Inside Amazon and Google's Vision to Power AI Sustainably | Dreamforce 2025

You. It's a pleasure to be with all of you today. My name is Sonia Norman, SVP of Impact here at Salesforce and we have two incredible guests and friends of Salesforce. I cannot wait for you to hear from them. We're of course all here at Dreamforce hearing about AI and we all understand the potential, the transformative nature of AI and, and as sustainability professionals, we can't escape also the reality of the environmental impact of AI, the energy consumption, natural resource use. Today, thrilled to have two leaders at the absolute cutting edge and forefront of this work. Chief sustainability officers from Google and Amazon. Truly, when we think about who to follow, who to look to for best, best in class examples, think of you. So thank you and welcome to Dreamforce. Would love to just dive right in. A question for both of you and Carol, have you start. You know, AI is powering incredible innovation across industries, uses a lot of energy and resources, as I just said. How are you thinking about the environmental impacts of AI as we're scaling, you know, generative AI a couple of years, you know, years old at this point. Where are you focusing? How are you prioritizing?

Yeah, well, super glad to be here. Thanks for having me. And it's always fun to be with Kate. We are thinking about AI as something that obviously is running across every single part of our business. And really, I mean, the big question is of course that we get over and over around energy use of AI. But I think the question we want to be asking is how can we use AI as a transformative tool for our business? And in particular from my lens, when we think about sustainability, we are really thinking about it as how can we particularly use sustainability and AI together to really tackle the problems that we're seeing in climate change, in what we're doing in human rights, even what we're doing on circularity issues, on biodiversity, on waste, on water. So we're using AI in every part of our business, in every part of these industries that we're in, and we're thinking about those solutions at scale. So first and foremost, we're building more efficiently and we're utilizing AI to help us run more efficiently. We're looking at our energy portfolio and utilizing AI to think about where can we be more efficient. We're utilizing it in the ways that we procure energy, we're using it in how we build. We're using it. And I can. We'll talk about specific examples later. But even in circularity examples in our materials procurement, we think about packaging, so we're using it in all aspects of our business to become more efficient. And then we're thinking about investing in new technologies. And so it's become really for us kind of a all in conversation. And we're really excited about the opportunities. It's not to say that there aren't certainly challenges when we think about that kind of scale across the big conglomerate nature of our business. But the other thing I want to talk about is also, we'll get into specifics on this, but how we're using it also in the communities where we live and we work operate because we are using it in disaster relief and we're using it to work with our nonprofit partners on some interesting, you know, ocean cleanup aspects and things like that. So for us, AIR is becoming a tool that we will see the increase in energy use, but we will also see huge opportunity.

I love that. And one of the things I want to underline is just your business scale is truly mind blowing. You just mentioned issues across practically every sector. So as a chief sustainability officer for Amazon, you're really diving into every single environmental and social issue, as you mentioned. And I love how you're mentioning the interconnectedness of the environmental and social issues because they really go hand in hand and you can't address any of these issues in isolation. Kate, different business also. Technology leader. How are you thinking about the environmental impacts of AI? Where are you all focusing?

Yeah, thank you, Sunia, it's great to be here with you and Kara, always wonderful to share the stage. Indeed, we at Google believe that AI is an incredibly powerful technology that can help solve some of the world's greatest challenges, whether that's enabling the growth of the economy, scientific breakthroughs, and indeed challenges like decarbonizing energy transportation, enabling us to deal with crisis resilience that's coming from extreme weather events. So we see this as a tremendously powerful tool, very much grounded in Google's mission, which is to organize the world's information, to make it universally accessible and useful, and to do that in a way that is as helpful to as many people as possible, and indeed rightfully so. People want to understand better how are we delivering this technology responsibly and efficiently. So we're looking at two big tools we have in our toolbox. The first is full stack efficiency from how we're doing AI hardware. We have a Google chip, our TPU, which is now in its seventh generation. That chip is 30x more efficient than the first generation in 2018. So we're seeing major efficiency gains in hardware as well as in Software, we're deploying solutions like quantization with how we're training our models. That's leading to 39% efficiency and training. And then furthermore, we need to deploy reliable, clean, affordable electricity. And so we have been very bullish on advanced energy technologies like advanced nuclear. And we've had first of a kind agreements with companies like Kairos Power who are deploying small modular reactors. We're even working on fusion energy deployment, long duration storage. So indeed we see also advanced energy as being really critical to that responsible delivery of AI.

Yeah, but this isn't new for you. All right, you've been on this renewable energy, clean energy, low carbon energy, 24,7 low carbon energy for a long time. Talk to us a little bit about where you started with that journey and where does AI take you as you think about the energy consumption piece.

Yes, indeed, we, we have a moonshot around 24 by 7 carbon free energy. And what that means is that we have been laser focused on this clean, reliable, affordable energy deployment journey for many years, going all the way back to 2010. So today where you find us is we have 170 projects around the world for about 22 gigawatts of power. And just last year we brought 8 gigawatts of power onto contract. And just to give you a sense back to your question around kind of the puts and takes of this conversation, that's four times more electricity than our load grow in that same year. So very important that we keep focusing on not only those clean energy generation sources and bringing on those next generations of technologies that I was mentioning, but also that we unlock the grid for everyone. And so that's technologies like advanced reconductor and advanced grid technologies. We recently partnered with a company called CTC Global where they're able to put new wires and existing power lines to double the throughput just a couple of months. Way less cost than building new infrastructure. And indeed, we're also looking at how can AI powered tools help unlock the grid. We have a moonshot for the grid at Alphabet called Tapestry. And what Tapestry is doing is working with grid operators like PJM in the mid Atlantic, helping them to speed up their interconnection queue. You might hear about this a lot. There's a lot of power that's trying to come online, but it's caught in interconnection cues. So we're building AI tools that can help free up and unlock the grid.

Thank you so much. And I know, Kara, you could also tell us for hours how Amazon's thinking about Renewable energy. But I want to switch gears a little bit to water. So can you explain for folks in the audience who might not live and breathe data centers and compute where water fits into the picture? And then how are you thinking about your water strategy?

Yeah, I think one of the things, I mean we think about resource consumption overall. I mean, you've talked a lot about energy and one of the things we know is, is energy intensive. And you know, I just, I do want to add, I think we're, we are also invested in the clean energy story. I mean, we have 600 projects globally. We reached 100% renewable in 2023. But we are investing in nuclear as well. We have four nuclear projects and small modular with our investment in energy. And I think these technologies that we're trying to accelerate will be part of the solution. But when you think about natural resources and water scarcity, you think about both kind of on the same, become efficient first. So yes, build better chips. We have Tranium and we have Gravitron and other things like that that are going to help us to build more efficiently. Water is something where you want to think about where you're siting data centers, how you're building data centers and using water most efficiently. So we publish water efficiency metrics, we are publishing our building standards and how we're building more efficiently, not only our data centers, but also in our entire fulfillment network. But we also set out to be water positive by 2030 for our entire AWS business. So that means that we're going to return more water to the communities where we operate by 2030 for AWS business and for our entire Amazon business across our entire global operations in India, we're going to be water positive by 2027 because we know that water is such an issue in that. And so I think when we think about those things, we're investing in water replenishment projects, we're working with the communities to understand those kinds of issues and really thinking about, again with the advent of AI, what's going to take more cooling, Investing in liquid cooling technologies and building better but, but also investing in those replenishment projects and understanding the issues of the communities face because we have such a scarce resource in that we're also looking at how AI can teach us more about where water leakages are happening because that's a critical issue. So much like Kate's talking about the infrastructure and you know, we need to modernize the grid and we need to use IoT and we need to use grid, modernizing technologies on the grid we need to understand our water infrastructure is antiquated and there's a lot of leakage happening. And so I can help us to address some of that. And we were working with some farmers in Mississippi and looking at grid, you know, looking at AI and how it helped us to save water leakage there. But we also have a facility in Glasgow where we used AI and we built something called Flow Ms. And we're saving 9 million gallons annually of water there by using AI and helping us to do leak detection. So these kinds of technologies again can help us to be more efficient with the water use in addition to building water replenishment projects. And then the third thing I'll mention is we launched sort of a center for best practice in AI and water usage. Most recently we announced it in the New York Climate Week. And this is in partnership with a number of utilities and University of Pennsylvania and Amazon together. And what we want to do is we want to encourage everybody who's working in the space of water and AI to put forward best practices and how we're going to measure what's going to happen in this sort of nexus of water and AI to put out all the information that we can so that we can accelerate faster. We think that any of this information that we're sharing, that when we work together and we put information out collectively is going to help us move faster.

Thank you. And I see that day to day you know, working in sustainability. I see the smaller companies and even large scale companies who are less mature in their sustainability journeys, looking to companies like ours to see what's possible and making those connections to prove out a concept at scale. You all have such tremendous scale is very powerful to show what positive innovation looks like, what partnerships can accomplish. So you already both started to share some of what's working, how AI is being used to address sustainability challenges. Kate, come back to you.

Yeah.

What's maybe one fun additional example that's inspiring you on what's working? What has feasibility to really create tremendous impact at scale?

Yes, indeed. I mean, I think the good news is we're seeing a lot of examples of this. So first I would point to one thing I hear a lot from sustainability professionals is a lot of challenge around data and especially supply chain data. This is a pain point I know that we all share and we've been seeing tremendous opportunity for AI tools to really unlock both data as well as to do better prediction. And a really nice example of this is a partnership that our Google Cloud team for forged with Unido and Tracemark to build a platform for coffee growers to enable greater transparency in that supply chain. And so this platform is doing a variety of things and I think this could be an example for a lot of other commodities as well. It's putting better digital tools in the hands of farmers so they can control their data and have better digital access. It's then also enabling those in their supply chain to get better transparency into their data. Then it's helping with EUDR compliance, with regulatory compliance that's coming, that's requiring companies to do more reporting. And then also we're seeing that weather is such a powerful tool. Our AI models are making weather so much more both able to be forecast in the immediate term with now casting and then getting you to longer term forecasts. This is also very powerful in the agricultural space where you can use that information to understand when to plant your crops, when to spread fertilizer, which also has both cost saving benefits as well as environmental benefits. So that's one place we're seeing AI tools unlocking a lot of opportunities for sustainability teams. Another example I would share that's shared by many in our space is reporting. In terms of how do you prepare your annual report? Probably you both have teams who do this. I know I do. And my team has really been looking at how do we use AI tools to not only enable us to generate the report to do that data collection, but also to make information more accessible to people. So, you know, if you, if you go on our website, you'll see not only do we have a traditional PDF of the report, but now you can go and interact with the report in natural language through NotebookLM, or we can give you a podcast that's generated about the report. So I think for those of us in the sustainability space, I often hear from CSOs, how do we better use AI in our, in our work? Those are a couple of examples I see as really powerful.

Yes, you are singing from the Data Nerd Songbook. For those of you who have checked out Tableau Play plus this week, we are talking about this all the time. How do you have conversation with your data? I think old school was you have this kind of static dashboard that shows you your rear view mirror, maybe sustainability data from a year ago, business data that's no longer relevant, frankly for management decisions. But now having a conversation be able to pull out those insights actually change how you do business is so powerful. And we're also so, so inspired by what's happening in ag. And one of the things that I'm most Inspired about is the use of SMS and WhatsApp. So, you know, these AI tools, these large language models, doesn't just have to be in a digital website or, you know, in the enterprise context within Slack or a product like that, or Google Cloud where I use it. A lot can happen at the farm level with people who have no technology skills beyond just keeping in touch with their families. Really incredibly powerful now also with real time translation, just an unlock when you are a multinational company or a multinational NGO working across Africa and Latin America, but being able to seamlessly get best practices to those constituents. Kara, same question for you. You've added already some great proof points. What's a recent example of something that's getting you inspired about how AI is going to change the world?

Yeah, there's so many already and I think you're making great points about meeting people where they're at. The teams that can focus on data interoperability and harmonization as well are able now to spend less time collecting information and more time just doing the work of decarbonization, which is what we all want to be spending our time on. But some of the things that, you know, that we're doing, they seem simple, but they can have huge impact, especially across scale, like we have at Amazon. So I'll give you a few examples both inside and outside of our company. One of the things that people talk to us a lot about is our packaging. And it's only 3 to 4% of our footprint. So it's not something that has the most impact in terms of from a footprint perspective.

It's visible, right?

It's very visible though, to our customers and we care a lot about it. So we've used a couple of different AI tools. One is a packaging decision engine and one is called private investigator. And both of these things are within our fulfillment network. The packaging decision engine helps our employees to make the right decision about which packaging to put onto a product. And that has saved over 4 million metric tons of packaging. So dematerialization of our packaging since 2015. And so taking that packaging out means when you get something from us, then shipped to you at Amazon, it's going to be shipped in the right size container, less packaging material for you to recycle and, and it still will be protected. The other one is about finding. It's an AI tool that helps us to find product that's already damaged as it flows through the network so we can pull that product before it ships. So what that means is one, a customer doesn't receive a damaged product. But from a sustainability perspective, what that means is we don't have to ship a damaged product and then either get the reverse logistics of getting that product back, which of course has a carbon impact of having trucks on the road and. Or get that product, you know, that maybe dispose of in a way that we don't want to a landfill or something like that. So there's a lot of sustainability impact in getting that product out of the system and having us take care of it in the right disposition channel in the right way, or maybe repairing it or fixing it before we ship it to a customer. So those kind of AI tools, they're great customer experiences and they have sustainability impact. We also have a tool that we're using actually in our business in India right now, which is around detecting food before it expires. And so when we look at it's an AI tool that allows us to see when fresh food is approaching its expiration date. And we're able to look at it, see if we should pull it from shelves and either sell it right away or again donate it and get it out into the community to people who can use it so that we reduce food waste. So those kinds of tools, really important for us, for our community and for the people who work for us, you can feel proud about. We're utilizing and handling our product correctly in the community. We have a couple examples. One is with Hurricane Helene last fall, we were able to have an AI tool. It hit, you know, in Tennessee and in areas where there was a large amount, sort of 28 square miles, where there's a large amount of damage, happened very, very quickly when a hurricane hits. And so we were able to create and use drone mapping with AI to map this sort of 28 square mile and immediately look at about 32,000 sort of images and analyze them very, very quickly with, with AI and do two things. One is to tell the first responders where they were needed most. And so to point their energy to the places their help was needed the most. But also importantly to tell them where there were dangerous areas to protect them from going into areas that were areas they could get in trouble in. So those kinds of things where we're able to use AI in disaster relief, we have a very large logistics business and obviously we utilize that to help in disasters all over the world. So our disaster relief team is super interested in how we can apply AI when we go in and we help with disaster relief and we help with our already what we're doing in logistics. And that helped us really, you know, save the lives of first responders. It helped us get help to the community in a much quicker way. So those kind of things are very, very interesting to us too, to think about how we apply our knowledge and our technology to the, the community. We've also worked with a group called Ocean Cleanup. So there's a great Pacific Garbage Patch, which many of you probably know about. And we've used AI to help them map that and figure out again from a resource efficiency perspective where, where are the biggest sort of floating pieces of plastic in that patch and where to direct their resources. So we've done that same kind of mapping of imagery and help them to point to where they need to go to right away. So I think those kinds of applications when we work with our nonprofit partners are really interesting in addition to what we're doing internally. And I could go on about different examples that we've had in sustainability around again, more utilization of our own operational efficiencies. But I think those are a couple where we're seeing some good customer benefit, community benefit, and I think sustainability impact.

Yeah, I love that so much. What really those examples call out for me is going again from reactive to proactive. So that example of getting the damaged product proactively out of the system or proactively serving information to your partners that they need to be successful. And I love what you underlined around like it's a better customer experience, it drives top line growth, it drives customer success scores. And that's ultimately, as Chief Sustainability Officer is what we're looking for, right that Win Win for Community, Planet and business. Kate, how do you think about that? Win Win? I mean I think about even just Google Maps, we don't inherently think of it as an AI product, but I know that you all route us more efficiently, which ultimately uses less gas, less pure emissions, things like that. But where have you had some of those winds of it's better for your customers, but ultimately better for sustainability outcomes?

Yeah, absolutely. And I really think if you look at a lot of the analysis that's out there of the size of this potential, both from a decarbonization perspective, what I can enable, but also crisis resilience. And I really want to come to that because that's another really powerful area. But you may have seen there was a paper published in the journal Nature not long ago by Lord Nick Stern and several other collaborators. And what they showed was, look, by 2035, so only 10 years from now, we could see up to 5 gigatons of emissions reductions that were enabled through AI. And what we're seeing at Google is just 5 of our products have already enabled 26 million metric tons of emissions. And just to kind of put that into perspective, enable them to be reduced, that is. And just to put that into perspective, that's more than double our own operational footprint. Wow. So we're doing that work through a variety of solutions. Several of them are oriented around transportation. So you mentioned fuel efficient routing in Google Maps.

Many of you may have used Google.

Maps to get here. We're using AI capabilities in the background to calculate, if it's not already the fastest route, what's going to be the one that's going to save you on fuel. And that's enabled us to reduce a significant amount of cost for our users as well as to reduce their emissions footprint. Similarly, we've worked with cities on something called Project Green Light, where we're partnering with cities to enable them to better sequence their traffic lights. If you live here in San Francisco, that's giving you that green wave effect like you get on Franklin Street. I love that when I'm late for dinner. And what that's enabling is 30% less starting and stopping events. So better experience as a driver, but also a 10% reduction in emissions. And what we know is actually air quality at stoplights is 20, 29x worse than on the open road. So both really good user experience benefits as well as sustainability benefits.

Less wear and tear on the vehicle as well.

Exactly. And then also to come to crisis resilience, which I think is another very powerful use case for AI and a really important one. As we all know, we're seeing seeing so many more extreme weather events in our daily lives. One area is flooding. Riverine flooding is something that impacts hundreds of millions of people around the world. And what our Google research team did is they built a new global hydrologic model that enables us to predict riverine flooding seven days in advance.

Wow.

So then we can send alerts to communities. We've partnered with the World Meteorological Organization to make sure we're getting that information out to communities so that people can get out of harm's way. And we're even now increasingly seeing aid organizations giving aid in advance of those crisis moments so people are better able to respond and prepare. Another critical one that I know is really personal for a lot of us who live here in the Bay Area is wildfire. We've teamed up with the Moore foundation and other partners to help launch a co constellation of satellites called firesat. And what firesat will enable us to do is detect wildfire ignition in 20 minute increments when wildfires are just about the size of this stage, like the 10 by 10 size that will be so powerful both in terms of getting people out of harm's way, reducing damage, and also reducing all the emissions that are caused by wildfire. So we're seeing both in the climate mitigation space and terms of ways we can decarbonize transportation, energy. We talked about AG earlier, but also in crisis resilience, just such powerful AI applications.

I love that so much. And it's so inspiring that both of your companies are so focused on disaster response and using your core competencies to look at those areas. So in our audience, primarily technologists, business users, folks who are likely not thinking about sustainability every moment of every day as we get the luxury to what would you say to, to our audience members as individual professionals, but also to their companies. If they're just starting out in their sustainable AI and efficiency innovation journey, what ways do they have to amplify and to accelerate what their companies might be trying to get off the ground? How could they be great customers? Many of our attendees are probably customers of yours. We're great customers of both of you. How can they help us do more? You know, this is a shared challenge, humanity's greatest challenges. Things like disaster, climate, et cetera. We all want to do better, faster. Where would you start, Kara?

Well, I think there's a lot that we can do in shared data and accelerating technology and sharing solutions. So there are things that we all just need to know about the products and services and how our companies operate, that some of these are just for business purposes and then some of them have a sustainability benefit or even purpose. So a couple other things that I'll point out. I mean we have a lot of reasons that we need to measure our cost, carbon impact, our waste, our water impact. These things have a business purpose as well. For cost, for business continuity purposes, for understanding where our business challenges lie, for understanding our products and services. So for example, we have a large devices business, we have a large retail business. We make a ton of private brand product. We do product LCA work, life cycle assessment work. Many of you will do this in your business, you know, whether or not you're in a sustainability role. These are costly to do. An lca, they previously take, you know, many, many months to roll out and they're done by almost every business who makes a product. As an example, in the past, these would take months and months, if not a year to do. They would Cost thousands of dollars, you would hire a third party firm. We've developed with AI a rapid LCA tool. We've done 4,000 LCAs in the last quarter.

Wow.

So we've taken it down now to where we can do one in about 15 minutes. So that LCA tool is going to completely change the game. So things like that, even if it's not, you don't think about it as a sustainability.

Knowing your product mix, knowing your inventory.

There'S so many ways you can build product. What we're doing with, with our mapping of our footprint, this is good information for our businesses to have to fully understand. I mean carbon and cost are good friends. So we've completely digitized our carbon footprint and we have a framework that we use internally where we sort of talk about digitizing and democratizing and delivering our information around with AI. And we want to think about this in ways that really give our business owners a lens into their business so that they can make good decisions. And when I think about sustainability, it's really one factor along with cost, continuity, you know, how they're operating, what safety, all these other things that business owners and operators have to think about every single day. And our tech folks are working on these solutions and now utilizing AI not only to go faster but really to disrupt the processes of how we operate. So when we think about where can you get started? There are so many places to intervene and disrupt the business and to think about how we move faster. And there are sustainability, what's good about it? A lot of it is pre competitive collaboration so we can share how we work. We launched something called the Sustainability Exchange where we open source a lot was formerly proprietary information. We've got 40 other companies who have put information on there. We have the Climate Pledge where we have 600 companies signed up to that with us who are committed to being net zero carbon by 2040. There are a lot of collective efforts out there where companies are coming together and sharing information in order to go faster. So I think you'll find a very open community, whether it's open, sourcing data, sharing data online, sharing information. And you can use sustainability as a way in to a lot of technologies, innovation and find a really welcoming community.

Thank you. We're at time. I could talk to you ladies all day. Thank you again for being lighthouse examples in the technology sector. Globally you are so inspiring. As individuals we always say the place to get started is harnessing your superpower for good. Whatever you do, whoever you are. And I see you as the best examples of harnessing your personal superpowers and your business superpowers. So thank you.

Thank you.

And you, too.