Data centers and the AI era

Speaker A: Hello, I'm Michael Hainsworth. Every day, thousands of petabytes of data flow through networks. That's the equivalent of 86,000 years worth of HD video watching about the same length of time human civilization has walked the earth and it's delivered every single day. And most of that is generated, stored and transmitted in data centers hidden throughout our daily commute. As AI takes off, we'll see a 50% increase in the number of public data centers over the next five years, driven by AI's need for more processing power and by the enterprise need to digitalize every aspect of their operations. Meet Nokia's Paul Alexander. He spent his career in telecom and helped companies develop their network infrastructure. For Nokia, he knows the power of the network effect.

Speaker B: When you come together with a partner that can really help you go to market, help you scale, help your strategy, help your objectives, then you get a network effect of bringing those people together, which is, are greater, the sum of the parts is greater than the whole, if that makes sense.

Speaker A: On, um, this episode of the Network Effect, we learn how existing data centers need to evolve to keep pace in the age of AI. Why companies and countries are seeing this as a strategic race, and why the future is built on people and partnerships, not petabytes. Just as data centers became old news, AI injected some new energy into the sector. How so?

Speaker B: Well, let's take a step back, Michael. Um, the first thing to say is that AI is a game changer and it's a race, by the way. It's changing everything. And the existing data center infrastructure that we know that's underpinned the cloud for the last few 10 years, 20 years, is radically going to have to change to keep pace with what AI is going to throw at it. Data centers are the heart, if you like, of the cloud era and the future of the AI era. Uh, and networks are going to be the veins and the arteries. All of that is going to need to be upgraded to enable us to use those applications that are going to rely on that data center infrastructure and technology. And let's just look at some of the figures we've seen that we expect that the hyperscalers alone are going to invest something like 290 billion just in 2025. And we expect something like 1.1 trillion to be invested by 2029, 2030. So this is a significant investment. They're placing massive bets. This is an investment in the infrastructure. Obviously the applications that we're going to see run on this infrastructure of a little bit yet to come. So this is A game changer. This is a race and the, and the certainty and the profitability and the long term prosperity of not only companies but countries depends on getting this right.

Speaker A: Consumers expect instant access to data and businesses require real time analytics.

Speaker B: So this is one of the things that are potentially driving the demand for edge data centers. Right. So obviously we've seen the growth in large data centers, mostly in the US actually which are to house and underpin the large language models that are the basis of the major investment that we've seen recently. But as AI applications and services become more ubiquitous, we're going to rely on low uh, latency to be able to deliver those applications to especially for things for like potentially autonomous vehicles, ah, you know, real time analytics of data. Think about the banking sector where milliseconds matter. So we're going to see investment I think in the edge data centers and that's not only driven by that need for ultra low latency, but by a number of other things as well, like cloud repatriation. So we're seeing the large enterprises come out of the cloud and start to look at a hybrid policy where they put the less critical data in the cloud and they put their crown jewels in a, uh, in their own premises or even in a data center, but much closer to where they are. So those edge data centers for latency, for colocation and by the way also for potentially sovereignty and where that data is located and stored. Yeah, that's probably one of the those things that will drive those edge data center technology in the future and I

Speaker A: can imagine it'll unlock some new opportunities and experiences like uh, immersive technologies that you couldn't get if you had a long round trip to take. Which for your ones and zeros.

Speaker B: Yeah, um, we used to talk about mission critical networks and those are the networks that essentially are life and death if you get it wrong. The rail network, the highways network, air traffic control and Nokia supplies a lot of the technology into those mission critical networks today. So we really understand that. But let's be clear, what is mission critical now? It's the user experience. Everybody's experience is mission critical, from the gamers to doctors, uh, to everybody using these kind of applications. So that user experience is going to drive a uh, competition for service and high quality of service. And that's clearly going to be one of the major focuses uh, of the future investment once we get this sort of initial investment infrastructure deployed.

Speaker A: So then what are the big challenges data center providers are facing?

Speaker B: I say there are probably a top three, um, one is finding a site for the data center in the first place. There are places already which you'd call the tier one locations that are already full. I can think of a few examples. Even Slough, for example, in the United Kingdom, there's no more space for data center because frankly there's no power. So even if you get a location for a data center, number two, you need power. And frankly power is the issue at the moment. To get 30 megawatts to a data center, you're talking something like three to five years of planning ahead and requesting that. That's significant. That's probably why you're seeing this huge demand for power, because it's being built up, a pent up demand for what they expect will be those demands in the future. You get those two things right, then you need the final thing, which is arguably in our opinion, probably the most critical and the thing that's not talked about enough and that's the network. You need connectivity. You have to have a high capacity, ultra low latency, highly reliable and by the way, super secure network to be to effectively to underpin those veins and arteries between those data centers which are the heart of AI.

Speaker A: But AI's appetite for energy knows no bounds. How has the data center evolved to accommodate for the increased need for sustainable practices?

Speaker B: Yeah, it's a guess what an area of major focus because again, if you can't get the power you want, you get the power you have or you can get access to. So things like liquid cooling, um, are becoming more prevalent. We're seeing locations for data centers being put into, for example the Nordics or colder regions where you can take cold air from the environment at no cost and cool your data center. Um, we're seeing um, other innovative things like in the design of data centers. Clearly the equipment that goes into data centers needs to be more power efficient. And that's something Nook is very focused on as ESG is a key part of our pillar, a key pillar of our uh, corporate strategy. So we work very hard to make sure that the power draw from our equipment is as low as it possibly can be. And just for example, in the last generation of IP we managed to save 75% in terms of power utilization. Um, and then there's the sort of focus about obviously not only cooling these data centers, but also where you could potentially get alternative sources of power, sustainable sources of power. So putting the locations of data centers maybe in Spain or sunnier countries and therefore possibly not the uk, but that's also one of the things that is uh, the data center providers are looking at AI is a game changer and a race, and everybody is racing.

Speaker A: So AI has made data centers sexy again, fueling 50% growth over five years and demanding faster, more agile infrastructure. Existing deployments must evolve to keep pace with AI's requirements and as much as a trillion dollars will be spent between now and 2029. That's a lot of green, but what about a different kind of green?

Speaker B: We say there's no green without digital. Uh, ESG is a pillar of our corporate strategy. We aim to make sustainability part of our, uh, effectively our competitive advantage. Look, we have Bell Labs. You've probably heard of Bell Labs. I hope most people have heard of Bell Labs. You know, inventors of similar, you know, fairly obvious things like the transistor, the laser, the first satellite in space, you know, pretty fundamental things to our lives. So Bell Labs is 100 years of history of an R and D and that's celebrating that this year. Um, but we spend around $5 billion annually investing in R and D. And we put that investment into finding the latest and greatest technology, pushing the leading the bleeding edge of what's possible. And as I mentioned earlier, we managed to save 75% power utilization on the last next generation of RIP and 60% per bit cost deliver reduction on our optical portfolio. So we're working all the time to drive down the power utilization and um, by the way, the size and space of our products so that they are uh, cheaper, easier to run and operate.

Speaker A: So you're green, but you're also pushing the needle into the red when it comes to record setting speeds. How do you accomplish that?

Speaker B: We do that by pushing the leading edge, the bleeding edge. And that's really part of that R and D investment. So we work with customers all the time to essentially test new technology and drive and push the boundaries.

Speaker A: You set regional speed records. Help me wrap my head around 1 terabits per second that you managed to accomplish in a project in the Middle east.

Speaker B: Yeah, so 1 terabits per second, I mean it's kind of like similar to the numbers that keep going up on the spend in data center and AI, isn't it? We're pushing the boundaries the whole time. So as you rightly say, 1 terabit per second delivered with our partner customer STC, over 850 kilometers between data centers. And that's one example. We worked with Colton Windstream recently and we delivered 800 gig over 8 and a half thousand kilometers across the terrestrial and under the sea to do IP and optical connectivity between Chicago and London. Um, and there are other examples as well. I mean, we provide the next generation of fabric to Microsoft. So our products are being deployed globally and we're obviously pushing the envelope as we do that.

Speaker A: And as they like to say in Silicon Valley, Nokia eats its own dog food by employing AI within its own networking infrastructure. So how does that help address the customer demand of AI on their side of the network?

Speaker B: Yeah, um, look, AI is a powerful tool. And so we embed AI into our management suite, our management tools to manage our products and our portfolios. Uh, AI will start taking on some of those more mundane tasks. And AI essentially, um, provided you get one critical thing right, by the way, which is you have a highly reliable network that you can build on scale and then autonomize. Provided you get that network infrastructure right, you can bring in a whole lot of autonomy. And that autonomy will be driven largely by AI. So we embed AI, uh, into our systems to help things like predictive maintenance, to predict failures, to avoid failures before they happen, into tools and systems which allow customers to perhaps optimize the use of their products. So, for example, if they're being underutilized or utilized below their full capacity, you can turn down the capacity and therefore turn down the power utilization. Place clever things like that that we do, we build into our portfolio so customers can maximize the use of that. But again, it's fundamentally based on having the right technology that's highly reliable to allow you to want to autonomize that now and build the scale.

Speaker A: When I think about the work you did in the Middle east with your partner, um, as remarkable as it is that you can achieve the speeds you did at the distance you did. I wonder, though, if edge computing, as we've discussed earlier, is going to end up being ubiquitous. Are the days of putting a data center in a farmer's field behind us

Speaker B: if that farmer's field had power and connectivity? Maybe not. Um, to our earlier discussion, but look, I think that edge data centers are going to become ubiquitous and there are a number of factors driving that. We talked about the latency issue early. We talked about applications like autonomous vehicles, real time video analytics, banking, for example, that are going to be so, uh, reliant on that ultra low latency that that's going to mean the data centers and where this, uh, inference is done needs to be closer to the end user. And that will be in the handset or the local, very local data center. We are going to see much more, much more of these edge data centers and we're talking to customers that are planning upwards of a thousand data centers, edge data centers, just for example, in the UK or a similar number in France. So absolutely, I think this is the way things will go. And by the way, also because power is so difficult to get to those large data centers, if you can put a number of smaller data centers out there, you can reduce your immediate power demand in that local area. You're more likely to get that power. But then you need to link all of those data centers together. You need to federate it not only so that you can utilize those resources, but then you can basically make that those individual data centers look like one big data center cleverly, and therefore flex your resources, flex your capability and be more, um, be able to offer customers something much more close to home and much more unique.

Speaker A: When you say close to home, I can imagine, um, data security is pretty critical, but also data sovereignty, and I think you touched on that a little bit. What role does the demand for security and data sovereignty play in this brave new future?

Speaker B: Okay, well, let's start with sovereignty, because what we're seeing there is regulation. Uh, we've heard NIST, M2DORA, for example, in Europe we're seeing other regulations likely coming, AI acts, et cetera. Um, and this is driven by a bit, the geopolitical situation, but also people wanting to make sure that the data, and by regulation the data needs to be where it's generated or where it's stored, acting under those local regulations of that country. So that geopolitical situation, where that data is stored, that sovereignty element is going to drive, um, specific networks. And we're seeing people starting to deploy sovereign networks that are very, very clean or very, very specifically for that type of purpose. So that's opening up a new avenue, coming perhaps to the security thing that is fundamental. Um, let's talk briefly about quantum Q day is coming. Q day is that day when a quantum computer is viable and able to be utilized. The estimates vary between 2027, 2029, 2033. I know Jensen Huang recently said it'll be 20 years, but absolutely they're pretty confident that it's going to be much, much before that. And uh, when those quantum computers arrive, they will have the ability to break all of the security that are run across these existing networks today. And there's this harvest now decrypt concept, you may have heard of that, where you harvest the data. Now you wait for your quantum computer, you decrypt it, and then you get access to that data, which is why we're absolutely encouraging our customers to deploy quantum safe networks today which we have built into our IP and our optical portfolio to give you a defense in depth strategy. So security is key. But the other thing is of course DDoS, DDoS isn't going away. DDoS is actually scaling and our friend AI is actually going to be our enemy because AI is going to be able to generate more and more of these DDoS attacks and going to do it at scale and it's going to do it 247365.

Speaker A: Right.

Speaker B: So we need DDoS capability built into the networks and um, for example we're working with links and other IXPs where they essentially are the underpinning network of the Internet. Right. So 100% of Internet traffic essentially travels through these Internet exchange providers. Nokia provides 6 out of 10 of the platforms, the IP platforms to these Internet exchange providers. And what we can do by deploying deep field, which is our uh, uh, defender solution which essentially is able to identify through analysis DDoS attack and scrub them in line within the network infrastructure. So I think again Noki's in a really kind of interesting position to help customers with those sort of DDoS threats and those cyber attacks that we see I suspect will continue to scale into the future. The growth in data centers is going to continue at uh, pace.

Speaker A: Scaling data centers to meet the demands of artificial intelligence needs real intelligence people. And what are the challenges that these people are facing? Well when it comes down to it, they're reshaping every industry from automotive and logistics to health care and education. So the thing is is once we bring these people together, what will the impact of the network effect be a decade from now? Wow.

Speaker B: I think um, I think the world is going to be radically different in 10 years time. I mean m. If we, if you and I perhaps Michael, are just, just about old enough to remember the world pre Internet. Well I certainly am. And the Internet has changed every industry, every ecosystem on this planet in the last 20 years. I think that's the same with AI. I think AI will change everything. Um, education for example, education will be radically different. My 18 year old has just completed his education. My 4 year old is about to go through the way she's educated with AI, with specific learning, with targeted focus on her abilities, uh, where she maybe needs some help. It's going to look very, very different. I think in the future medical industry is going to completely change. We maybe will get access to GP appointments which we're struggling to get to now here especially in the uk uh, an AI doctor is going to get you through that first step and probably weed out a number of those, uh, those people that don't need to be seen first. So the medical industry is going to change radically. Transportation is going to change radically. We're going to have autonomous cars, autonomous trucks, autonomous trains, maybe autonomous planes. I know they fly by wire a bit now. So, you know, you cast the lens of AI onto any industry and you can say industry 5.0 will be the era of AI, uh, and it will change everything radically.

Speaker A: So how do we bring people together to make this future a reality?

Speaker B: Yeah, that's going to come down to partnerships. The one thing that I've learned very much about the cloud ecosystem in the last year is it's an ecosystem. There are thousands of players in this ecosystem. Not one party does everything end to end. You need a whole group of companies and partners to come together to deliver this complete ecosystem from storage to networks to GPU to building the data center and on and on and on. Um, so those partnerships are going to be fundamental and in fact, look, we're fundamentally pivoting our focus as our business to work much, much more with partners as we go forward, as we pivot much more towards this data center. As our CEO recently said, we're investing 100 million to 300 million over the next few years to really go and drive into this growth area of data center. So we won't be able to do that without partners. And therefore that's going to be a major focus of our business going forward.

Speaker A: So what's the solution to the hardest part about making partnerships work?

Speaker B: I'll give you one word, trust. I think trust is what builds partnerships. Having built partnerships before with uh, uh, existing organizations, I know the one thing that they really want to know is if we're going to work together, can we trust each other? So that means we need an aligned go to market, clear strategy, clear objectives, that we're working genuinely as one together with one. And one plus one, as I like to say, should equal three when you're working with a partner, not two. And that's where you bring value to the end users, to customers. And so I think it's that one word, trust.

Speaker A: Describe to me what the network effect means to you.

Speaker B: I think the network effect means none of this is possible without highly scalable, highly secure, ultra low latency networks. And that is where Nokia comes in. And that's what I think our value is to this growing industry. Networks that you can rely on one plus one should equal three. So when you come together with a partner that can really help you go to market, help you scale, help your strategy, help your objectives. Then you get a network effect of bringing those people together which is a greater the sum of the parts is greater than the whole if that makes sense.

Speaker A: Can we do it without the network effect?

Speaker B: Absolutely not.

Speaker A: Global demand for artificial intelligence is soaring, propelling data center expansion at unprecedented speed. But moving fast and breaking things is not an option. Networks must deliver capacity reliably and at low latencies while meeting sustainability demands. The top three challenges securing locations, sourcing power and ensuring high capacity. Ultra low latency networks With ESG at the forefront, new data centers embrace energy efficiencies from liquid cooling to custom silicon edge computing is the new normal driven by data sovereignty. Real time analytics and ultra fast user experience is all powered by partnerships an ecosystem woven together by trust and shared innovation. As AH AI transforms every industry, the network effect becomes the multiplier for the future. I'm Michael Ainsworth, thanks for watching.