for more of this energy—hungry infrastructure. also on the show, we're going to get the view from the us tech giant ibm. yeah, they've seen their demand for al skyrocket. so what are they doing to cut down on their carbon emissions? wherever you'rejoining me from around the world, once again, a big hello and a warm welcome to the show. artificial intelligence. yep, ai. it is everywhere nowadays. and the big tech companies like google, meta, microsoft, and now apple, they're muscling in to be at the head of the ai race, unveiling ever more sophisticated chat bots and other machine learning tools. but one aspect of the ai race that's perhaps not talked about as much is the impact on the environment. these language models, like chatgpt, are some of the most energy—guzzling technologies. they need vast data centres to store the amount of data required to train the chat bots. so let me just show you some of the numbers. emissions from global cloud computing account for around 2.1% to 3.9% of all global greenhouse emissions. that's more than the emissions from global aviation. but that number — it's growing by up to 36% every year due to the energy required to run ai tasks. and that means that, by 2027, according to researchers at vu amsterdam university, ai could consume up to 134 terawatt hours each year. what's that? well, let me put it this way — that's about as much as the yearly energy needed to run a small country like the netherlands. and that's a lot of energy consumption. so what are the big tech companies saying about all of this? well, databricks, it's a multinational data analytics and ai company. and they've told us that demand for data storage has grown massively in the last few years. since generative ai hit mainstream sort of accessibility for people, our entire uptake of that has gone through the roof. um, before chatgpt went live, we had about 400, 500 use cases doing some of the most boring ai type of things in production. um, using some of the generative capabilities. but in the last 18 months, that'sjust gone through the roof. as a company, we've always looked to squeeze out the marginal cost of using compute. we really try to be incredibly efficient with the compute and the resources used. well, meanwhile, microsoft has admitted in a report last month that its energy consumption has gone up by 30% since 2020 due to its push to ai. and according to an independent study published last year, if google were to switch its whole search engine to ai, it could consume as much energy as the whole of ireland in three years�* time if nothing is done to mitigate it. 0k, we're going to hear more in a moment about what the big tech companies are doing to help bring carbon emissions down. but first, let's delve a bit deeper into al's carbon footprint. my first guest is an expert on helping digital industries make their activities more sustainable. dr david pugh, great to have you on the show. and, david, can we start with this? can you give us a sense of the environmental impact of ai and...and also, how do you actually quantify it? yeah. so ai obviously will play a very key role in getting the...the... the world to net zero, whether it's in looking at soil management, whether it's in enabling energy flexibility, but there is a carbon cost for running ai and digital technology in general. 0ften you'll hear people talking about, "we shouldn't fly, "we should have more sustainable transport routes." but it's very rare to say... to hear someone say, "turn your phone off because of the environment," or, you know, "close instagram." so it's a really important conversation to be having. and the way we quantify this really sits with how we manage, store and process data. so data, as you... as you work on it, you have a very small laptop, but there is a data centre somewhere with rooms and rooms of computers that is processing all of that information, storing that information so that, as a user, whether you're in industry, whether you're at home, whether you're on your phone, you can access that data in milliseconds. and just for the uninitiated, david, can you also explain what it is in particular about al that requires such vast energy consumption? and of course, not only that, when you look at these data centres, the amount of water that's needed to, well, to cool them. so when an ai model is processing, that's taking a huge amount of information from something that we call a test set. so that's a lot of information that's been trained or been used to train that...that ai model. um, and that's all whirring away in some data centre somewhere outside of a city. um, and as a result, powering that data centre has its own issues. but those. . .those centres get really hot. and so there's a huge amount of energy and water that goes into cooling those data centres just to make sure that they, um, they don't overheat, the data centres where data is being processed, or the server rooms where that ai kind of development is happening that has um, they don't overheat, the data centres where data is being processed, or the server rooms where that ai kind of development is happening that has a huge energy footprint. 0ften often we see data centers away from water sources, nevada is an example where there are huge suites of data centers, so it is to be shipped into the desert to get the data centers, and it's same with the energy, if you are not connected to the grid, there is a reliance on fossil fuels that ramps up the energy requirement and the environment to emissions that come from running those data centers, and running those ai activities. so just to be clear, david, essentially it's the training of these language models, such as chatgpt, that require so much energy, rather than say, well, a user like mejust putting in a query with an ai chat bot. it's a little bit of both. so when you're...when you're developing an ai model, you have a training set. so if we were looking at image recognition of cats, you'd have 10,000 pictures of cats and you'd go through and you say, "this is a cat, this isn't a cat." and that's an awful lot of energy. that's a large, large proportion of the... the energy required to get that al model running. but then every time you check in, you're checking in and there's a small piece of energy, but you're one person doing that a couple of times, we scale that up across the world, then there's lots of people using a small amount of energy relatively often. so it's not just. .. it's notjust the training orjust the use. it's a combination of the two. david, i guess it begs the question, just...just how bad could it get if we continue developing these ever more sophisticated machine learning programs, which are, well, are going to require, one would imagine, just more and more data centres being built? there are a variety of predictions on...on how this will go, but without any sort of environmentally responsible technology interventions, we could see that, um, more than 50% of energy consumption globally goes towards digital technologies. you can see, think about, that is just you can see, think about, that isjust one you can see, think about, that is just one slice of that thing we use energy for, today that would then result in a whole host of different issues that arises, regarding managing indeed digit sensors and increasing the amount of energy used to maintain these technology. 50%, david. i mean, that's an astonishing number. but let me ask you this. the big tech companies — the likes of microsoft, google, meta, now apple as well — they've been pushing really hard into the ai race over the past few years. and i'm kind of wondering, do you feel that they're taking into account the carbon footprint of this big push into ai? is it something that they're sensitive to? so over the last few years, we've...we've seen movement from the big tech companies. google have committed to making all of their data centres run on renewable energy 100% of the time. and that's...that's not a light you can just turn on. if you think about the data centres that these companies have all over the world, it's very different powering a data centre in southern california with renewable energy compared to southern africa or east asia. different data centres will require a different mix of solutions to actually get them to net zero. and so we've seen the big companies... big tech companies moving in the right direction. we've maybe not seen them move fast enough, but that is something that is now being ta ken account for. and, david, let me end on this. in your view, what do you think is needed to...to make ai more sustainable? what we need, um, is clear guidelines on the best ways to use ai, both from a from an innovator point of view, if you're building ai algorithms, but also from a user perspective — what is the best and most environmentally conscious way to use ai? and what that will help is it will help new businesses entering the market. it will help people using ai. it will help established companies that are looking at how do they bring in things like large language models to their businesses and going, "ok, this is clearly... we always look at "what's the cost roi, but how do we make sure "that we're doing it in a way that respects nature?" well, on that point, dr david pugh, director of sustainable industry at digital catapult, great to have you on the show. thanks very much for your time. thanks very much. 0k. as the demand for al keeps growing, so does the demand for the infrastructure. yeah, those huge data centres. microsoft recently announced a $3.2 billion investment over two years into cloud and artificial intelligence infrastructure in sweden. microsoft's president and vice chair, brad smith, said at a press conference that one of the reasons the company chose sweden is because of the country's plentiful access to green energy. the big part of the reason we're able to do this is because of sweden's forward—looking energy policy, the plentiful access to green energy, whether it's carbon—free energy or renewable energy. it's because of a partnership with a company like vattenfall that has been so important to our growth here. and so part of what we're doing here is committing ourselves to run these data centres with the world's best water conservation and the broadest commitment to the use of green energy. but, will commitments like that be enough to reduce the carbon emissions from the growing number of data centres we need around the world to satisfy all of those ai demands? well, let's find out, because i decided to catch up with the big boss of one of the world's biggest data centre providers. andy power, a real pleasure having you on the show. and, andy, let's start with this, cos i'm kind of wondering to what extent have you seen the data centre industry change over the past few years with this big push towards ai? we've been in business for over 20 years now, but certainly seen quite the inflection over the last several years. so today we are supporting 5,000+ customers across 50 metropolitan areas on six continents. two — everything has just gotten larger. the smalls became mediums, the mediums became larges, and the larges became extra, extra larges. so hyperscale was a word that came to the industry and some of our facilities now you can literally park one to...or close to two aircraft carriers. so everything has gotten bigger and bigger and bigger. that's phenomenal. so, andy, let me ask you this, though. do you expect to...to see that sort of demand just continuing to grow rapidly? almost all the data that is being created in today's world, er, through our devices and then often hooked up to various cloud computing, is going back to a data centre in one shape or form. er, you think about what's on your phone or your laptop or your ipad, and it doesn't just sit there permanently. it's getting backed up on a recurring basis to the cloud. er, the misnomer of the cloud is, the data doesn'tjust float into the aether. it gets connected through fibre—optic networks, back to data centres that are large—scale facilities with significant power filled with our customers' servers. well, let me ask you about power demand, because it's certainly one of the biggest challenges facing data centres, and your industry. it's already overtaken the carbon emissions output from, well, certainly, say, commercial aviation. and as you say, andy, demand just continues to rise rapidly. so the question has to be, how can all of this be sustainable? today we're a 60—billion enterprise and we are working on projects as we speak that are north of $5 billion of new data—centre capacity. and we have the long—term runway at our fingertips as a company to double the size of our megawatts, double our company, across our 50 metropolitan areas. er, we do so in a sustainable framework and we're very... we're not in this for a trade. we're in this for the long haul, to make sure that we're building in a thoughtful way, and not speculatively building, but signing long—term contracts with our customers that are really inventing the next technology. gen a artificial intelligence essentially will be some of the technology that will help us grow in a sustainable framework. we've been consistently building, owning and operating for our customers in these markets and really future—proofing their growth in a sustainable framework. andy, are you doing much in the area of moving renewable power sources to your data centres? we are a leader when it comes to sustainable data centres. that's how we design, that's how we operate, and that's certainly how we power. er, we are the largest data—centre provider in terms of power purchase agreements. so, as a reminder, those are long—term financial agreements we enter into that facilitate the creation, the building of solar farms and wind farms that are bringing green electrons to our grid around the world. we're on the road map to get to nearly 80% green power in the next 12 to 18 months. and we're not stopping there. we're pushing the needle. er, and that's part and parcel with our business. we need to raise the bar in terms of the sustainability impact, given the asset class that we are experts in. well, on that point, andy power, the big boss of digital realty, great to have you on the show. thanks for your time and we'll check in with you soon. thanks so much, aaron. hey. what more should big tech companies at the forefront of the ai race do? here in london there's been a big summit taking place with senior execs from all the tech giants coming together to discuss the future of the industry. so i sent him with one of them, the head of dayton ai and ibm. so i sat down with one of them, the head of data and ai at ibm. ritika gunnar. a real pleasure having you on the show. and, ritika, let me start with this. i guess, quite simply, how much of a role does ai play in ibm products? well, we believe that al is central to ibm's strategy. ai and generative ai has really infused every part of our portfolio, whether it be the infrastructure that we build, the software that we build, and how we infuse that to empower more productivity for our users, our ai portfolio itself, which other clients can use, and our consulting practices. so throughout all three we have seen ai make an immense acceleration. it is about how we help all organizations infuse ai across their most essential systems. they are different things that you think about their nonconsumer technologies that are used everyday.— are used everyday. what are the ? are used everyday. what are they? examples, _ are used everyday. what are they? examples, if- are used everyday. what are they? examples, if you - are used everyday. what are they? examples, if you put. they? examples, if you put generative _ they? examples, if you put generative ai _ they? examples, if you put generative ai and - they? examples, if you put generative ai and your- they? examples, if you put| generative ai and your most essential systems, you think about how they can scale and perform, how they can do so in a way that is efficient and most importantly what we see in the major inhibitors is in and around how you govern and manage, especially with it as of the regulatory and compliance concerns. when we think about generative ai, it's about infusing them in your central systems while managing the compliance and risk aspects with trust and transparency with trust and transparency with how you build your ai and how you put that out there for everyday use. ok, so no doubt ai, cloud, huge part of your business. i've got some of the numbers here. the latest numbers. the last quarter, the last three—month period, ibm raked in, well, $14.5 billion. i think that was a 3% jump or increase compared to the year before, thanks in large part again to ai and cloud. so given that, it made me wonder about data—centre capacity, um... can you give me a sense ofjust how much data—centre capacity ibm uses around the world? well, we have data centres all across the world. and what i will tell you is that the way in which we build these data centres, we really look for that sustainability from the beginning. 74% of the energy that our data centres consume come from renewable sources of energy, which is something that... we put sustainability at the heart of what we do and how we build out these data centres. one of your competitors, microsoft, it's been quite transparent about how much its emissions have increased because of the huge demand in al. and, in fact, microsoft says that their emissions have increased by 30% since 2020. i'm just wondering, have you, has ibm done those sort of calculations? from our perspective, we are still on track with our goals for our 2030 no—emissions target that we have. six years. carbon neutral. yes. and you've already said that 70%... so far, right? 74% of the energy. 74% of our data centres today are using renewable energies. that's ibm's data centres. yes. correct? yeah. do you put pressure on the other data centres that you're a customer of? well, yes, of course. we put a lot of sustainability practices in everything that we do, all across not only how we build our software, we hold our partners to a higher standard in their sustainability efforts as well. and can i ask you this? what other areas has ibm done to help mitigate the environmental impact of... well, it's a cloud business, you know. yeah, when you look at especially in the ai space, some of the things that we're seeing in the market, to be able to have and minimise the environmental footprint, there are really four things that we see in the market. first is really about the model selection that you have when you're using ai in your applications and in your systems. you know, people always think that bigger is better. that is not the case. bigger is not always better when it comes to model selection. you can have a very fit—for—purpose kind of model that is actually smaller and can yield better results for you. so therefore, the footprint that you need in that data centre to be able to get the outcomes you need is a lot smaller. yeah. the second is, you have to think about generative ai and ai as an ingredient in an overall application and in an overall workflow. and when you do things like that, what you're actually doing is you can optimise the networking, the data access, and if you can keep things closer, you're also minimising that footprint that you have in the data centre. and finally, the infrastructure actually matters. yep. and so if you're picking out the right infrastructure for that right use case, you can minimise that footprint. and if you're open with all of these metrics, you can use that to create an iterative feedback loop and always create an efficient type of footprint. so it's about the efficiency, it comes down to. but is all that, ritika, is all of that enough, do you think, given the huge demand that we've already seen and everybody tells us that that demand is just going to continue going through the roof? i think in each of these areas, there are going to be continuous new innovations and improvements that come up. i'll give you an example. in the areas of models, we're seeing techniques like quantisation, which allow you to take a larger model and to make it smaller in nature, running in a smaller footprint and a smaller infrastructure, yet yielding you very similar outcomes and similar performance. and in these types of innovations that are coming out into the market, we're going to see this will continually help the sustainability footprint of a lot of these generative ai use cases. i've got you sitting in front of me. you're a big boss at one of the tech giants on this planet. so i want to ask you, ritika, how much responsibility do you feel that large tech giants like ibm have...? the responsibility they have in mitigating the environmental impact of the huge rise in al. it's a big responsibility for the tech giants, and that's why i think we at ibm are always looking at having sustainable ai ethics at the heart of everything that we do. we have an organisation within ibm that really focuses on making sure that we're on the trajectory to hit our targets, as we talked about, our 2030 target, and that we are continually innovating to be able to drive those targets as well. and, ritika, let me end on this. i'm not axing you to talk or are there tech giants, you are in the field, are you following the same path, all the tech giants are doing the right thing? i giants are doing the right thin ? ~ �* , giants are doing the right thin ? ~' �* , ., , thing? i think there's a desire to do the _ thing? i think there's a desire to do the right _ thing? i think there's a desire to do the right thing, - thing? i think there's a desire to do the right thing, we - thing? i think there's a desire to do the right thing, we talk| to do the right thing, we talk to do the right thing, we talk to our tech counterparts, we are part of an ai alliance that are part of an ai alliance that are funded with many other organizations, and apart from having open and trusted ai, we also talk about not only the ethical considerations but the sustainability considerations as well. to sustainability considerations as well. ., . ., as well. to what extent do you think i -- _ as well. to what extent do you think | -- ibm _ as well. to what extent do you think | -- ibm are _ as well. to what extent do you thinki -- ibm are leading - as well. to what extent do you think i -- ibm are leading the| think i —— ibm are leading the way? thinki -- ibm are leading the wa ? �* , thinki -- ibm are leading the wa ? v ., thinki -- ibm are leading the wa ? �*, . . , ., way? it's a great question, if ou way? it's a great question, if you think _ way? it's a great question, if you think about _ way? it's a great question, if you think about how - way? it's a great question, if. you think about how enterprises are adopting artificial intelligence, there are things they need at the core of the ai they need at the core of the ai they are using, then either ai they are using, then either al to be open and transparent and trusted, and we recently released our grand night series of articles as open source so that any enterprise can understand the data used to train it so it is there in the open. at ibm, we are customer zero, we are playing these techniques within our hi —— hr organizations, or procurements, organizations, or procurements, or sales organization, all throughout. and, ritika, let me end on this. let's say in five years' time, what does the...? what does the ai and the data centre picture look like around the world? i think we're going to find a lot more sustainable ways of being able to continually increase the efficiency of how ai and data is used, but still have the productivity that organisations need to be able to drive their business. well, on that point, ritika gunnar... thank you. ..a real pleasure having you on the show. thanks very much and good luck with the summit that you're visiting here in london. thank you. it's a pleasure. well, that's it for this week's show. i hope you enjoyed it. don't forget, you can keep up with the latest on our global economy on the bbc website or the smartphone app. of course, you can also follow me on x. x me, i'll x you back. you can get me @bbcaaron. thanks for watching. i'll see you soon. bye— bye. hello. saturday brought more than our fair share of heavy thundery downpours. for some of us on sunday, there probably won't be quite as many showers, particularly across england and wales, whereas for scotland and northern ireland, we'll see cloud and outbreaks of rain sinking southwards. 0n the earlier satellite picture from saturday, you can see clumps of cloud all rotating around the centre of an area of low pressure. this low still with us for sunday, but loosening its grip a little bit — so the winds will be easing, especially down towards the south. some showery rain first thing across some central and eastern parts of england. that should clear. then we will see some spells of sunshine, a scattering of showers, but not as many as we had during the first half of the weekend. conversely, for scotland and northern ireland, cloud will sink its way southwards, with some outbreaks of showery rain. quite breezy across the western side of scotland, down into the north coast of northern ireland, lighter winds down towards the south. temperature—wise, well, 13, 1a degrees in the north of scotland, 19 or 20 in parts of south wales and southern england. now, as we head through sunday nights, we'll see this band of increasingly light and patchy rain sinking into northern england, parts of north wales. a lot of cloud filtering into western scotland and northern ireland. that'll hold the temperatures up — 11 degrees there in belfast, a little bit chillier where we do see any lengthy, clear spells. we will see some spells of sunshine on monday, perhaps the best of those across parts of wales, the midlands, down towards the south of england, the channel islands could just see one or two showers, and then we will see some showers breaking out once again through the day, always with a fair amount of cloud across the western side of scotland and northern ireland. temperatures north to south, 13—21 degrees, so not far off the average for the time of year. this clump of showers likely to move across the channel islands, perhaps clipping into southeast england on tuesday. some showers popping up elsewhere. a decent amount of dry weather. temperatures still in that range between 12—13 in the north — 21 degrees down towards the south of the uk. now, this ridge of high pressure is going to try to settle things down through the middle part of the week, but there's still a bit of uncertainty about how long it will hold on. this frontal system trying to push in from the south, this one trying to push in from the west. either or both of those could bring the return of some rain later in the week. there are some tentative signs that things could turn a little bit warmer later next weekend. live from washington, this is bbc news. israel says eight of its soldiers were killed in an explosion during an operation in gaza's southern city rafah. world leaders gather in switzerland with ukraine's president volodymyr zelensky to discuss how to bring peace to the country. the princess of wales makes her first official appearance since being diagnosed with cancer at the king's official birthday. hello, i'm carl nasman. israel's prime minister has vowed to continue the war against hamas after eight soldiers were killed in a blast in southern gaza. the incident happended during what the idf called �*operational activity�* in the city of rafah in southern gaza. it identified one captain and said the names of the others would be released later. israeli forces have been attacking rafah for weeks, saying hamas�*s last remaining battalions are based there. meanwhile, the hamas—run health ministry in gaza said 30 palestinians have been killed by israeli forces in the past 2a hours.