Well, hello. I'm Aart De Geus, and I have the privilege to talk to you about a big challenge: the dual challenge between climate and energy. And this is a challenge that's accelerated by the very fact that AI is now wanting a lot more power, which falls in the energy category. Now, my background is not in energy or in climate. I grew up with semiconductors. Actually, I had the privilege to lead a company for many decades called Synopsys. And we essentially provided tools to design most of the advanced chips in the world. Now, to give you a little bit of a sense of the roadmap, here's a simple model. Earth's resources, standard of living. They're connected via two pathways. Climate, which of course determines how you feel and the temperature and all that. And then at some point in time, humans discovered fire, and it completely changed the picture and improved many things for the humans. Since then, of course, fire falls in the category of energy, and energy does have some effects on climate. These are understood through science, and science has made fantastic advances, and actually, there's still much more to learn, but we have a good sense of the impact on climate. On the energy side, it is technonomics. Technology advances the state-of-the-art, and economics closes the loop in order to be able to keep investing and changing things moving forward. Now, there's one more change in this picture, which is AI, as mentioned, that is both changing the standard of living, hopefully for the better, and at the same time, having impact on the need for more energy. So, there's four things I would like to briefly talk about. One is chip technomics, AI, power, and so on. Two, the impact of emissions and the dual challenge that results from that. Three, have a look at what we can do on electrification. And lastly, give a few calls for action.
So, let's get started right away with some very big words: exponentials and technonomics. Well, of course, the exponential that is best known in the history of mankind is Moore's law, essentially predicting that on a computational chip, we would add more and more and more transistors. Actually, we'd be doubling them initially every two years or so, and then later a little slower, more slowly, but still very much going forward. Actually, you can see it here on a log graph because an exponential log graph is a straight line. And what is quite amazing from the time of the first prediction, where it was about a few hundred transistors on a chip, today we have a few hundred billion transistors on a chip. And you can see at the very end, it's not only single chips; multiple chips are coming together in very close proximity, even stacking them on top of each other. And just for kicks, here is Synopsys since 1986. And we've touched pretty much all the chips that you see on the right side there.
Now, if we look at this picture, we quickly find that some of the most advanced chips—here's one that was just announced recently this year—has 336 billion transistors. That's a lot. But at the same time, the way you design this has really not changed. There have always been three metrics of success: make something better, get it sooner, and make it cheaper. And actually, it's exactly in that order. You start by creating a chip that can do something that is amazing versus the last one. And then the question is always how quickly can you go to market and how can you make it cheaper over time for its success. And the technonomics of that are very similar to what we saw many, many years ago in the industrial age. And I'd like to compare that now to the information age. And fundamentally, there are two steps. The first step is you invent something, and the second step is that you then improve it or you scale it so that you can build many more of them.
The invention invariably is some capability that is new, and then there's this great hope: will it be successful? Can we sell many of them? Can we build many of them? But in order to make it real, you have to really polish and improve and remedy issues and so on. Go around and around and around. And actually, this has been true since time immemorial. It's interesting that in the industrial age, in the late 1700s, the steam engine was invented. And then still by 1860 or so, every so often, some of those steam boilers would literally explode. And that suddenly looks like a fixer-upper from my perspective. But if you look at the first transistor, it looks a little bit like a high school science project. And who would have believed at that time that today we would have 300 billion of these on essentially a very, very small chip.
Well, in order to complete the economic loop, you have to get the economics right, meaning can you make it cheap enough and sell many enough so that you can reinvest in the next round and reinvest and reinvest and do this over and over again. And indeed, if we look at the advancements that have been possible, massive missions have come out of this. And the comparison is sort of interesting. You know, the industrial age essentially had a mission to augment and scale human muscles. The information age has an even bigger ambition: augment and scale the human brain. Quite astonishing, really. And one of the reasons we can actually think this is possible is because systemic complexity that grows over time is not one exponential. It is really many, many, many exponentials. Every time something gets slower or has a problem, you invent something new. And when you aggregate all of that and you see the semiconductor evolution over 60 years, we have arrived at a capability of computation to literally do artificial intelligence, which in turn creates a whole new set of exponentials going forward.
Now, all of this has one thing in common, which is both sides of this picture requires more energy, or in the case of AI, more power, which is really electrical energy. Okay, let's have a look at AI compute and power. What does that really mean? Well, you know, I was sensitized by actually an article in the Wall Street Journal last year. And it's sort of a cool one because it said that OpenAI was going to buy a large amount of chips from AMD. And actually, the chip was announced earlier this year. By the way, notice the number of transistors: 320 billion transistors on that one. But that was not the reason I got so interested. There was a little piece of text here that essentially said that they were going to buy six gigawatt worth of AMD chips. Well, that's a little bit the same as if you go to the butcher and you say, 'Yeah, why don't you give me four and a half acres worth of beef? That's a lot of beef.' Well, this is a lot of transistors. And the very fact that we talk about compute centers in terms of the amount of power used is really a change of perspective. So much so that a couple months ago, I had the opportunity to look in one of our own compute centers. And you know, these centers that they have these rows and rows of what's called racks. And I asked them to open up the door of one of those racks—that's the front and the back side of it—and the first thing to notice here is actually just one of those uses about 15 trillion transistors. Everyone has to be working and use this power of 95 kilowatt. That is substantial. And actually, that explains, of course, that those red and blue, those power lines, they are substantial and they go to the front and the back. Actually, I was just lying to you. Those are not power lines at all. These are water cooling lines. And actually, water that's coming literally from underneath. You can see it over there. Goes straight to the hottest chips. So it flows right next to the hot chips. And by the way, there's also air cooling through the whole thing. And so cooling is in itself about 25% of the total energy used on these racks. But the fact that these are 100-kilowatt racks and that you have rows and rows of them explains why this is going to be a lot of energy. And so ideally, you would want to have one of those compute centers and literally sort of just plug it in and get running.
Well, let's get to that, but just wait one second. So, this brings me to the dual challenge. The challenge of energy and emissions. Let me explain. You remember this picture where energy has really had a lot of positive impact on the standard of living, but at the same time, had side effects that essentially changed climate, and which climate actually changed the temperature of Earth. That's quite amazing. And in order to understand the impact of that, we quickly realized that the conjunction of both energy and climate are a very systemically complex system. They intersect. And I found that for systemic complexity, there are three things one needs to look at. One is what's big and what's small. The second one is what's fast and what's slow. And the third one is where are the feedback and feedforward loops. Let me give you a simple example of a feedforward loop. You know, temperature is up. You're living in Phoenix. So, what do you do? Well, duh. You crank up the air conditioner, of course. What does that do? It uses more energy. The energy has a side effect. And here we are. This is a feedforward loop. And so, understanding these has a very big impact on how we deal with the issues.
And it's in that context that we look at the externalities. And Open Minds took up something that I think initially was formulated by the National Petroleum Council, which is this notion that there is a challenge on these externalities that needs to be managed at the same time. And the challenge—the dual challenge—it's actually great terminology, I love to use this—is that we do want more energy, actually a lot more, and at the same time, we must have a lot less emissions. And the your part is actually augmented now by the very fact that AI says, and we want it very soon. And therefore, the dual challenge is absolutely on today. And so it's in that context that I've joined actually Open Minds. I love the organization because it was started by a number of people that took this dual challenge and brought together a rich set of people from energy, from climate side, and a number of other disciplines. Essentially, people that had in their own life already had many great accomplishments and were experts, and brought them together through the lens of energy and climate and the ability to impact markets, policy, and technology through their respective disciplines. So the mission of Open Minds is actually to help with the dual challenge as much as possible. Fantastic organization. And so if we take their perspective, and I love this dual challenge notion, the question is, well, what is the gravitas of the emissions? How bad is this? Well, let me show you a graph here that really had a big impression, made a big impression on me. It's the global atmospheric CO2 over the last thousand years. And the first thing, of course, to observe is, you know, nothing happened till 1800, and then clearly, it started to move up pretty rapidly. Actually, earlier I mentioned to you this exponential thing. Let's do a quick little experiment. Let's just take a copy of that. Here we go.
And lo and behold, this thing exactly looks like an exponential. And when I saw that, you know, that is a photograph of myself in that moment, because an exponential is an extremely dangerous curve. Now, I'm not saying it is an exponential, but it sure looks and walks and quacks like an exponential. And actually, if you look at the numbers, they certainly give a different perspective. By 1800—and actually this goes back to the last 10,000 years or 800,000 years—it was two trillion tons in the atmosphere. By 2024, it was three trillion tons. Obviously, the difference is one trillion tons. Now, it's not just one trillion tons. This is on an exponential that says that half of that trillion tons was just in the last 40 years. Now, CO2 is not the only gas that we're dealing with. The second most potent one is methane. And here's the methane curve. And what do you see? Well, again, flat until 17, 1800. And then what happens? Let's do that experiment one more time. Can't resist that. Here we go. Astounding. Astounding the fit. Except in this case, it's even worse. Methane increased by two and a half times. Now, it's important to understand what the impact of those gases are. And there are a few others, but in general, CO2 and CH4, that's methane. Here are the curves. Essentially, what the impact was, it has increased the temperature, and right now, we're already at 1.55 degrees Celsius above 1850. If we look at the impact, obviously, it changes the climate and therefore the weather and many side effects. And so you can look at the economic impact of this as well. And actually, this is just for the United States, a little messy graph. It's actually two graphs. One is the number of events, an exponential by the way, and the other is the cost of those events, over $200 billion per year. So this is very, very, very serious stuff. So let's have a little closer look at the impact of these two gases because they're not the same. It turns out that between CO2 and methane, roughly 70-30% impact on the global warming, and there's a reason for that. If you look at the two gases first from the perspective of their global warming potential over a 20-year span, CO2, let's say that's 1x, and so it's sort of a winter coat that you put around Earth, right? Well, methane is like 80x—you know, that's a super high-class, you know, down jacket that you put around it. Now, there's another difference between the two. If you look at the half-life, well, there's no half-life really for CO2; it stays forever. With other words, we keep accumulating and accumulating and accumulating. Methane, on the other hand, has a roughly 12-year half-life. Meaning after 12 years, half of it is gone, and then another 12 years, half of half is gone. But meanwhile, you saw the curve. We're still accumulating more. With other words, we're putting more in than is actually passing away by on its own behalf. And so these two extremes need to be treated differently. And that leads to my first call to action. Obviously, CO2 is the very big danger for the long term because it keeps accumulating. And we must transition as fast as possible to non-CO2 power sources.
For methane, there, we can take action right now and should, because if you use gas or oil, there are leaks that happen around that. We should remove those as fast as we possibly can. Now, that brings me a little bit to the overall energy sources picture. And if you look at this graph, you can quickly see that things have been growing over the last, what is this, about 30 years or so, 35 years. And when I looked at this, the first thing that surprised me was looking at 1990, and you see there was about 15% non-fossil, and those happened to be hydro and nuclear sources. And then, of course, the so-called renewables started to grow: wind, solar, and some others. And here, what surprised me was, well, it grew only from 15% to 20%. Well, the reason is that the overall amount grew massively to 170 petawatt hours per year. That's the overall amount of energy. And so still, wind and solar have been amazing successes. And so let's zoom in a little bit into the solar one. And here, let me show you one of my very favorite graphs. A friend of mine sent this to me. And what I love about this is look at the yellow lines. Those were the predictions every few years of the growth or the total installation base expected. The black line was the real one. And of course, I can't resist to do our little experiment one more time. And you look at the black line—that is an unbelievable success for solar. And actually, in many places, solar is now the cheapest energy source that you can get. You cannot get it 24 hours, but when you get it, it's very, very, very cheap. And so if we look at that, that brings us to a topic of looking at the overall evolution. That externality I talked before in CO2 is about 40 gigatons per year. Well, let's just multiply that by 25 years. That means in the next 25 years, at the present rate, we will add one more teraton. With other words, that move up is continuing with great, great danger.
Well, where is this coming from? Well, if you look at the supply, you can quickly split between non-fossil fuel and fossil fuel as we saw a minute ago. But I'm most interested now in what's happening on the demand side. Oh, here are the numbers. By the way, non-fossil is 20%, fossil is 80%. As we saw in the previous graph, if you look at the demand side, here we go. What we see is 20% of that is consumed in electrical form and 80% in non-electric. And this is very important because, of course, this is where AI is saying we want way more electricity to do our job. And interesting enough, that 28% is the world number. Here's the United States 16%. Here's Europe 22%. And quite amazingly, here's China, that's actually by now close to 30%. So how is that possible? How did they change that so fast? Well, this is called industrial policy. Meaning there were a number of decisions made in the 2000, around 2000, to actually accelerate this in China, whereas the rest of the world sort of just kept going. And of course, this is a change that we will now see all over the place by necessity. But it's also an indication of a great, great opportunity, which is that electrification has a great impact. And remember how Open Minds essentially looked at this and said, 'Hey, we need to deal with this dual challenge and get more energy and less emissions as quickly as possible.' Well, Open Minds decided that the first area they wanted to push forward on was electrification, and they took a subtopic referred to as the grid, which is essentially what connects everything, as the topic to move on forward on.
So, let's talk about electrification, and I'll focus only on the United States for this part of the presentation, but of course, it applies everywhere. And so if you look at electrification, the first question is, well, what's the US electric power demand today? Well, you can see the full line moves up to about four and a half thousand by the time we're in 2025. And there's a very big change happening. The rate of change is actually about 10x going forward, which is by no means easy to do and largely stimulated by demand. And AI is a big piece of that. But electric vehicles, HVAC, robots, all of these things are going to be more and more electrified. And actually, if you look at the real numbers, essentially what we need to add in the next 10 years, 2025 to 35, is 1,500 terawatt hours per year. We're right now at about four and a half thousand terawatt hours per year. And so half of that is about what will be used by computation, and the rest for these other things. And so that brings the question of, well, how can we execute on that? And we pretty quickly find that the grid is probably one of the issues that is in the way, or is very helpful to this. Let's look at the grid.
The grid is a fantastic example of a truly technonomic system, and it's made up of really two parts: mostly transmission, which is mostly the long distance, and then distribution into the neighborhoods, but also bringing down via transformers the voltage to the appropriate level. And one way to look at that is also to say, well, you know, it has a form and a function. And the form is really the physical connection between sources and usage. And there, we're literally talking about, you know, how many terawatt hours per year can go through the system. The function is quite different. It's actually an economic connectivity between supply and demand. And there, at the end of the day, we're interested in what's the cost per kilowatt hour. And so if you look at these things, they both have the same challenge: how can we massively increase this by this number, and how quickly can we actually do this? Well, this is where we truly are talking about massive system complexity because this is a system that essentially was built early this past century and has organically continued to grow and technology continue to evolve. And so one way to represent that to you is through a number of pictures. And so this is actually what the grid looks like. Connectivity. These are all the lines. The brighter the line, the higher the voltage. And actually, I've seen this at one of the organizations provide this. You can literally click in until you see the line that goes to your house. And of course, at the appropriate voltage for that part.
The challenge here is we have to cram 1,500 terawatt hours more per year through this. Here's a different map. These are all the energy sources, and you know, the black ones is coal and the sort of yellow-greenish is solar, and so on. This is continually evolving right now. And here, of course, our challenge is what can we do as we evolve to move the dual challenge in the right direction and use more and more clean energy. Here's the next one. Jurisdiction. Actually, there are many maps of that, any more than what I have here, and they all have sort of the same problem: in simple terms, permits. How quickly can you actually authorize the building of the next pieces that are necessary? And this is in many, many situations the very reality of systemic complexity. You touch one part, it has impact on another one, and so it's non-trivial to do. My favorite map, this one also has the compute centers. Those are the grayish circles there. And of course, the white lines are actually the photonic connection between those things. And here, of course, this picture now is dominated. So what does AI need? What can it do? And to what extent can compute centers adapt themselves to the need of the grid, and vice versa? And then we get to the bottom line for many people, which is electricity prices. And you see different colors, and you know, California is sort of dark because we had a lot of fires, and of course, there's a lot of rebuilding that needed and a number of other investments. But the reality is each region has its own challenges and can need improvement. And so the grid is in the middle of that, but affordability is one of the issues that people have, and they're very worried that compute centers are going to force a lot of cost that will fall onto their pricing. At the same time, a number of the hyperscalers have definitely committed to say, 'No, no, we will pay our own cost of transformation and make sure that the affordability is not hit.' To be seen how it works, but a lot of goodwill, I think, is there, and that should be harvested.
So if we take this picture, that leads to the grid vision that Open Minds has put together and is still continuing on. And of course, we understand we have to make it better, we have to have it soon, and cheaper. And here's the thing that we have to accomplish, but those are the metrics of success. The question is what are the drivers that actually really govern change, and there are three: technology, market, and policy. Same objective of change, of course. And if I look at them, I think that on technology, we're actually making fantastic progress and many great things to be done. On the market, it's a little bit more challenging because the question is who pays for what and how does it adjudicate it. And lastly, on policy, the challenge is there are a lot of permits in the way, and you know, my check mark is there's always hope, and we need to push on local and overall government harder to make these changes as fast as possible. Bottom line is that Open Minds has three categories of advice or input in its vision: modernizing the existing grid, building the one for the next decade, and increasing industrial policy urgency.
In the first one, there are many ideas, and here too AI can already have impact on how to dynamically optimize the utilization of the grid. By the way, reconductoring, very promising, going to high-voltage direct current, and the high voltage could use different cables that then can use the same pathway. So you have much less permitting need. And then in a minute, I'll talk a lot more about grid batteries because they're very exciting, what the changes they're bringing about. If you look at the building, well, there's certain things that we can do right now. They're partially already permitted, and that is fix segments that are bottlenecks in the present system. But just as important is actually for the long term: we need to improve the speed with which we build new lines substantially, about 10x more than what we're doing today. And lastly, we need 3 to 5x faster planning and permitting if we want to just be on par with what China is doing right now. And in a number of cases, for large, large buildings such as nuclear power plants, for example, we need to accelerate the availability of financing. So these are the things that Open Minds is pushing forward.
So let's talk about the grid but now from the supply and demand side. And so here's the picture again, and underneath you see the different supply sources. To the far right, you see coal and gas; on the left, hydro, solar, etc. And you know, let me add a little color—that or really colorful editorial—because the sources on the right, which is by the way about 56% of this, they actually generate CO2. And you see the number on the very left: the US generates 1 gigaton of CO2 for just the electrification. That's quite a large number when you compare it to the 40 gigaton worldwide for all the sources. Having said that, this is the dual challenge. We want to get even more power. Clearly, we need to move from the right as much as possible to the left. Now, part of that is instrumented by looking at the top at the uses. Now, you say, 'What is this picture?' Well, let's zoom in. I love this picture. A friend of mine gave this; they had done this at Google, actually. This is a digital twin of one of the US regions for a one-week period. And you can see sort of the seven days there. And let me explain a little bit what these colors mean. And so at the bottom, you see what's called clean firm. And so clean firm would be hydro or nuclear, for example. You can count on it any time of the day. Then, of course, you see wind in blue and solar in yellow, and they appear every day in different configurations. But the thing that I love really is this thing, the batteries. And what happens is when you have an overproviding of energy, you can actually charge batteries. Could be from solar, could be from wind. And the reason this is valuable is because when you don't have enough, you can actually discharge this. And notice that the red part right above the discharge—that is where actually gas was being burned. And so obviously, we would like to minimize that as much as possible. And so it's in that context that things like batteries really help in the life cycle of solar and wind and a number of other sources. And so that's the challenge. How can we make sure that this gas is going to be minimized? By the way, here's a picture of a large solar field. And by the way, you can see a number of batteries, and you see some of the numbers. They are quite substantially large. And so let me zoom in a little bit more into that because suddenly the battery is not just a technical capability. It turns out that it's truly a technonomic emphasis on anomic arbitrage. And let me explain this to this wonderful picture. This is actually the spot pricing in the United States. This picture gets updated every 15 minutes. So it's the function of how much does it cost, right? And what you see is the color coding. And so it's very cheap when it's on the blue side, it's very expensive when it's the red side. And this is middle of the afternoon in the central place. So they probably have the air conditioning on over there. Except when you look at this picture, there's something weird at the bottom. It's not only cheap. It's actually people will pay you to take their energy, and the reason for that is there are many energy sources that you cannot just willy-nilly stop. You know, you have a nuclear power plant, and suddenly there's less demand. You cannot just, you know, turn this thing down a little bit. It keeps running. And so this is, of course, where batteries come into this game because they can actually buy energy when it's very cheap, or even buy means they get money to take it, and then sell it when you're in the red period. And so another way of saying that is when you look at economics, it also helps in optionality and arbitrage. Let's say you have a compute center, and really what you like is optionality, in of course you know the supply, but also optionality in the distance where that supply is, and optionality in the time when to use it. And if you look at supply optimality, there are many different sources. If you look at those sources, as said, there's some that are better from a climate perspective than others, so we encourage everybody to move to the right there. And at the same time, you have the grid that is the mechanism to connect to all of this. And then batteries are now appearing in many places, not just in individual smaller configurations, but literally grid batteries. And so the connectivity is through the grid itself. And you see capacity and price arbitrage there on the sources, but you can also see on against the batteries. And sometimes batteries are built right in the field. You saw that with the earlier picture with the solar cells. And so that all helps with pricing the peak arbitrage.
And last but not least, you know, the new compute centers say, 'Well, you know, I don't have that much time to get permits to connect to the grid. Why don't I build a private grid? Maybe I can even build some nuclear power.' And actually, they're investing in that. But they also have essentially brought back some old nuclear power plants, like Three Mile Island, for example, and they have big parks of batteries. And so their own little grid can temporarily work, but really, they would like to have the reliability and the broader choices of the overall grid. And so you can see how important batteries are in the middle of this picture. But there's another reason I wanted to talk a little bit about batteries because this is what has happened to the cost of batteries in the last 10 years or so. It's not that long, actually. And this is the amount of storage capacity that has been added during that time. And of course, you know where I'm going here. My favorite curve. Let's just try it one more time. And isn't that amazing? That is called success, of course. But that is actually not the reason why I'm showing you this. The reason I'm showing this is this is a great, great example of technonomics. The tech, because it has brought down the price of these batteries through new technology. But the anomics have further brought it down because of the learning curve when you actually manufacture a lot of them. And that learning curve itself is valued very, very highly because it is continuing, and in my opinion, the next 10 years, we're going to see much, much more of this. So, technonomics in action, it governs everything we've talked about, which brings me to the call to actions.
And there are really three calls that I would like to make against the backdrop. We talked about the exponential and the race forward. AI, it is a trigger point. It's a driver to take the opportunity and hand to make changes. The necessity to deal with the CO2 equivalent. So CO2 and methane, and the dual challenge that comes out of that—that has a high degree of urgency. And so below that, I would propose that there are three steps that we should absolutely take. They all fall under the same category: electrification as much as we possibly can. It's the most efficient form of energy once you use it. And the more we bring it into that form, the better. That means driving technologies such as evolving the grid and batteries and a number of others. Secondly, rushing forwards to clear clean firm sources of energy. And third, killing methane leaks right now as fast as we possibly can. Now, the way to get there, of course, is through industrial policy at a government level, at a world level. That is not always happening. And so I think the opportunity exists for leadership coalitions to be audacious enough to team up together and make changes for the future. And that is why our industries, the ones that you all working in, should team up and make this happen one way or another. The urgency is high, but the opportunity is also fantastic. And so given that that takes actually some foresight to make things happen, I would like to have the following call to all of us. It's a phrase that I've liked to use for many years, concluding every one of my presentations. And in simple terms, it says the following: They who have the brain to understand should have the heart to help and the courage to act. That is all of us. Thank you for watching and listening.
