The Two Sides Of Technology’s Environmental Impact | Forrester Podcast
- Hi, I'm Jennifer Isabella. - And I'm Stephanie Volores. - Your co-host for Forrester's podcast. What it means where we explore the latest market dynamics impacting executives and their customers. Today, we're joined by VP and research director, Glenn O'Donnell and analyst Abhijit Sunil, to debate the impact of emerging technology on sustainability efforts. Welcome both. - Thank you. Great to be here.
- Thank you. - So naturally there's two sides to the story of emerging tech and sustainability, and we love a healthy debate around here at Forrester, but before we get into debating the benefits and the risks of some of these emerging technologies, tell us why this issue is such a hot topic amongst clients right now and tech leaders in general. - Organizations across the board in every industry are taking action that puts them on a positive path towards climate action. Many organizations in the Fortune 200 in our survey, we found have committed to a net zero or carbon neutrality pathway, and to enable companies to do this they're exploring technologies, especially emerging technologies and trends that help in assessing and reduce carbon emissions in their scope one, scope two and scope three. To define what they are quickly. Scope one emissions are direct emissions that an organization has from their premises.
Scope two emissions are emissions through purchase electricity and scope three emissions are indirect emissions in their up and downstream value chain. And we see emerging technologies have a role to play in carbon emission reduction across the board for organizations. - Yeah, and our report on the Jekyll and Hide report highlights the fact that there are two sides of the coin here. There is,
of course, the, you know, the good side where, you know, technologies are helping the planet, but, you know, technology has an impact too, and that impact can be negative. So, you know, these two forces balance each other and hopefully balance each other to the positive, they don't always and we'll talk more about that. - Yeah. I mean, that's interesting, cause I think whenever people talk about emerging tech, there's always this tech optimism, which is tech is gonna solve all our problems. It can solve all of our climate change issues. It's gonna help us reduce our carbon footprint.
And then no one really looks at the downstream impact of the technology itself as an actual generator of carbon footprint. So let's actually take one category of emerging tech and maybe some people consider it emerging. Maybe it's just tech in general, which is semiconductors and processors. And I guess there is some advances there that would put it back into the emerging tech category and maybe it's the, how processors are being just deployed in everything today. Just about every device.
Every object has a processor in it as we begin to network things into smart infrastructure. Let's start there. What other benefits on reducing carbon footprint and then what are gonna be the, you know, the downsides that people aren't taking into account. - Chips are central and a backbone to almost every technology and the way semiconductor chips are manufactured and used will have impacts in every industry. Now processes and chips have long been looked into for power optimization, but now more so than ever before and power management functions in semiconductor chips are now becoming more dynamic and examples include turning off circuitry when it's not needed, embedding systems that earlier used to exist in embedded systems into broader computing systems as well.
We are also seeing the emergence of application specific circuits that are used for not only particular applications for broader use cases. We are also seeing the advent of power optimized architectures like risk, the most notable being the advent of Apple's M1 chip, according to apple switching to Apple's M1 chip for the Mac mini reduce the carbon footprint by as much as 34%. - Yeah. And that's notable because the so-called risk architecture reduced instruction set computer that's behind the apple processor. It uses less circuitry to accomplish much of what it needs to do. And these are more efficient chips.
So that is a positive development and others have to follow suit. And, you know, this topic of semiconductors is a good place to start this discussion because everything else, all these other emerging technologies, whether we're talking blockchain or AI or IOT, you name it, it's all built on semiconductors. And this is why this is such an important thing. These things have been technically around since the 1940s, but we're entering new phase where everything, you know, everything you can think of now has a chip in it.
And, you know, that means, you know, lots more chips and you know, it has all those benefits that Abhijit was mentioning that, you know, it has the ability to have more control over power consumption, but there is a downside of semiconductors and the most notable one is these things are environmentally abusive to make they require an intense amount of electricity, electrical power, a lot of water, you know, this is not simple stuff. This is high impact manufacturing. And you know, the manufacturers are taking a lot of measures to minimize the impact. You know, a lot of things like solar energy and other renewable energy sources are helping and water reclamation to save water, you know, they're recycling water at an astounding rate, but it's still, you know, we can't ignore the fact that this stuff has a pretty heavy environmental footprint. And the other thing that comes out of this is that we have to watch is more chips means burning more power in general, you know, chips consume electricity and more chips means consuming more electricity.
So the power footprint of all of this extra Silicon that's floating around the world, you know, we can't ignore that. - Yeah. It seems to me, based on the discussion that you guys have is there's some advances that will make the chips themselves be much more power efficient, more dynamic, like Abhijit mentioned, you could turn off circuitry when it's not needed, but at the end of the day, all you're doing is making this slightly less evil or slightly less bad. I mean, to your point, Glen, the manufacturing process itself consumes a lot of energy, it may or may not be clean or renewable.
You know, you have to consume elements and resources. It takes a lot of water. You can get more resource efficient, you can get more water efficient, you can use renewable energy in the manufacturing process, but at the end of the day, this is still generating carbon footprint. All you can do is actually just make it much more efficient.
So to me, this one doesn't sound like a solution at all. If anything, it's just, you're minimizing the impact that you can. And to some extent in the tech industry, we just have no choice if really we wanna have, you know, everything that this will enable smart devices, the smart infrastructure, smart cities, you name it. This is at the core of making things, "smart" - Is it a necessary evil then? - Yeah, I think it kind of is. And, you know, maybe we don't look at it so much as evil. I mean, yes, I think the net impact here is negative on the environment, but, in all the use cases that are emerging and many of them are good use cases and we'll talk more about that later, but you know, smart, everything is inevitable.
And let's look at some of the other developments that are going on. Everybody gets excited about electric cars, you know, as opposed to fossil fuel burning cars. Well, electric cars need a lot of chips and a lot of power, and that does not come as a freebie.
So I think yes, overall negative, but necessary. - Yeah. That's a good way to think of it as the necessary evil.
But so I guess to some extent then too, when chip manufacturers talk about making them more power efficient, that's good and we should celebrate that. But sometimes I become wary where they make it sound like it's become some sort of solution. This is a necessary evil that you have to reduce the impact of. Right. So we've talked about semiconductors. Let's talk about another emerging tech that again, techno optimist love to celebrate, which is machine learning and AI, you know, which is everyone is excited about because it leads to more insight, faster decisions. It can power genomic research, financial risk modeling, weather prediction, everybody loves AI and machine learning.
So let's start about the specific benefits that can be applied to sustainability and climate action. - Sure. And you're right, Stephanie, in the way AI and ML is talked about at this time, they're talked about heavily in terms of predictions and optimization and in the report, we call out how AIML has broad applications and energy use optimization.
We are already seeing AI based solutions in predictive power usage, monitoring in the data center, in predictive maintenance for equipment, maintenance and security purposes and even creating sustainability benchmarks. So if excess energy is produced in the grid and AI engine can learn that and a machine learning engine can then direct that excess energy into where power is needed. So some use cases like these have already emerged, and we are seeing, for example, well, we saw the example of Google partnering with DeepMind AI and implementing a dual network in the data center, which eventually enabled them to predict the amount of energy that is used for cooling, according to the amount of power that is needed in data halls and overall reduce their energy usage by 40% driving down their PUE by 15% in the data center, which is power usage effectiveness. And we are seeing AI's application in a variety of other use cases, including in predicting, as you mentioned, the weather or therefore forest fires. And we are also seeing how in manufacturing organizations, especially in the energy and utilities and oil and gas industries use AI engines to look into safe operations and predictive maintenance of equipment.
So AIML will have many more use cases as we look into the next decade - All right, I'll buy that. And of course there is the AI that's used in climate risk analytics itself, which is, you know, a lot of the impact of climate change we won't be able to reverse, even if we went to zero emissions tomorrow. So it's dealing with the impact of that, like understanding sea level rise, understanding changes in weather patterns, droughts, wildfires, like you mentioned, et cetera. - I think that's where we're gonna see some of the real benefits here is the ability to make smarter decisions about what we do as a human race.
You know, a lot of the reasons we're in this pickle with sustainability and environment is, you know, we just made a lot of dumb decisions as a species and if we make better decisions going forward, we'll be much better off and I think that's the real benefit that we get out of this kind of stuff. - Okay. My techno optimist give me the bad news. - Well, I'll give you the bad news. Doing this kind of processing takes an awful lot of compute power. You know, the most common processor used for this kind of compute is the GPU, the graphical processing unit that, you know, gamers use these things to, you know, to render on their screens, aliens that you wanna blow up and things like that. But it turns out that that architecture works really well for the type of machine learning models that we're doing.
The problem is these things burn a lot of power. They burn so hot that you can't put your finger on it. You know, if you open up a computer that has these things in, there's a big fan on it, and a big heat sink to draw the heat away. What that's telling you is, you know, for all of the electricity going into that processor, it's generating more heat than it is doing compute. So what we need to do is make those processors again, coming back to semiconductors, we need to make those processors more efficient because all the heat that you're getting off of there is wasted energy. That's not doing any good.
So the more of that we can use the better, but in the meantime, doing this kind of work takes an awful lot of compute power. In fact, the data that we've been looking at is startling. You know, much of this is going on in data centers, whether they're corporate data centers or the Cloud data centers and data centers use about 1% of the world's electricity. And you think, oh, 1%, that's no big deal. For just one particular use, that's a lot that's insanely high.
So all of the data center providers, whether it's, you know, Cloud or again, corporate data centers, one of the big pushes here is to just make them more efficient and not generate as much heat because this does generate a lot of the CO2 equivalent. You know, we saw data that says, you know, a lot of these large AI models can emit five times the lifetime emissions of the American car. That's mind boggling.
So we have to get better at this. And I think if we make that more efficient, we'll be better off. But again, this is not something we can ignore.
- Yeah. And I think that data center use at 1%, I've heard one to 2% of world's electricity, I think that's conservative. And that's just today that doesn't take into account growth trajectory over the next five to 10 years either. But I guess, let me throw it back to you. And I tend to be, you know, more techno dystopian, but in this case, I do have some questions though, which is, you know, you mentioned hyperscalers and Cloud providers.
You know, a lot of them actually do claim that they're running their data centers completely on renewable energy. So if you have a data center that is running on completely renewable energy, does that offset a lot of what we're talking about here? And then, you know, you have data centers that run renewable. You also have data centers that can optimize their geographic location to reduce power and cooling requirements. I mean, you've seen, you know, data centers and Nordic regions, couple of the high hyperscalers talk about actually building a data center under water, to reduce power and cooling costs. So in this case, like, can you truly offset the carbon footprint of both the energy requirements, as well as the energy that you need for power and for cooling in this scenario? - So some of these specialized applications that requires AIML, the amount of energy that AIML power applications require are the reason basically why these cutting edge research and development efforts in the data center is happening in the first place.
So when I recently spoke with a service provider who develops AIML machine learning related applications in the healthcare space, they mentioned that some of these algorithms are so cutting edge and at the same time are so impactful that the notion of sustainability is not even being talked about yet. One example is the use of machine learning in looking at scans for patients in radiology and the applications themselves are no doubt, much more energy intensive computing intensive than a regular computer application so to speak is. However, the impact of the application and the use case itself is so cutting edge that the notion of optimization is not in the conversation yet. However, other applications that may be more mature, for example, face recognition on devices and user devices like phones, have come along in their maturity so much that we are now able to optimize that for more than actually just thinking about implementing it. - Yeah. I guess the question is like, regardless of how compute intensive the actual model is, if you're making the argument that the data center is running on renewable energy, and it's also maximizing its power efficiency, maximizing cooling efficiency, again, could you argue that the net impact on carbon is minimal? - Well, it will be minimal if the data center does all of these things, but it's not all data centers are able to do all of these measures.
Now, this is one big reason why we are seeing a lot of enterprises explore the public Cloud model or turning towards larger co-location players who are able to put into place some of these measures within their data centers surely because of their scale and because they are able to put into place certain measures like this, the example we talked about earlier of Google experimenting with an AI engine that will predict energy usage in data halls is one such example. Smaller data centers that are operated by enterprises on-premises may not have enough scale to experiment with cutting edge technology to lower their PUEs, or to have their data center run all on renewable energy or to make partnerships like that with utility firms. - Yeah. My guess though, is if you don't have the kind of heft to be negotiating with partners or to demand renewable energy in your data center, you're also not the one who's investing in some seriously cutting edge AI either. So I don't know for me,
I'm gonna go with good for this one. - I think potentially very good. Yes, potentially good. Because again, it's gonna help us make better decisions. In fact, one of the problems we have with this is people just willy-nilly apply machine learning to anything.
So making better decisions is going to help that, or should we do it this way? Ironically, it's going to take machine learning to help us make those decisions. - So let's move on to the next emerging tech. Let's talk about edge computing and IOT, and, you know, it's interesting this discussion that we've had so far, I mean, we've talked about these emerging technologies as distinct, but to some extent they're all incredibly interrelated and even build on each other.
We talked about semiconductors and we talked about AI. But really, those two things are actually what you need for all the kinds of edge computing and IOT use cases that people think about. You think about what's being done in agriculture and increasing crop yield. You think about what's being done in manufacturing, in smart cities, in smart infrastructure transportation. There's a lot of exciting things that are actually enabled by every device having a chip in it and being able to distribute the processing closer to where the data and the intelligence is needed. You know, there's some exciting use cases for this tech specifically from a sustainability perspective, I'm curious Abhijit, what do you see as any possible benefits of this? - And yeah, you are right about edge and IOT being the example of how all of these emerging technologies are interrelated to each other.
We spoke earlier about chips and how chips are a cold part of IOT devices. And even the chip shortage that we've experienced over the past couple of years have had major implications on the availability of IOT devices and in the edge computing space as at Forrester, we've written about. Now edge and IOT, we say have active roles to play in data gathering and optimization of networks.
And that's the core of how edge and IOT can help with sustainability. IOT devices can be embedded in remote locations to gather critical climate data and edge can help process this data where they originate and thus reducing loads on networks, as well as emissions back at larger data centers. Edge computing can help enable some of the AIML use cases that we talked about earlier, right at the edge of the ecosystem of a technology organization to analyze this data and take action at the edge without having to transfer data back to Cloud location or to a larger data center. So edge and IOT use cases have relevance across scope 1, 2, 3 emission reductions, and especially in scope three emission reduction where edge or IOT devices can help with monitoring and penetrating as deeply into the supply chain as possible to gather data about, for example, in one use case, we saw Chiquita explore AR/VR to illustrate to customers sustainability journey of banana. So in agriculture we use cases like that, and in other use cases such as in fleet management, smart cities, smart lighting, smart building solutions even.
So we'll see the emergence of edge and IOT services as a major enabler for sustainability in the coming years. - So then Glen, what are some of the downsides that people don't think about? - Well, as again, we make everything smart. We need more of that instrumentation out there, more chips, more devices, and more devices means more power consumption.
Now, granted, we talked earlier about the chips for AI burn, an enormous amount of power. For the most part IOT devices burn very, very little power, but where you may have one processor for AI, you might have a hundred or a thousand IOT devices out there? So what is the aggregate of all of those things? That's the big question. In fact, we're trying to quantify that now. It's not an easy thing to do, but you know, lots and lots and lots of little things add up to a big thing and that cannot be ignored.
But I think the bigger issue here is, you know, in the consumer world, we tend to be, you know, everything disposable. So if I get a nest doorbell and okay, well that thing has a certain lifespan to it. At the end of its lifespan. I throw it in the trash.
That's the typical mode of doing things. And, you know, all of this electronic waste is gonna be a big problem if we don't manage it. And we also have in a commercial setting, especially industrial settings in a longer lifespans. And what that means is, you know, we in the tech space, we're used to throwing things away after two years. You know, if you've got a mobile phone, that's two years old, it's ancient.
In the industrial world, we have machines that sit there for 20, 30 years doing the same thing over and over again every day. And as we make that smarter, the technology that makes it smarter is going to also be there for 20, 30 years. And after a certain amount of time, this stuff just does not take advantage of newer technologies that can be more efficient. So there's a lot of waste in older electronic devices. And we see that across the board. You know, many of our industrial machines have programmable logic controllers that are 20 or 30 years old, literally, and they're not power efficient.
So we need to manage that. And as we develop the products that are going to go into these situations, plan that the maintenance is going to take care of such things. So, yeah, I think, again, it comes down to decisions.
Edge and IOT allow us to make smarter decisions, but in the process, maybe some of those decisions we're making are not so smart after all. - So I'm curious then to offset some of that, I mean, I can recognize the massive waste problem. I mean, we're talking thousands, tens of thousands of devices and sensors.
I don't know if this is an easy fix per se, but is this where we need much more robust kind of e-waste programs and recycling programs and refurbishment, not just throwing stuff away, but refurbishing them and using it again? - Yeah. We, as a society have to get a lot smarter about that and it has to become part of our daily lives and recycling anything it's a pain, you know, taking your plastic bottles out to the recycling center is an inconvenience, but we do it because it's a good thing for the planet. Not everybody adopts that mentality. So we have to make it part of the social fabric that, all right, anything that has a battery or a plug at some point it's gonna die and you don't want to just throw it in the trash, or if you do, then that means the trash companies and waste management and everybody like that, as this stuff comes in, they're going to have to have some way to sift through all of that garbage and find the stuff that needs to be recycled or refurbished or whatever. It's a big problem, but there are solution if we, as a society step up. - I feel with this one that it's almost neutral.
There's so many use cases for this technology that will make our world more efficient and enable significant advances in transportation and agriculture, you name it. But then at the same time, more devices, more manufacturing, more energy consumption. So without a corresponding advances in renewable energy, e-waste management or recycling or refurbishment, you know, in best case, it washes out, but without it, it could actually have a worst footprint for the world and for the environment. - Yeah. I would agree with that. I think it is pretty much a wash.
It gives us a lot of feedback, but there aren't downsides. And let's also remember a lot of this stuff is battery powered, you know, think of a smoke detector. Now you buy a smoke detector, it has a 10 year battery in it.
You don't have to replace the battery, but after 10 years, that whole thing is shot. So how do you dispose off that? A battery is not a happy thing in the environment, and you know, the tiny little bit of radioactive material that detects the smoke is also not something you want going into a landfill. - It's interesting. So all these emerging technologies, the categories that we've talked about, some of them have, I think, a more compelling case that they're on balance more good than harmful to the environment, but I think what I've come away from all of this is it's not that easy to put them in a simple bad or good category that the approach is actually to consider the total impact and to try to minimize it. And there's very few things that aren't going to have some kind of impact.
So what is the best plan to balance all the positives and negatives as you start to consider deploying this emerging tech? - It's important for us to discuss the negative or risks associated with emerging technologies so that these risks are not in the blind spots as we adopt and as we see the rise of these technologies. We discussed edge computing and the risks it poses for e-waste management. It also disseminates the carbon footprint of an organization to the edges, so it's harder to monitor and put into place the types of measures we might otherwise put into a data center at the edges, because we may not be able to put a AI engine for predictive energy measurement in an edge data center, small and moderate data center, for example. So the takeaway is for us to think about use cases backwards. So as a technology leader who is looking into emerging technologies and how we can help in their sustainability strategy in climate action and mitigation climate action and climate change adaption, what types of use cases would these emerging technologies fit into? Like for example, the adoption of edge and IOT, or how will you include sustainability related efforts into every digital transformation and then work backwards into finding the right balance as Glen often puts it into a quality locks space of the benefits that technology can provide versus the risks that it can pose.
Simple example is the use of AIML. Is there an alternative to using such an AIML engine for something and perhaps a less carbon intensive alternative. And in the report we explore other technologies as well, such as blockchain too, which also has similar sites to the story. And the important takeaway is to think about use cases backwards and to see if there is a real need for using technologies that may be more compute and carbon intensive than others.
- Yeah. It's important that we achieve that balance and balance is something that seems to be foreign in the society we live in today. But you know, let's not be Pollyanna about this stuff.
Everybody gets excited about things like electric cars and all that, okay? It's good, but there's a downside to it too. And you have to account for all of these factors. This is not a panacea. It is not, you know, the apocalypse. There is a middle ground and let's find that middle ground and just make the right decisions as you're doing all of this stuff. These technologies are coming, whether we like it or not, let's just be smarter about how we're going to deploy them.
- Thank you both for joining us today. - Thank you - Thank you. - If you like what you heard today, subscribe to Forresters, what it means podcast on Apple podcasts, Google podcasts, or your favorite podcast player.
To continue the conversation, follow Forster on Twitter, Instagram, and LinkedIn, or drop us a note at email@example.com. Thanks for listening.