin. It's great to be here.

Kristin Hayes: Great. So, for my usual introduction, I'd love to ask you to start by telling our listeners about you and about how you became interested in issues at this intersection of energy, climate, and of course the digital technology world.

George Kamiya: So I guess I stumbled into this topic accidentally. So I'd been working at the IEA for a couple of years on climate change policy, and our executive director came to us and said, "What's this digitalization topic and buzzword I'm hearing about?" And he asked me and my boss to look into the topic. Then we ended up writing a report that we released in 2017, looking at how digital technologies were impacting the energy system: everything from heavy industry to buildings, transport, electricity, oil, and gas. And in that report, I was fortunate enough to work with one of the top experts, Eric Masanet, on the chapter that looked at the energy use of information and communications technology (ICT). I guess I'm new to the topic, but I've been fortunate enough to learn from some of the leading experts who've been working for decades on this topic. It's fortunate for me because I've always been interested in technology and I've always been interested in the environment and climate change energy. So it's fun to bring all of it into one thing right now.

Kristin Hayes: That's great. That's great. Well, this is a big ask I realize, but it would be great if you could set the stage for us just a little bit more in terms of the universe of information and communications technology, or ICT. So what that universe that we'll be talking about today looks like—can you talk through that ecosystem for just a second?

George Kamiya: Yeah, of course. So when you talk about the ICT sector, when we're talking about energy in the ICT sector, we're really talking three main areas: data centers, data transmission networks, and devices. That all sounds abstract. So maybe we can think about how energy is used as listeners are listening to this podcast. So first, there's energy that's used at the data center. These are large server halls, warehouses, that basically have a lot of computers and storage that stores things like this podcast recording. Then we need to use energy to transmit that data for the recording, from the data center to the listener. So it might be going through fiber optic cables, routers, maybe mobile network towers.

And then finally there's electricity that's used by the user. Maybe they're listening to it on their smartphone or a laptop. So there's electricity at each step of that chain. And of course there's energy use to manufacture all those pieces of equipment. So all of the storage equipment, routing equipment, computers, phones, and of course at the end of the life of those equipment, there's going to be disposal and or recycling. So yeah, it's a big area. I think it's best to think about these three main buckets, I guess.

Kristin Hayes: Great. Okay. That's a really helpful grounding. And so with that background, let's talk about which parts of that ecosystem consume the most energy. I've certainly seen headlines about how data centers in particular are gobbling up energy, including a suggestion that they could consume up to a fifth of the world's electricity by 2025, which would of course be tremendously impactful on the energy system. Can I ask you how much energy is consumed by data centers globally, and how has that changed over time? And maybe a follow up is: does the picture of energy consumption look quite that stark?

George Kamiya: So I guess the good news is I'm pretty confident that data centers won't be consuming a fifth of electricity in just five years. So to put that into context, that'd be 5,000 terawatts a year, which is more than the entire United States today. The media seems to love these mind-boggling numbers as headlines, but, fortunately or unfortunately, the facts of the reality aren’t as exciting. We're seeing these headlines every few months and it's been around for a long time. There was this 1999 Forbes article that I dug up that one of my colleagues Jon Koomey refers to often, where this article was saying that the internet could consume half of the US electricity over the next decade. So that's in 1999. So by 2010, the internet could consume half of the U.S electricity system. It turns out that was way off. It was more like 1.5 to 2 percent.

Kristin Hayes: Okay. It's a big difference. Yeah.

George Kamiya: So there's been lots of these headlines and it's hard to follow and it's hard to keep track of all the headlines. The good news is that based on the best available research we have and the most recent research we have, data centers are consuming about 200 terawatt hours last year. This is around 0.8 percent of global electricity use, so nothing close to a fifth or even a fiftieth. And the good news is that it's been pretty flat at around 200 terawatt hours since 2010. So over that time, over the last decade, internet traffic has grown about 12 and a half times, and data center workloads-which is a measure of data center demand—has grown by seven and a half times. Despite all those growing very quickly, fortunately overall data center electricity use has been pretty flat.

And that number 200 terawatt hours can appear big or small depending on how you compare it. Right? So there might be a headline saying, “Oh, it's equivalent to the entire consumption of a medium sized European country or a big US state.” But if we compare it to some other end uses, like electric vehicles right now are consuming about 80 terawatt hours a year. So just to put that into context, space cooling consumes about 2000 terawatt hours a year right now. It really depends how you compare those, but yeah, 200 terawatt hours is about just likely less than 1 percent for data centers right now.

Kristin Hayes: So George, how did we get here from a point where the expectation was of considerable electricity consumption to a relatively modest amount? How did we manage to reduce that consumption compared to what we may have expected?

George Kamiya: Yeah, so it's a really impressive energy efficiency story, and there's really two main drivers here. One is that the efficiency of computing hardware has improved very rapidly. There's this trend called Koomey's law, which describes how the efficiency of computing has doubled roughly every two to three years. This is a very rapid rate of efficiency improvement. Then the other main driver is this huge shift going from smaller, traditional, so-called “enterprise” or corporate data centers. These are something that a medium-sized company or governments might have, and we're seeing more and more of those data center services outsource to a cloud in hyperscale data centers—which are much larger, much more efficient. I've heard upwards of 10 to a hundred times more efficient, and these are run by companies like Google, Amazon, and Microsoft. So it's really these two trends, both at the hardware level and then more broadly shifting the type of data center from a very inefficient traditional system to a much larger, much more efficient system.

Kristin Hayes: Interesting. Okay. All right. Good to know. So I want to move from that large-scale data center part of the ecosystem to something that probably is all too familiar to many of us these days. That’s streaming video, another thing that's shown up in the headlines particularly in this year where so many of us have been at home and it's only risen in popularity. There are similar headlines about the amount of energy we consume from our evenings of “Netflix and chill.” I cribbed that from one of your presentation titles, but it works very well here. So can you talk a little bit about that particular element just as another case study and how much energy goes into those nights of movies streaming?

George Kamiya: Yeah. Some good news for all the climate-conscious binge watchers out there. Streaming video is not that bad compared to everyday activities that we do. So, there were a few headlines in the media—I think late last year—saying that watching a half hour Netflix show emits as much CO₂ as driving four miles, so about 1.6 kilograms, which just out of hand seemed way too high to me. So I dug into the numbers, dug into the assumptions, and then published this fact check on Carbon Brief.

And it turns out the numbers that the headlines were quoting were about 30 to 60 times too high. So instead, four miles watching a half-an-hour show on Netflix is more like driving a few hundred feet or a city block—so not four miles. There were a number of problems with this study. One of them was just using very out-of-date assumptions. And like I said, the energy efficiency is improving so quickly, both on the data center side and at the data transmission side, that even using numbers from four or five years ago can have a huge impact on the overall numbers.

Kristin Hayes: Interesting. And I think we're going to talk about this a little bit later, but the question of what's actually driving that efficiency. Essentially, you referred to the law that drives that, but are there other factors at play, other corporate commitments, other investments? Is it in their best interest to improve that efficiency? Yeah, so let's come back to that. But I'm happy to know that my catching up on Hamilton and seasons of Game of Thrones is not ruining the planet. That is great. So what are the factors that go into play in terms of putting together an analysis of the energy consumption associated with streaming?

George Kamiya: Yeah, so it really depends on a couple of factors. It really depends what kind of screen you're watching on. So, whether you’re watching on a large television or a smartphone, your network connection, and the resolution you're streaming. So to take an example, if you're watching on a 50 inch LCD TV, that television is consuming about a hundred times more electricity than a smartphone. Similarly, if you watched in 4K Ultra High Definition that uses about twice as much data as HD, and if you're on your smartphone, the connection really matters. So using your 4G mobile connection would use about four times as much electricity as through wifi. And then of course, for carbon emissions, it really depends on the grid mix, right? For instance, I live in France so that half hour show, we were talking about emitting something like four grams of CO₂ because our electricity system is a very low carbon.

So I need to watch something like 30,000 hours of Netflix. So 24 hours a day for something like three years to emit the same amount of carbon emissions that I would emit on a round-trip flight from Paris to New York. So just to give consumers a better sense of how their choices affect their overall emissions, I've republished that commentary on the IEA website. And there's an interactive tool where people can enter the number of hours they've streamed, is it on a TV or a laptop and those kinds of things. And just to get a sense of, how does it compare to boiling a kettle—which is something British researchers have looked into actually—driving 10 kilometers, and then of course the round trip flight from Paris to New York.

So I think maybe the main takeaway here with the article is that watching a video has pretty low emissions compared to other everyday activities, especially driving. So long story short, you can ignore those headlines saying that the two hour Netflix binge is worse for emissions than 15 miles of driving, or another one I found was how cat videos could cause a climate change nightmare.

Kristin Hayes: Well, we're not getting rid of cat videos. So I'm glad to hear that they're not as bad as we might anticipate. So that's actually a great lead-in to my next question, where you mentioned that the grid matters a lot here too, and there is this connection between the amount of electricity for consuming and of course the resulting greenhouse gas emissions, but what is generating that electricity really matters. So what else do we know about the electricity that's fueling all of this growth in demand for internet services. The growth and the services aren't necessarily mirrored by growth in electricity demand, but nonetheless, that has to come from someplace. So what do we know about the electricity that's fueling all of this internet consumption?

George Kamiya: So again, I guess I have a bit of good news in that. Some of the big data center operators are also some of the world leaders in corporate renewables procurement. So you might've seen coverage around this, but one data point is that over the past five years, these ICT companies have accounted for about half of global corporate renewables procurement. And just last year, the top four corporate off-takers for renewables were all ICT companies. And Google alone signed 2.7 gigawatts worth of power purchase agreements (PPAs). That's one side of the story. The other is that typically these big data center operators will try to locate in regions that actually already have low carbon electricity. Of course, the electricity consumption is so high that it's one of the main costs for them. So it makes sense to try to locate in regions that are stable, that have typically cheaper electricity, which is usually low carbon in colder climate.

I see more and more big data centers popping up in Denmark or Sweden. So on the renewable side, it's looking good, but of course, even if they were to buy the equivalent amount of renewable PPAs to match their consumption, that doesn't necessarily mean that every kilowatt-hour of electricity used by that data center is powered by renewables, right? So Google and Apple have achieved this kind of a hundred percent matching over the last couple of years, but they still need to match going forward, and match for time and geography. Of course, wind is generally more plentiful at night, and solar more plentiful during the day. Just because you have a PPA doesn't mean that it's matching the data center energy demand. So I think these companies are thinking more and more about how to reduce their greenhouse gases (GHG) impact.

So, Google's announced this 24/7 carbon-free electricity strategy where they're now trying to think about the grid electricity, their generation or their PPAs, and figuring out how they can manage their demand. So how can they, for instance, forecast the wind that's going to come online so that they can shift their data center demand or workload that can take advantage of that lower GHG grid? And I think other companies are looking at the renewables in their supply chain. I'm really happy to see that it's going beyond these renewables commitments. More and more companies are signing on and they're getting closer to a hundred percent, but it's nice to also see companies pushing the envelope and to realize that they can do more for the system. And then the other thing is the ICT sector as a whole committed to a science-based target earlier this year. So, they've committed to reducing their emissions by 45 percent between 2020 and 2030. And I can see certain governments starting to act on this issue as well. So hopefully in the next months and years, we'll see more and more commitments and actions ramping up.

Kristin Hayes: Yeah. Just two quick follow-up questions on that then. So I just want to confirm that PPAs are power purchase agreements, is that right?

George Kamiya: Yep. Yeah.

Kristin Hayes: Okay. So you referenced a number of commitments to wind and solar and dealing with the intermittency of those sources. Are there other, what I suppose I would consider clean sources, if not completely renewable sources, that can hold onto some of that clean electricity that these companies are also investing in?

George Kamiya: Yeah. That's a great question. I don't know the specifics of the company strategies, but I can imagine that energy storage would be a key part of their energy portfolio. So the reliability and resilience of these data centers is extremely important. Right now, a lot of these data centers have big diesel generators for backups. So if they can instead have battery storage or longer-term storage that can displace any diesel consumption they might have while also maybe even participating in the grid. That could be a really interesting proposition.

Kristin Hayes: Well, we've talked mostly as I anticipated we would about energy consumption and associated emissions, but I feel like I'd be remiss if I didn't at least ask you about other types of environmental impacts that might be driven by all of this ICT infrastructure as well. And I guess I'm thinking about things like water consumption, land use impacts, and biodiversity disruption. I realize these are a little bit out of your mainstay, but I would love to ask if you are aware of other perhaps non-energy related concerns that the sector is grappling with as well?

George Kamiya: Yeah. So let's start with water use. So I think the big data center operators are starting to become more transparent and report around their usage and improvements around water use in their sustainability reports. The other big impact that I think is worth looking at is materials extraction, electronic waste. So I think it's fair to say that in most places in the world, we need to do a much better job on electronics recycling. So to reduce both the extraction that's coming out of the ground and then to reduce the toxic waste that we might be putting into our landfills. And of course, one way to do this is to use the equipment that we have for a bit longer. Taking an example of a smartphone, about three quarters of the environmental impacts come from production and disposal. So not in the use phase where the consumer is actually using it. Just by using a phone for a year longer or two years longer, we can really spread out those life cycle impacts.

Kristin Hayes: Great. So now I want to turn to the uses of the internet that are probably less familiar to many of us. We talked primarily about things—streaming, regular data usage—that are quite familiar, but there are a number of high energy consumption activities that are beyond the realm of most of our day-to-day reality. And in particular, two that have come across my radar are this concept of Bitcoin mining, really could use a primer in that particular topic, and then also the growth in use of artificial intelligence and the amount of data processing and associated energy consumption that might go along with that. So can we talk a bit more about those things and what can you tell us about energy consumptions and trends for what I'll refer to as those advanced computing activities?

George Kamiya: Yeah. So again, it's like the headlines that we saw with streaming video, with data centers, there were a number of headlines around Bitcoin emissions or emissions coming from Bitcoin mining in the last couple of years, especially as we saw Bitcoin prices going through the roof. I think it was late 2018, maybe. So there's one article in Nature Climate Change for example that said Bitcoin emissions alone could push global warming above two degrees Celsius. So in response to some of these headlines, I did another fact check reviewing: why does Bitcoin use energy? How does it use energy? What are the efficiency trends? What is the global electricity consumption look like? So I published a review last year in the summer. It turns out Bitcoin actually uses about 40 to 80 terawatt hours a year, last year and this year. So that's around 0.2 to 0.3 percent of electricity use. So again, it's like if you compared it to a country that would be like a smaller, medium-sized European country, which is large, but it really depends what you're comparing it to, to give it more context.

And then in terms of emissions, a lot of the Bitcoin mining is actually happening in China historically. So they seem to be aggregating towards regions that have excess renewables production. So we could use a global average grid to estimate GHG emissions, but it actually might be lower than using a global average grid. So in terms of emissions, Bitcoin emissions definitely won't be pushing us above two degrees Celsius. If that's the case, we've probably messed up in other sectors like transport, electricity, or industry.

I think it's dangerous when we try to focus too much on these new emerging technologies, and it distracts us from the policy and technology and innovation that we really need to pay attention to the big sectors, right? Not to say that the ICT sector doesn't warrant attention and efforts to reduce emissions, but we really need to cut emissions in power, in industry, and in buildings and transport. So it's a tricky communications exercise. It's important to highlight that emissions and energy use of the ICT sector are significant in some cases they're growing, but we can't let that distract us from the bigger fish in the ocean that we need to deal with.

Kristin Hayes: And I do think that clearly every sector that has associated emissions has some role to play in reducing those emissions. But given that so much of this is driven by electricity consumption as you point out, a lot of the solutions here are in fact one step further back in the chain and about cleaning up that power sector. So yeah, that's an important reminder that there are foundational building blocks to a lot of sectors’ decarbonization efforts that start in other places.

George Kamiya: So I hope it's coming across that things change really quickly. So Bitcoin, that's something that emerged in a couple of years very quickly and suddenly is consuming 0.2, 0.3 percent of global electricity use. So it's really, really hard to predict how the energy use of ICT is going to evolve over the next 10 to 20 years. Things like blockchain, artificial intelligence, 5G, these things might shape new applications and new demands in ICT that we don't even know about. The other thing is that efficiency gains that I mentioned earlier are starting to slow. And at some point, we're going to run out of efficiency opportunities, right?

So I think companies and governments need to really think about investing in R&D to figure out newer technologies that will get us to even higher efficiencies as the demand for all these services is just going to continue to go up. It's really the three prongs, I think we need to continue to push efficiency as we have—more and more renewables to power the remaining energy that's needed. And then this next generation technology that's much more efficient.

Kristin Hayes: Okay. Well, as the title of the podcast noted, this conversation is fundamentally about the environmental impacts of the ICT sector. So we've talked a little bit what I would characterize as the negatives, but I get the sense that there are also some real potential positives. So I want to close the substantive part of our conversation, with an optimistic view and get your reflections on how some of these digital technologies can actually help reduce emissions.

George Kamiya: Yeah. Thanks so much for bringing this up. So, in that 2017 report I mentioned on digitalization energy, we tried to highlight some of these real opportunities that digital technologies could and are having in reducing emissions in industry and transport, buildings, and the electricity system. And I think through this COVID crisis, we're really seeing some of these real world examples of how ICT and digital technologies can reduce emissions. For instance, we're on a video conference now that can cut business travel flights. There's a lot more working from home now, which means less commuting and energy use. In fact, earlier this year, a couple colleagues and I at the IEA did some analysis around: what if everyone who could work from home in the world did work from home once a week?

And we found that we could see significant energy savings and emission savings just by cutting the commute. But of course, there's going to be some increase in home energy use from cooling and heating, but I think it's interesting to think about these scenarios. It's really the question where some of these ICT technologies could have very large rebound effects or unintended effects. So let's take the example of working from home. I think if it's once a week, maybe people's general behaviors are not going to change, but if I could work most days of the week from home, I might be willing to move further and further away from cities because maybe I can live in a bigger home, which actually might require more energy. I might end up driving more in general. So, there are these direct effects and then indirect effects and then more structural, long-term effects that I think could go both ways, right?

So it could be good or could make things worse. Same for things like e-commerce making it easier to purchase online. Maybe it's cutting down on the travel that we do to go to the store, but maybe we end up consuming more, and the same goes for the streaming videos. If we were renting it from Blockbuster, there'd be so much more effort to rent that DVD that we're not going to be watching the amount that we're bingeing today. And then finally, another area that I'm particularly interested is this future of automated and shared mobility. So, a couple of years ago, there was a lot of hype around autonomous cars and then this very large uncertainty about whether they would help cut emissions because maybe they'd be shared.

We could reduce private ownership of vehicles, reduce parking spaces, kind of like this optimistic perspective. And then there's a very pessimistic perspective where maybe everyone's just going to be living further away from home, being chauffeured in their autonomous car for longer distances and willing to take that extra travel time. We really don't know. And I think maybe the way to approach this as not so much predicting what could happen or what people think will happen, but rather thinking about what are the important policy leavers that we have to get to a better world where we have lower emissions, lower energy use. And that gets to what I think I'd like to be my final point around digital technology is that digital technology is not a silver bullet for climate change.

So in my opinion, digital technology is really climate agnostic. It could be applied to reduce emissions, but it could just as easily be used to increase emissions. For instance, we could use it to better forecast wind, but at the same time we could use sensors and AI to extend the lifetime of coal plants or make it even cheaper to extract oil and gas. So it really depends on the real world policies that I think shape where these technologies are applied and where it makes business sense to you. Yeah, I think I'll close on that final substantive thought.

Kristin Hayes: Well, that was a great final substantive thought. Thank you very much for bringing the conversation full circle in such a helpful way. So George, it's been fascinating. I really appreciate your time on this topic, it's so relevant to so many of us. So I will close with our regular feature, “Top of the Stack,” and I'd love to ask you to recommend more good content of any type to our listeners. So George, let me turn it to you. What's on the top of your stack?

George Kamiya: Okay. So first I'll say a Carbon Brief. It's so hard to follow all the climate and energy news out there. And I find they provide a really good newsletter that summarizes the important issues and news of the day or the week. And of course, they publish my commentary. The other thing I've just started reading is a book from one of my grad school advisors, Mark Jaccard. So it's called The Citizen's Guide to Climate Success: Overcoming Myths that Hinder Progress, and it attempts to debunk some of the common myths around climate policy and actions. So I think that's worth a read and there are some nice New Yorker-style cartoons in there about climate and energy from one of my labmates from grad school.

And then the third one is not really energy and climate related, but there's this long form stick-figure illustrated blog called “Wait But Why.” It covers everything from climate change to procrastination, how to pick a life partner, AI revolution, a really wide range of topics. And it's really well-researched. There's a couple on climate, so that might be worth a read.

Kristin Hayes: Well, that's a great recommendation, the diversity of topics we all need these days. Well, George, thank you again for coming on the podcast. I really appreciate your time and the information you shared. And I look forward to staying in touch.

George Kamiya: Thanks so much, Kristin.

Kristin Hayes: You've been listening to Resources Radio. Thanks for tuning in. If you have a minute, we'd really appreciate you leaving us a rating or a comment on your podcast platform of choice. Also, feel free to send us your suggestions for future episodes. Resources Radio is a podcast from Resources for the Future. RFF is an independent, nonprofit research institution in Washington, D.C. Our mission is to improve environmental, energy, and natural resource decisions through impartial economic research and policy engagement. Learn more about us at rff.org.

The views expressed on this podcast are solely those of the participants. They do not necessarily represent the views of Resources for the Future, which does not take institutional positions on public policies. Resources Radio is produced by Elizabeth Wason, with music by Daniel Raimi. Join us next week for another episode.