Will AI power up or slow down sustainable development?

Jennifer Martin
Welcome to The Angle from T. Rowe Price where curious investors can gain an information edge on today's evolving market themes. Sharper insights on the forces shaping financial markets begin here. In this season of the Angle, we're focusing on the exciting world of artificial intelligence, otherwise known as AI. We explore its rapid rise and its potential to reshape industries, sectors and economies.

With many companies racing to acquire new AI capabilities and refine existing ones, we assess how we can all make the most of AI potential. I'm Jennifer Martin, a global equity
portfolio specialist here at T. Rowe Price Associates. And I'll be your host as we learn more about this intriguing and rapidly evolving area of the global economy and financial markets.

In this episode, we explore the role AI can play in creating a more sustainable future, but also the risks and challenges it poses. On one hand, AI has the potential to be a powerful tool when it comes to maximizing efficiency and mitigating environmental impact. On the other hand, it also raises salient questions over the displacement of established industries and jobs, not to mention the drain on energy resources, and what this means for climate change.

To discuss these dynamics and the apparent contradictions, I'm joined by Maria Elena Drew, the director of research for Responsible Investing at T. Rowe Price Associates, an
individual I have admired over the last decade, as she has shown real entrepreneurial spirit in building out the capabilities for our firm.

Well, Maria, welcome.

Maria Elena Drew
Thank you. Thanks for having me, Jennifer.

Jennifer Martin
Well, I'm so energized. This is going to be a great discussion. And let's lead off with the idea that AI is clearly here to stay. In many ways, the current technological revolution and sustainable development are intrinsically linked with investors leaning into AI advancements across different sectors while keeping a closer eye on regulatory
developments in this space. Maria, I want to talk to you about the benefits and the risks through an ESG lens. Perhaps let's start with the positives. In what ways can AI innovation help tackle climate change and other sustainable outcomes?

Maria Elena Drew
Well, AI is a tool that essentially helps us capture more information or data. It helps us
analyze that data, and it even can make decisions on that data for us. So we've had forms of AI operating for a while, but the emergence of generative AI has been a game changer as this technology allows you to delve a lot deeper and interpret complex data in much more creative ways.

So going back to your question, if we think about an ambition to limit global warming to less than 1.5 degrees Celsius. There are lots of ways that AI can help us achieve this goal. So, as you know, limiting global warming to 1.5 degrees means that the world would need to achieve the very, very formidable challenge of lowering greenhouse gas emissions by 50% by 2030 and then reaching net zero by 2050. For those that that don't know what reaching net zero means; net zero is the point at which greenhouse gas emissions are balanced by the greenhouse gases that are absorbed by natural land sinks, like forests or manmade carbon capture technologies. So, there's a mathematical formula known as the Kaya identity, and this formula states that greenhouse gas emissions can be expressed as the product of four factors: human population, GDP per capita, energy intensity, and carbon intensity.

One easy AI application is that it can help us model these four factors with much greater depth and speed than we've ever been able to before. So, if we hone in on just one of these four factors. So let's take energy intensity, you know, which can translate just to energy efficiency. So, AI technologies can help us be much more energy efficient, say, in our homes or our offices. So it can turn on your dishwasher when the grid is operating only on renewable energy. I happen to live in the U.K., where wind power is more prevalent in the evening. You know, I could see in the future maybe my appliances are going to turn on automatically in the middle of the night. So, they run on wind power and not, you know, daytime power when we might have some fossil fuels in the grid. Someone who lives in Florida might experience that at midday, because their grid is more solar powered. So it just allows all of our machines to be a lot smarter, and think for us.

Autonomous vehicles are another really good example of the benefits of AI. So with AI
powering an autonomous vehicle, they can ensure that they're operating most efficiently – you know, much more efficiently than a human would operate. They might, you know, go at the right speeds. Whereas a human might go a little bit faster than the speed limit and a little bit slower at certain times. That AI can make sure that they're using the vehicle as efficiently as possible.

There are also, you know, really innumerable social benefits that could come from AI. So, sticking with autonomous vehicles, they're safer as they can absorb more information than a human driver ever could. And also, they can react a lot faster than a human driver can. So they can reduce the number of accidents that we have on the roads. AI in general will have a lot of safety applications as that can really help to reduce or eliminate human error.

Another really positive application is in the health care space, where we see drug
companies using AI to help make new discoveries because they can just absorb a lot more data and analyze much more data to come up with new drugs and new therapies. Also, you know, we see AI applications being used in order to evaluate patient scans or other tests that they're taking, in that they can do it much faster, and they can do it with better accuracy than a doctor could.

Jennifer Martin
Those are awesome examples. I mean, using grid power, at you know, off peak levels,
electric vehicles, and then, of course, health care. And, you know, I think anyone who sees the radiologist probably appreciates having a second look by an AI assistant.

For investors trying to align themselves with the United Nations Sustainable Development Goals (otherwise known as UN SDGs), how interlinked is AI in terms of achieving, or in some cases, destabilizing these goals?

Maria Elena Drew
So we just talked through how AI can drive positive sustainable outcomes by capturing more data, analyzing that data, and making decisions on it. The flip side is that AI can also drive more negative, sustainable outcomes by capturing more data, analyzing that data and making decisions on it. So, if we look at the U.N. Sustainable Development Goals or the SDGs, these are 17 different environmental and social goals adopted by UN member states in 2015. They include things like reducing poverty, ending hunger, reducing gender, ethnic and other inequalities, creating affordable and clean energy and building sustainable cities. Each of the 17 goals have specific policy recommendations that underpin them.

At T. Rowe, we think about the SDGs as a really good framework for identifying the world's pressure points. So, if an issuer is selling a product that helps solve one of the U.N. SDGs, it's likely that they're going to have the regulatory wind at their back and they're going to have a consumer base that's there and ready and willing and wants the product that they're selling because it helps fix a problem for them. Conversely, if an issuer's product is undermining an SDG, they might face a lot more regulatory pressure and maybe less consumer demand for their product.

So one area where many people believe that AI may be a problem is with inequality. So let's start with the fact that previous technological revolutions have been all about replacing physical work with machines. Well, the information revolution is about replacing mental power with digital technologies and artificial intelligence.

Historically, every major technological revolution has created more jobs than it destroyed. But those new jobs have been created in places that were not previously predicted at the start of the technological revolution. So, for example, if we look at the auto revolution that took place. There you saw a lot of jobs displaced, but all the new jobs that were created were really around building and supporting suburban life – because all of a sudden you saw a lot of people move out of urban areas and into the suburbs. You had to build supermarkets, houses, roads, all kinds of things. So there was a huge economic boom that followed the auto revolution, but the jobs were not remotely what people would have
predicted from the auto revolution.

We see that trend take place with all of the technological revolutions that have happened previously. There was a paper by the International Monetary Fund or IMF, where they found that 40% of global employment is exposed to AI. That's a very high figure when we compare it to job disruption in previous technological revolutions. What's even scarier is that because advanced economies have more high skilled jobs in the developed world, 60% of the workforce is exposed to AI. Now, AI could potentially displace jobs in some cases, like, let's say, call center workers, but in others they could enhance productivity.

The IMF estimated that the split was about 50/50. So, if you think about it, we could be
looking at 30% of a developed economy potentially having lower or no wages, and another 30% of that same economy becoming hyper efficient and therefore making even higher wages. So without proper management from governments, we could have inequality gaps that grow at very fast rates because of AI.

So another good example of the double edged sword that AI wields is when we think about energy demand, because the data centers that power AI are extremely energy intensive. Let me just end this answer on a more positive note. So I'm going to focus on SDG2, which is Ending Hunger and SDG 15, which is Life on Land. AI could have a tremendous benefit and great applications to help solve the problems in the world's food supply. So, AI can help analyze whether environmental, commodity and other factors, so that farmers can optimize the crops that they plan to grow and also deploy, you know, precision agriculture. This would help eliminate excessive fertilizer use, which currently contributes very heavily to greenhouse gas emissions, and also help reduce biodiversity loss that we currently are experiencing through soil degradation and the ocean dead zones that are created as we have the fertilizer runoff that goes down into the rivers and then eventually down into the oceans.

Jennifer Martin
Well, thanks, Maria. You brought up some great examples and, you know, specific on energy demand – one of the stats that's been circulated is that some of these higher-
powered GPUs are the equivalent of one U.S. household in terms of electricity demand. And that's, you know, I think a statistic maybe people don't appreciate. And then when you think about the data centers that we're building, it's like we're building small cities in the middle of nowhere, but definitely drawing more on energy, which might be a great transition to the next set of questions. A ccording to the International Energy Agency, data centers and cryptocurrencies accounted for nearly 2% of global power demand in 2022. What are the ESG implications of this massive water and power consumption?

Maria Elena Drew
Great question. One and a half percent of global energy demand is coming from data centers – add cryptocurrency, that takes us to nearly 2% of global energy demand, because when we look at energy markets, it's actually very small fluctuations in supply and demand can have huge impacts economically. And so if we look at data centers themselves, they actually account for 1.5% of global energy demand. That's an IEA estimate. And the fact that, you know, we've gone from them not really being on the radar screen to all the way being one and a half percent in a very short period of time is really alarming.

A couple of weeks ago, we had a big discussion about data centers and energy demand. And one of the big takeaways from the investors was just kind of the shock that we are so, so early in AI development and all the data centers that we're going to need to power AI and the fact that it's already starting to register on the power grid is a bit of a shock. So, we can expect to see some disruption there.

If we go back to maybe the heart of your question and looking at kind of the supply demand picture, there are a couple of different factors going on. What the markets have been grappling with the last decade or so is the whole concept of energy transition. We've known that we're going to see huge changes to energy and power markets because we needed to decarbonize and we also needed to electrify the grid. So we're seeing people move from combustion engines to electric vehicles. And that means that the energy that used to be purchased at the pump or at a gas station now is coming off the power grid.

So we already knew that the power grids were going to be getting a lot more demand as we saw this shift towards electrification, which would eventually allow us to decarbonize the economy. nobody really had the fact that we would have a tremendous amount of demand coming from AI on their radar screen when they were doing all of these models. The fact that we already thought it was going to be extremely challenging for the grid to meet its decarbonization targets, it's now getting even harder if we add a whole lot more demand from AI to the equation. I do think that, you know, there are some interesting kind of things to think about here. The first is a lot of the analysis that's being done right now is really only focusing on the actual energy demand that's coming from the data centers themselves.

And we do need to think about the fact that there's going to be more energy demand needed for AI than just what's coming from the data centers. So you're going to have upstream and downstream energy demands that come into play. So, the upstream
demands could mean, you know, looking at building a lot more chips and equipment and all of the things that go into making the technologies that are going to power AI.

And from the downstream standpoint, if we look at the end user, they're going to be using, you know, much more advanced computing technology at the end of their day. So one example I like to use is my son, who obviously loves his iPad, he's 12. And in the evenings we let him go on it. And he used to listen to audiobooks and do relatively simple things.

This year, his school has released their own safe chat GPT AI for the students to use. So one of his favorite activities is talking to chat GPT, the one from his particular school, and writing various scripts and stories. So just that little example there is you have somebody who previously was not using much compute power at all, has already taken it up a notch, you know, just for their own entertainment. And if you think about amplifying that down the road, you know that that could lead to a lot more energy demand.

Another factor to keep in mind is we're still just so, so early in this stage. So, what we find within energy markets is that people learn how to be more efficient over time, and they tend to learn that very quickly, especially when the price of power goes up. And it will, because the market's going to be squeezed, as we just talked about. So I think that, you know, we shouldn't be looking at the current demand that we're seeing right now and assume that into perpetuity. We're probably going to see a lot better efficiency gains. You can see a lot of changes taking place kind of in the chip design to make them more power efficient down the road, and there'll definitely be the incentive to do that. Also setting algorithms differently so that they can be much more efficient in terms of the power that they need in order to run AI.

I think another interesting mix here is before AI really hit the scene and people started to recognize the energy crisis that might come from it, the story for nuclear didn't seem all that positive. So what we found is nuclear…is a zero carbon fuel, so it's positive from that standpoint. But obviously nuclear waste is a is a negative that comes with that technology and generally people don't like to have nuclear plants just down the road from them. So it's quite difficult to build them. I think that what we might see is, extension of existing nuclear facilities as we need to have more clean power sources available.

And also we'll have to see if that particular fuel source becomes more in favor around the world. There are places in Asia where they've been able to build nuclear at more economic costs. But if we look in Europe and the U.S., the more recent nuclear builds have been stratospherically expensive. So, it is a difficult technology to see come onstream. But I think the case for nuclear will probably change if we look at a very different picture for energy demand.

Jennifer Martin
Really helpful. Well, we've been observing that some of the hyperscalers are buying farmland in Pennsylvania and hooking up directly to the grid, which is I mean, to the
nuclear facility, and thereby staying off the grid, which is a pretty remarkable development and one that probably wasn't in people's estimates as they were thinking about demand and supply for energy.

So it's clear that while AI can effect positive change, it comes with significant climate related risks. Questions also remain over the value created by the data center boom in some parts of the world. With fears around this dystopian future of job displacement and privacy breaches, what are the key social challenges and risks?

Maria Elena Drew
I think the data center boom is an interesting one to look at. So if we look at the value of a data center, you are going to create some jobs just to build the data center itself within the region that it's in. And there will be some tax revenue that comes from that data center, but not very much. So for the most part, a data center alone is really just going to be taking more from the local economy than it's giving probably because of the power demand issues associated with it. And for the reasons why we just talked about how we're expecting that we might see a power crunch taking place as we see the build out of a lot more data centers.

And if we look at how the market is structured right now, what we see in the U.S. is a lot of the computing and kind of the intellectual capital and where the company is really making money is located in places like San Francisco or Silicon Valley or other more urban places, whereas the data center is going into other parts of the world. So you have to question kind of the value that that data center is bringing to the local community. And if local communities will want to have those data centers there. That could create a problem for companies as they do this buildout, and I think that they need to be thoughtful about that. So how do they make sure that all their stakeholders are benefiting from their presence?

Jennifer Martin
And if anyone's driven by some of these data centers, they're really, really noisy.

Maria Elena Drew
Yeah, exactly. We talked about most people don’t want to have a nuclear plant down the road. I think you’d probably rather have a data center than a nuclear plant, but it’s not the ideal neighbor for you. So that is one consideration to have. And generally, you know, they’re almost like energy tourists, these data centers! And that doesn't seem like something that's going to be a long-lasting phenomenon.

Jennifer Martin
The current technological revolution is clearly highlighting a complex balance between fast-pace innovation and job security. What can be done to mitigate the risks?

Maria Elena Drew
I think this is an area where we are really going to have to lean in and understand what's happening, because how governments and regulators react is probably going to have a pretty significant impact on economies. I think what's so exciting and frightening about the AI revolution is just the exponential growth and impact that it's going to have. We already talked about the number of jobs that are exposed to AI and in developed markets, that being 60% of jobs. That's just amazing. It's more than half of the workers of the economy are going to be impacted by this. So we need governments to be able to be in a position to help workers that end up potentially losing their jobs or going for lower paid types of jobs and helping them upskill and move into new places and really take advantage of those new opportunities that are coming.

Jennifer Martin
Just that statement, the task of writing, I guess, different regulatory pieces or laws for technology that will be used in different ways for different purposes and different
industries. Sounds really, really difficult.

Maria Elena Drew
Yeah, exactly. And the other part of it is when we think about just the use of AI itself and the fact that, you know, AI is replacing mental power, right? It's replacing humans on some levels. And what kind of decisions is that AI going to make? And are those decisions going to be beneficial for society? We do have some regulatory measures being taken around the world, but we are in just such early stages for it.

I think that everybody kind of appreciates we’re so early on the journey, it would be difficult for governments to be prepared for this. What's a little bit worrying is if we look at all of our previous regulation around technology is also a bit limited, shall we say, and maybe governments around the world have not really been in a great position to be out in front of technology as that industry has developed.

Jennifer Martin
I think you're right, lot of good examples of it's so challenging, particularly by geography, to be proactive. It's a lot easier for governments to be reactive. And I think that's what you just said.

Maria Elena Drew
Exactly. You put it better than I did. So if we look at some of the regulatory steps, we find that they're really split in that we see governments like the EU trying to take a much more centralized approach to AI, and coming up with requirements that would be used across the board. Whereas if we were to look at, say, the United States, the approach there is also very early, but it's more about setting non-binding principles and voluntary guidance and then looking more at an industry level to identify kind of what measures need to be in place.

Because if we think about maybe the regulation that you would want around a company that's actually producing a large language model, and the types of governance and oversight that they have on that, it's going to be really different than a company that's just using AI models. And just to maybe bring home some of the impacts that could happen is, you know, we talked about the fact that AI can analyze huge amounts of data and then make decisions or recommendations based on that data.

Imagine they're going through and analyzing credit scores and credit data, and they're then making recommendations on who's eligible for a credit card or a bank loan. You can imagine how that can start to create some issues with inequality pretty quickly, if you don't have proper governance in place or proper human oversight on the decisions that are being made.

Jennifer Martin
Yeah, you just highlighted how these AI systems in some ways are reflecting a combination of, and producing, social relations and understandings of the world that may or may not be correct. And the other thing that while you were talking, particularly about geographical regulation, brought up a memory of a conversation I recently had at a U.S. pension plan where they're already considering asset allocation decisions related to AI and regulatory, which I thought was really an interesting intersection of asset allocation discussion.

Thank you, Maria, for sharing your thoughts on the fascinating links between AI and sustainable outcomes.

Maria Elena Drew
Thank you. It was a great discussion.

Jennifer Martin
Well, it sounds like regulation will play a crucial role as AI and other technological advancements continue to evolve. If I were to summarize today's discussion, the key takeaways for me were that there is no doubt that the fast-paced technological
advancements we've seen from generative AI will help in meeting ESG goals.

Certainly, AI innovation will help in developing cutting edge energy transition solutions.
However, the rise of AI also brings key issues to the fore, such as significant job
displacement and lack of economic value in the regions where energy hungry data centers operate. Yes, new jobs will be created as part of this revolution, but whether these opportunities are suitable to the existing skillset of the unemployed remains to be seen.

This could be where company regulations can take center stage.

Thank you, Maria, for sharing your thoughts on these fascinating links between AI and sustainable outcomes. And thank you for listening to The Angle. We look forward to your company on future episodes. You can find more information on artificial intelligence on our website. Please rate us and subscribe wherever you get your podcasts.

Disclaimers
This podcast episode was recorded in April 2024. This podcast is for general information and educational purposes only, and outside the United States is intended for investment professional use only.

It does not constitute a distribution, offer, invitation, recommendation, or solicitation to
sell or buy any securities in any jurisdiction, or to conduct any particular investment activity.

This podcast does not provide investment advice or recommendations, nor is it intended to serve as the primary basis for an investment decision. Prospective investors are recommended to seek independent legal, financial and tax advice before making any investment decision.

The views contained herein are those of the speakers as of the date of the recording and are subject to change without notice. These views may differ from those of other T. Rowe Price companies and/or associates. Information is based upon sources we consider to be reliable; we do not, however, guarantee accuracy.

Investing in technology stocks entails specific risks, including the potential for wide variations in performance and usually wide price swings, up and down. Technology companies can be affected by, among other things, intense competition, government regulation, earnings disappointments, dependency on patent protection and rapid obsolescence of products and services due to technological innovations or changing consumer preferences.

International investments can be riskier than U.S. investments due to the adverse effects of currency exchange rates, differences in market structure and liquidity, as well as specific country, regional, and economic developments.

The acronym ESG refers to environmental, social, and governance.

The source for the statement about 40% of global employment being exposed to AI is the International Monetary Fund's Paper, titled: 'AI Will Transform the Global Economy. Let's Make Sure it Benefits Humanity, January 14 2024.

The source for the statement about data centers and cryptocurrencies accounting for nearly 2% of global power demand in 2022 is the International Energy Agency, Electricity 2024, Analysis and Forecast to 2026.

The source for the statement about 60% of jobs in developed markets being exposed to AI is the International Monetary Fund's Paper, titled: 'AI Will Transform the Global Economy. Let's Make Sure it Benefits Humanity, January 14 2024.

This podcast is copyright 2024 by T. Rowe Price

ID0006880
202405-3502280