Enhancing Energy Security In A Time Of Rapid Change: Perspectives From The US And Taiwan

>> Kharis Templeman: Well, we might as well go ahead and get started, we've got our quorum here. I'm Kharis Templeman, I'm the program manager of the Project on Taiwan in the Indo-Pacific region here at the Hoover Institution. It's my pleasure to have with us today Laura Smoliar, who is the executive director of the Stanford Taiwan Science and Technology Hub on campus.

 

She's gonna kick us off with some opening remarks and then we will go ahead and jump right into our first panel. We have a large audience online, I'd like to welcome you, wherever you may be, to our secret lair here at the George Shultz Building at the Hoover Institution.

 

And I hope you can hear us well and see us well, and we look forward to engaging with you over the next two hours, all right, without further ado. Okay. The floor is yours.

>> Laura Smoliar: Thank you, greetings to our audience here in person and online. I am Laura Smollier, the executive director of the Taiwan Science and Technology Hub at Stanford.

 

We are pleased to co-sponsor this event with the Hoover Institute's Project on Taiwan in the Indo-Pacific region. Co-chaired by the esteemed Senior Fellow Admiral James (Jim) Ellis and Senior Fellow Larry Diamond. And managed by Research Fellow Kharis Templeman, who is one of the first people I met on campus when taking on my role.

 

At that time, Kharis's pointed out how much science and technology, especially semiconductors, intersects national security and geopolitical issues. So began our close collaboration leading to today's seminar on energy, which is also at that intersection, as is artificial intelligence. We were joined by David Fedor, Policy Fellow at Hoover, who is part of the George Shultz Energy Policy Working Group, to put together today's program for you.

 

In fact, this is the ninth meeting at the Hoover Institute where the Taiwan Science and Technology Hub has had a seat at the table. And I think that it is important for the science and technology community in both Taiwan and the US, especially in today's challenging geopolitical climate.

 

Let me start by saying a few words about the Taiwan Science and Technology Hub at Stanford. We are a lean team founded by our chair, Professor Ying Zheng, who splits her time between Taiwan and Stanford, where she is an adjunct professor in Chemical engineering. We are fortunate to have our co-chair, Chemical Engineering Professor Zhenan Bao, our executive committee member, Electrical Engineering Professor Philip Wong.

 

And our Program Manager, Ellen Zhang, we are in our second year of operation and the first office of its kind on the campus of a Taiwan US University. We foster people-to-people interactions between Taiwan and the US With a focus on science and technology. We do that in a few ways, our events have been as large as 500 in-person attendees and more intimate gatherings of 40 to 50 with additional online access.

 

Drawing on speakers from the US and Taiwan, including our university president, ministers, premiers and vice presidents from Taiwan, Nobel laureates. And deans from Stanford and Berkeley, as well as leaders in the investment, corporate and startup communities. We have sponsored or co-sponsored 10 such events in the US so far, serving more than 2,000 people in our audiences.

 

We sponsor postdocs to come from Taiwan to Stanford in any area of science and technology. They currently are in the schools of Medicine, Engineering, Sustainability and Humanities and Sciences. They are Stanford postdocs when they are here, giving them full access to a rich variety of Stanford resources. Which is a very different experience from a visiting scientist, for example, we sponsor speakers from Stanford to speak in Taiwan.

 

We have sponsored or co-sponsored 11 such events so far, covering diverse areas such as healthcare, biomedicine, AI, quantum and human-machine interfaces. Over 4,000 people in Taiwan have participated in our events so far. We serve as a landing pad for visitors and delegations from Taiwan and facilitate connections for scholars and students.

 

We help students who want to have internship experiences in Taiwan or the US, facilitating their connections to programs. We are developing one of our own for 2025 due to the great demand on both sides of the Pacific. Today we take on the topic of enhancing energy security in a time of rapid change with perspectives from the US and Taiwan.

 

For example, how do we navigate the polarities of energy-hungry AI data centers while aiming for net zero, and how do we ensure energy security at the same time? We will have two panels, one bringing the US perspective and one bringing the Taiwan perspective on this complex topic. The first panel will be chaired by Admiral Jim Ellis and the second panel will be chaired by Kharis Templeman.

 

 

>> Kharis Templeman: And with that, I would like to hand the floor to the esteemed Admiral Ellis.

>> Admiral James Ellis: Well, thank you, Laura.

>> Admiral James Ellis: Thank you for this partnership, we're very excited about this, we see a real Boolean overlap in engineering terms between your efforts and our own. And so it's great to have this opportunity to gather to focus on some more specifics of the energy challenges.

 

Here at the Hoover Institution, we've had a long-standing focus on energy led by the late Secretary George Shultz, who was absolutely fascinated by it. And felt it was critical to our shared prosperity and security going forward. And that's the context in which Hoover began the conversation about energy well over a decade ago.

 

He was more present than perhaps he realized in predicting the centrality of energy and energy policy to our nation's security. And I would pause here to indicate that we define national security much more broadly than just the military context, we think there's a huge economic piece to that as well.

 

We believe there's an environmental element that has to be considered and that the national security writ large is what we're here to talk about. But we also are aware that in addressing our energy circumstances, we as Americans have in a real sense found ourselves in a place we've never been before.

 

With opportunities that we've not heretofore had, we grown up, many of us, if you're as old as I am, in a situation of energy privation. Energy shortfall, lines of gas pumps, spiking oil prices due to geopolitical situations around the world. We're now in a much more different and more stable environment from a United States perspective.

 

We find that there are opportunities to use our new found energy, if you will in different ways. We've had a long-standing interest here at Hoover and, First, defining what a national energy strategy might look like. Arguably, I would say we've never really had one. We've had various political administrations come and go, as is the norm.

 

Each of them brings a focus and a prioritization. And we have obviously a large element of individual elements or sectors of the energy environment, each of whom have their own needs and demands. And so we've tended to balance things or tried to in a more even-handed fashion. But in point of fact, we really don't even have an agreed to model for optimization, as my wife, who's a professor of engineering here at Stanford would tell me to highlight what should be the ideal mix for the opportunities and challenges we have now.

 

And how would one characterize that fairly for a longer term intervention. So we still believe that that's a goal that we ought to pursue. But we're also much more pragmatic, and we now see, as Laura has indicated, some real opportunities here to talk about the needs that are growing.

 

Heretofore, we've had a relatively flat electricity demand in the nation. We're now seeing clear indications, and we've got a number of panelists who can address that based on their personal experience of an increase in electricity demands. Depending on who you're talking to, the amount or the magnitude of that may vary fairly significantly.

 

We're hearing 2.5% growth annually, but others are saying no, it's gonna be closer to 5 to 6%. And even when you pulse the experts, as we have done in looking at the Secretary of Energy's advisory board, they even hedge a little bit and say, well these trends are gonna grow.

 

But you know, perhaps we can understand how to do this more efficiently. Perhaps, these data centers or the AI use can learn to be more efficient. Perhaps we can, rather than co-locate them, we can spread them and diversify them, which of course then requires infrastructure to connect them because this has to be a collaboratory device.

 

So there's already cautionary folks out there saying yes, we're seeing some increase, but it may or may not continue. So one of the things that we're gonna have to focus on today I think is the centrality of all of this, and what do we need to do to make this all work?

 

OpenAI has a paper out last week. Some of you may have seen it about the infrastructure blueprint for the United States. And in pretty dramatic terms, they lead by saying revolutionary technology drives advances in infrastructure. Capital flows determine where and how the infrastructure is built. These decisions determine whether a nation leads or lags in technological innovation, often with far-reaching consequences for economic competitiveness and national security.

 

This is why infrastructure is destiny. Well, that's a little bit different approach and a lot more forceful than we're hearing out of the Department of Energy. So where, as I'm fond of saying, in every conversation, there's always three sides to every story. There's yours, there's mine, and then there's the truth.

 

And so what we're trying to do here is get to that truth, or at least as much as we can in the time that's allotted to us. So I look forward to the conversation. We've got an exceptional panel that's joining us for the first phase. If I can find my assembled biographies here to make sure that I introduce them fairly.

 

One is a colleague here at Hoover and briefly a colleague in some board service elsewhere. Paul DeBar, former undersecretary of the US Department of Energy from 2017 to 2021, where he was the Department's principal advisor on fundamental energy research, energy technologies, science, and commercialization of technology. A real expert in the current state of play.

 

To my right here is Jared Dunman, co-founder and Chief Scientist of Startups and a non-resident fellow at the Columbia center on Global Energy Policy. He's previously served as the technical director for Artificial Intelligence at DIU, the Defense Innovation Unit. And so he has an understanding of what the needs are in the real term, in the near term for energy.

 

And then finally, I believe that Jen Huffstetler is with us, the Chief Product Sustainability Officer and Vice President and General manager for a future platform strategy at Intel. Obviously someone who's well grounded in the, in the current state of play in the technology industry and the energy needs.

 

So as part of my remarks I will ask a few questions and if appropriate, perhaps our panelists can choose to address them. But the first is back to the variance we see in quantification of future electricity and energy demands in support of our AI initiative. Is this demand growth real or is as we used to say in the old television ads, is it real or is it memorex, and so what's your state of play on that?

 

And to what extent is the Department of Energy right, could this be affected by technological changes in how we do large language models and the training of artificial intelligence and the like? Could we, do we? How likely is it in the near term we're gonna see a dramatic left or right turn with new technologies and new innovation that might significantly reduce those energy demands.

 

And really does that matter if we need to hedge anyway in case that miracle doesn't occur on the timeline that we wish? There are some real questions about the costs associated with this infrastructure, we've talked about? Who pays and how do we do that? Does this go on the back of the classic rate payers of our regulated or unregulated utilities?

 

We're seeing some new innovations as you've all seen with some of our technology companies hoping to get behind the meter and have direct access to energy sources, small modular reactors, the restart of the operative unit at Three Mile Island and these types of things are in play? But what are the financial constraints and the costs?

 

But then who bears those costs, and how are they born? And are they born fairly by the population and the rate payer populace? And then what about the generation technology? Can renewables continue to meet the need? If it has to be nuclear, then are we talking the Gen 4 plus small modular reactors, or are we going back as we're now seeing?

 

Some people say, well you know, there are economies in scale after all and Gen 3 large reactors really are the most efficient and lowest cost way to do that. If we can break the code from our recent experiences and go that. And then finally there are the visionaries that say well we can just hang on using fossil and wait for fusion to arrive.

 

Well, I've been doing that for 25 years, folks. I'm not sure that's a path to success. But you know, what are the generation options that confront us? Otherwise, we're gonna have to continue to build gas fired power plants and peakers as we call them, to provide this in the short term.

 

And that doesn't seem to be something we wanna do for a number of reasons. And then finally, what are the real barriers to success here in the United States? And again, in this first panel we're talking about America's policy not in an effort to say we are something that ought to be emulated, but so that our Taiwan.

 

Partners in this conversation can learn from our own experience as they, too, try and chart their own way going forward in their aspirational goal to really dramatically increase their involvement in artificial intelligence research and creation. Rather than the preeminent role they play just in producing the microchips that fuel it.

 

So is it access to capital that is gonna slow us down? Is it the permitting processes that come with this? And what are those barriers? I mean, there are permitting processes about everything, not just generation technologies, but about transmission technologies and pipelines to provide natural gas to natural gas cogeneration plants or whatever your generation model is.

 

All of these things come with permitting delays. As Paul would know, there are some companies in the country that have tried unsuccessfully to permit major infrastructure such as natural gas pipelines, and found to their surprise and chagrin, there were 60 entities that you had to get approval from and any one of them could say no.

 

And that generates tremendous costs, tremendous delays in rapidly responding to these crises. And then finally, what is the timeline we need to meet? How rapidly do we need to have this electricity delivered? And if that's the criteria and that's the metric, then we can take a look at what options might be most timely, which might be a selection criteria rather than that which is cheapest, that which is most sustainable and the like.

 

If the race is really on and we need to be able to do this quickly, then timeline may be a consideration. So again, broadly, the questions are, is the demand growth real? How do we finance? Who pays for the creation of these generation capabilities? Are there clear winners in any of the generation technology that's available now or likely available in the next three to five years?

 

And then what are our barriers to success? So with that, I'm gonna stop talking and start listening. And I'd like to begin with Paul because he's just number one on the list. It's not an issue of priority. But Paul, if you've got some thoughts on this based on your experience, we'd be delighted to hear them.

 

 

>> Paul M. Dabbar: Okay, great, Admiral. And as Admiral mentioned, we overlapped on a board of a power company that is the largest supplier of power to data centers in the world. So I was glad to get handed off Admiral Ellis' board seat on that. So we spent our respective time significantly on this.

 

Whoever's controlling the slides, you can go ahead and move to the second page. So there's a number of drivers. I'll hit a number of the Admiral's points about drivers for locating AI and data centers. When I was undersecretary, my program built three number one supercomputers in the world, and we were the first user of GPUs in the world for a big computer, which was at Oak Ridge.

 

And we rolled the dice on NVIDIA on their GPUs. And that turned out to be a good bet on behalf of the United States and obviously has taken off very significantly in just about six or so years. So the first point of location is energy availability. This is the one that most people are talking about.

 

A number of people, including President Trump in a number of his speeches, has commented that in order for the US to maintain AI leadership, that we need to double the amount of power generation in ten years. I'm not 100% for certain it's every power pool in every location in the country, but that's a pretty good estimate.

 

And certainly in Northern Virginia, that's certainly the plan. There's give or take about ten different locations that are basically needed by the various hyperscalers and other big uses of AI where they're trying to target 5 to 7,000 megawatts in each of those approximate 10 locations. That is a stunning amount of power.

 

5 to 7,000 megawatts is more than most states in the United States. And so the only way that you can do that, frankly, to get to 5 to 7,000 megawatts, really, is all the above. You have to kind of build everything to get to that sorta concentration in a specific location.

 

Self-generation, which was looked at when people were looking at 100 megawatt facilities, maybe could be doable. But when you're talking about 5 to 7,000 megawatts and you're talking about, give or take, 5 to 7 AP1000s of reactors solely dedicated to one point on the map, it's quite stunning.

 

In addition, the capacity factors are very high. They're north of 90% at the data centers, and therefore intermittent power without backup is just not doable for the customers, for the use cases. If you go to a data center now, if you go to 100 megawatt data center, which by the way is quite large, they have 200 megawatts of backup Rolls-Royce or Caterpillar generators.

 

So they literally have 200% backup of generators onsite, and they also have lead acid batteries on top of that. So the amount of value or the amount of backup that they want to literally have 100% uptime is they will pay for a lot for that sort of upside.

 

Admiral Mel also mentioned about availability of power, about efficiency. Every generation of chip that we would build at the Department of Energy at the big supercomputing would be energy-efficient significantly. The current systems are about 5% of the power, 95% less power than ones five generations ago. So that's a good metric of how much efficiency has already improved in it.

 

What I do worry about is that NVIDIA and AMD and others are not really focused that much on the next generation improvement because they're just so busy just producing the customer product as fast as possible. And I do worry about the efficiency improvement curve that we used to enjoy.

 

The second one is energy prices. Many people may not realize, but the energy prices are an incredibly important part of the location. I was talking with the Senate Foreign Relations Committee this last week about where in the world and advantages just based on prices as well as availability.

 

If you look at, for example, the contract,

>> Paul M. Dabbar: The contract data center companies like Cyrus One, they're basically a distribution utility in a large building. They take wholesale power from a big utility like Southern or Dominion, and then they distribute it to architecture, racks on the inside of the building, and they just mark it up.

 

So they're buying megawatt hours and they're selling kilowatt hours, so to speak, inside the building. And since they're not selling real estate, they're not selling square footage, they're selling kilowatt-hours. The lower the price of the kilowatt hours, the better. And so that's why you're gonna see a location in lower priced regions of the world.

 

And including in the United States, for example, you're not gonna see a lot of discretionary data centers in California. The power prices are way too high. There'll be some, but not the really big ones, not the discretionary ones. You're not gonna see it in Japan, you're not gonna see it in Europe.

 

And this is a really big issue for Europe right now because of their energy prices. The third point is certain customers want low latency or data sovereignty. So a lot of the reasons why you see it in New York or in Beijing or in Washington D.C. is that there are various either national security customers or financial customers who want very low latency.

 

The financial markets, the stock exchanges, they want low latency for their use cases. So you'll see them located certain places in the world for that. So even though London energy prices are high, you'll get a certain amount of data centers outside of London because of the London Stock Exchange.

 

Nothing discretionary, but what they need for that local side. You'll also see certain users for data sovereignty. You see this in Paris, you see this in Madrid, and so on for government customers or for making certain that certain data wants to stay in a certain country for one reason or the other.

 

So that's another location point. The fourth point is the availability of dark fiber and interconnections. The interconnectivity is important. So once again, Northern Virginia has a tremendous amount of interconnectivity, which helped set it up to be by far the largest current location in the world. The next one is a really interesting one that I find is about workforce.

 

So it turns out that a lot of the workforce for the design, construction, putting together the internal networks in the data center, the interconnecting of that to their architecture is a pretty specialized workforce. And it turns out, I'll give the European example here for example. It turns out that most people who graduate with comp sci, or applicable electrical engineering, or other sort of degrees from great universities in Europe, a significant amount of them move to the United States because the top tech companies in the world are in the US.

 

They learn the skill sets for those people who are part of the data center business. And if someone wants to build a data center in many places in the world, they fly Americans over who, and they may not be Americans, but people who work in America for American companies are the experts.

 

They fly over there, they set it up, they wire it up together, and then they fly home. And so the workforce is very interesting. US is a big sucking sound of talent for the tech industry, and that's certainly this area. Then there's availability of cooling and water because it obviously generates a lot of heat.

 

And then the last one that the admiral mentioned about capital availability. So a 100 megawatt data center is about $1.2 billion for the building. It's about two and a half to three times that amount for the architecture for the chips. And then you have to build a power plant somewhere for it to deliver 100 megawatts or a series of power plants for 100 megawatts with 100% effectively reliability.

 

That gets you to about 4.5 to $5.5 billion for each 100 megawatt. So going back to my 5,000 megawatt, and you want 10 of those sites as an example, you're just to go build one 5,000 megawatt large language model cluster in Indiana or wherever, you're gonna need about $200 billion of total capital.

 

And if you're gonna do ten of them, like the big Hyperscalers in total wanna do, that's about $2 trillion. So that is lot of money. And so you're seeing investment banks setting up financing structures for the debt components. The Hyperscalers, obviously have a lot of cash and infrastructure funds.

 

It's far beyond what national governments can spend, so the private sector is certainly gonna be driving this. That's it, Admiral.

>> Admiral James Ellis: Great, Paul. Great insights, and thank you for that. And we'll move very quickly to Jennifer next, she's on the demand side of this. And so perhaps she can kind of align our thinking on what are the real priorities that the corporate leaders as they craft these strategies that Paul talked about, what are they thinking about, and where are they willing to trade?

 

In other words, often we, we hear that, well, I want all this energy, but I want it emissions free. Well, if it's a question of one or the other, how do you, how does one think about that? And, and so I'd be very interested in how the private sector is, is thinking about this as, as they're gearing this up, they have to get this right.

 

I mean, as Paul indicated, this is not going to be built by the government. This is not going to be, we can't subsidize our way to success in this area. It's going to have to come from the, from the private sector. And as John Rowe once told me decades ago, in conjunction with the nuclear industry, at the end of the day, it's a business, it's not a religion.

 

And so how do we think about this Jen, or how should we think about it?

>> Jennifer Huffstetler: Yeah, I apologize. My loading of the slides that I would share doesn't seem to be working, so I'll just be descriptive. On your first question around the demand, is it real? The answer, obviously, it's yes.

 

We're all experiencing this tremendous growth that's happening. But some of the data points that I want to put out there to share is, you know, and from 2015 to 2022 compute demand grew 340%, but the energy demand only grew 45%. Nested in that is that innovation factor. And Paul mentioned some of it, right?

 

In terms of what's happening inside the chips, the chips are getting more efficient. Intel did a summary where we just looked from 2010 to 2020 how much data center energy efficiency was delivered in that decade. What if you never built another chip? We estimate. From the hardware alone, it was 1,000 terawatt hours saved in that decade.

 

The thing about the way these solutions run. So I wanna make sure we talk about both. Cloud computing is still growing at a 20 to 30% CAGR to run everything in our society in addition to this AI growth. There's the software layers on top of it and that also has had a lot of innovations in these last 10, 15 years, including virtualization, containerization.

 

And I like to use the example back when the Apollo went to the moon, they had very limited hardware resources and that software was tightly coupled. I actually think one of the innovations that we have ahead of us is having AI to help us solve for a tighter coupling, to help make more efficient software.

 

One of the startups that we work with in the UK, they can make the energy efficiency of a software program 20% better by going in, evaluating the code that's used and auto replacing it with sections. That's just one example of where you would get efficiencies inside that software stack.

 

There's other areas in terms of the data types that are used in these large language models. There's one I will just give an example of, is a one bit large language model and this is a slightly less precise model, but it's nearly as accurate, and it can run 2.7 times faster than the current solutions.

 

And it can use 72% less memory, which when you think about the way a computer works, it's getting more done within those limited resources is the key. It also uses 94% less GPU energy. So this is just one innovation that's already out there. It's being tested in research groups from Microsoft and beyond.

 

And you start to think about chip innovations, software innovations, these new data types. This is going to come and help to bring down, and in my opinion, Intel's opinion. But you're gonna level off what happens because you're going to now have AI can help to lay out a chip better, it can help to write the software more efficiently.

 

We already know today that across every vertical industry, AI can be deployed today to lower the energy consumption that can be in a building's lighting usage based on traffic and the cooling in Google's data centers alone. Way back in 2016, they used AI to apply a ML model to just the cooling portion of the data center, and they saved 40% of the energy of their cooling systems.

 

This type of technology is not very widespread today. So when we think about the US as a whole, how do we help to accelerate the deployment of these technologies, at the end of the day, it ends up decarbonizing the buildings, decarbonizing different industries. But you need to go first into pilots and then into full deployment.

 

Intel's done some pilots with local carriers or telecom, where you're basically monitoring that mobile traffic and throttle down this equipment when it's not needed. So I think there's a lot of opportunity in that space as well. And then lastly, and I'll hand the time over to the next person in the utility sector.

 

I think this is an area that we're not spending enough time thinking about or talking about why technology exists today to help them out of the current deployed generation get more efficient at using it. But there's a lot of barriers and hurdles in place to deploying that technology.

 

You mentioned some of it. Who's gonna pay for it? Is it going to be resilient and are we gonna be able to keep the lights on? A lot of those things have been proven, but there's still resistance. And so I think finding ways to deploy existing technology more quickly, we can get more gigawatts out of this current infrastructure that we already have.

 

 

>> Admiral James Ellis: Great, Jen, that's fantastic. You've teed up an important concept, I think, here. And you know, with all of the demands and the requirements and the infrastructure and the challenges and the regulatory uncertainties, we tend to view AI as the problem. But in reality, AI is going to be a huge part of the solution in order to help us get this right.

 

And so I think that's an important takeaway. The other thing, as I reread the missive that I was quoting earlier from the Department of Energy, they're a bit dismissive. They're not sure that we have enough visibility into the private sector planning for new model training. And they see the possibility of speculative and duplicative requests for data center capacity that may go unfulfilled, and the like.

 

My point here is you have a huge incentive in the private sector to get this right, because you don't wanna buy more than you need. You don't wanna fund the creation of generation, that's unnecessary and the like. And so I know Yogi Berra told us that predictions are difficult, especially when they concern the future.

 

But you talked about the leveling perhaps of the demand. But in the near term, are you able to see what timeline that might be on or as IT levels, I'm assuming it's going to level at a stabilized rate of growth. And we're now talking about the second derivative.

 

And now, I'm wondering about the first from my engineering background. Do you have a sense of how real, how close that horizon is and how accurate the forecasts are that you and intel are using to plot the course going forward?

>> Jennifer Huffstetler: Yeah, I think you're absolutely right. There's an insatiable demand for compute.

 

We have so many problems and challenges yet to solve, and we're able to get those answers even faster. So the slope is always going to be up and to the right. It's at, what is the degree of that slope? I think it's just a couple of years out in the three to five years out that we're gonna feel this bigger constraint because there is a lot of generation coming online in the longer term.

 

But what I forgot to mention, in the design of the chips, when you know the workload so well, you can design the chip for the workload. This is what Apple is doing in their products, this is what Microsoft is doing in their Surface laptops. They also have new chips for AI.

 

Amazon has chips, Google has their own chips. And that provides much less waste than a general purpose processor, right? By definition, it's doing a lot of things well, but nothing great. And so when you start to get these ASICs or these purpose built designed chips for the workload, even you could envision those could be designed for.

 

Pharmaceutical workloads. It's going to become easier and cheaper to do that. And you see many chip companies out there are attacking specific verticals that becomes part of this total compute envelope where we, we call it the build your own is going to be always and financially incentivized. I heard you ask a question.

 

Are these merchant suppliers, which I would call the AMDs, the Intel product side versus we have the foundry side where we're building everybody's chips, are they incentivized to make these more energy efficient? Not as much as those hyperscalers that are building their own because every penny they save is back to their bottom line.

 

 

>> Admiral James Ellis: Great insights and very much appreciated. We're now gonna go to Jared Dunnmon. Jared, it says here on my paper DUI, I hope that's the wrong source of letters. I think it's DIU.

>> Jared Dunnmon: My mom makes that mistake.

>> Admiral James Ellis: Yeah, I know. Yeah, I bet her friends wonder what is her son doing for a living.

 

But in all seriousness, you've been part of forward leaning, if not visionary organizations. You've seen, you've attempted to peek over the fence into the future, look behind the curtain, what do you think from your perspective based on the conversations we've had is a likely direction? And what are some things we ought to remember and perhaps things we ought to be cautious about going forward?

 

 

>> Jared Dunnmon: Absolutely. And I'll try to hit on some of the points that you brought up as well. So my particular perspective, which is where this comes from, as Samuel mentioned, the kind introduction. I spent some of my life building energy systems, my PhD was building things that blew up and made energy.

 

I did a postdoc in the AI lab here at Stanford, I came from kind of the ML systems community. So thinking about how do you take things, run them more efficiently, you know, take models and be able to get more out of your hardware. And then I spent some time in the private sector doing that and then some time on the government side, on the AI side of the house.

 

And so my, the perspective I'll give you comes from, comes from that combination of things to start with on the demand side. I think Paul and Jen hit well a lot of the major trends that are interacting. And I'll give you my gestalt on how I think that that complex interaction is gonna play out.

 

But just to lay it out, there are companies on the training side building bigger models. They're serving those bigger models to larger numbers of users and that requires more and more and more energy. Furthermore, as Even as those models get cheaper to run, I would be very surprised if you didn't see something similar to what you saw in the 1800s with coal.

 

If you ever remember from Jevons Paradox from economics, as folks thought the cost of coal went down, it turns out that demand really, really, really went up. And this is, in this case, you have a situation where the price of a service goes down. And it doesn't mean that just like costs go down, it means that people use more of it.

 

And so you're competing with that because I can tell you, as someone who sits around writing code, sometimes there are services that I wouldn't pay 100amonth for, but I'll pay 20 or I'll pay $10 and those things are going to keep getting better. So you're competing with that as well.

 

As prices come down, I expect demand to go up. So the other piece that you're seeing is a lot of things that Jen was mentioning, right? You've got ASICs, you've got your chips that are designed specifically to run AI workloads efficiently from big companies as well as startups.

 

Have you ever seen Grok, or Samba Nova, or Sarah Brass, or I can take your pick. There's a bunch of folks who are designing chips for specific workloads or types of workloads. You've got data engineering going on to say, look, I actually don't need a general purpose model for a lot of workloads.

 

I can actually use a much smaller model if I make it a very specifically well scoped one. You've also got innovation going on on the architecture side, both in the US where you've got folks working on things like state space models and model distillation to make bigger models smaller and use fundamentally different kind of, I would say, physics approaches to model what's going on inside neural networks.

 

But you've also got folks who have, for instance, hit, I would say, if a lot of US companies are now starting to think about, wow, the engineering wall I'm hitting is power on the, in China. Actually, the wall that they've hit is chips in a lot of ways.

 

And so you're starting to see really good engineering happening where folks are able to get more out of the chips and the power they're using. So that's just to say the engineers engineer around the first wall they hit and you're actually seeing folks in different parts of the world engineer around different walls.

 

And for that reason, like you will see some more efficiency. So, so those are all the things that are playing out at once. My personal opinion is that you'll probably see at the end of the day, like the price going down causes more demand. Thing is gonna win out here, and you're gonna see increased power demand in the long run from this.

 

So, that's the qualitative assessment. Quantitatively, there's a metric that I like to look at or a measure that I look to like to look at. Which is at the end of the day, if you assume that when people buy accelerators, so they just buy AI chips, you assume they're buying them for a reason, not to sit there.

 

And so you assume that even if they're running at a cruddy capacity factor, say it's 50%, say it's a cruddy capacity factor. Well, right now, if you have clusters that are 100,000 GPUs, or say you move to a million GPUs and you call it, that's a 5 gigawatt cluster.

 

IFP did this analysis a couple of weeks ago. Right now TSMC forecasts that this year, I think 2023, they shipped 4.8 million accelerators. So that's like 25 gigawatts worth of demand. And if you believe those projections, by 2030, they're planning on having capacity for 40 million per year.

 

Per year, that's 200 gigawatts worth of chips. So even if I'm off by an order of magnitude, that's 20 gigawatts per year. So it's not small. And if, and if those projections are right, you're looking at like 200 gigawatts worth of just chips for you. So whatever the number is, I would argue it's big.

 

And it's big enough that the energy challenges are gonna be important. So that's that piece on the, on this, on the question of, you know, where are these things going to come, you know, where's the generation going to come from? What are the barriers? From a US perspective, let me just kind of frame it this way.

 

I think the numbers that I've seen, a lot of the projections that I've seen say, hey, we're gonna move from something, data centers being double the percentage they currently are of US power load. The IFP folks, I think they did some of the work, and they're like, hey look, by 2030 you're gonna have maybe 130 gigawatts of extra power you need worldwide to run AI workloads.

 

And this is, this is that estimate. And of that 30 gigawatts is likely to be in the U.S. that's a really interesting thing to say because the vast majority of that by the way, not in the US. What's also would imply the vast majority of that compute in that scenario would also not be in the US.

 

So to kind of some of the points that were made earlier in the US There are dynamics here where you have an aging and maxed out grid. We've lost as a country muscle memory for how do we increase the amount of energy we're supplying to the grid? The demand has pretty much flat lined for the last two decades, and we're now seeing refusals to cite new data centers because of concerns about electricity price increases.

 

If you just saw recently, FERC. Kind of denied Amazon's, attempt to go and buy a nuclear reactor for their use specifically because it was gonna take the price up of power for ratepayers locally by a pretty substantial factor. And so you're starting to see that not only is it hard to do, but there are regulatory and other barriers.

 

On the other hand there's also the risk for the hyperscalers, the hyperscalers want computer and the risk of an underbuild. And I think Most of the CEOs have said this, at this point they see the risk of an underbuild as worse than the risk of an overbuild. And so to power that I think it was that earlier.

 

There's certainly an all of the above aspect to this. But if you cycle through the options, solar and wind have an intermittency issue that hasn't yet been kind of economically solved with storage at the scale that we're talking about here. Nuclear, there's obviously a lot of attention on nuclear right now because in the intermediate term you can get the scale of electrons you need out of that.

 

And, and then the, the assumption is that if those, none of those things work, you're going to end up in a world where you're just in the US you're going to build gas and you know, and then if you just don't build those things, as we said earlier, the data centers are going to move.

 

So what do you do here? I mean there's a couple of things to do. And on the, on the, from a US perspective, one would be think about the demand management side. All the things that folks have said earlier and that I mentioned earlier about how do you just reduce demand.

 

Like you should do those things when you can. Number two, you should think about how do you make the grid better able to function as a higher capacity network. So that's like improving the kind of the capacity of conductors and various other things. You probably want more firm power like non intermittent firm power.

 

And a lot of that looks like in my opinions, yes, there's gas, but there's nuclear and advanced geothermal like the non-gas options. You want utility incentives to be moved to not just be toward building new generation, but also making more efficient use of current generation. And then you got to really think about how you're going to interact with partners and allies like around the world.

 

So you've got folks in areas like the Middle east who have A lot of capital, who have a lot of interest in moving their economies towards, you know, towards some of the, you know, some of the growth drivers that AI brings. And right now, I mean, folks, I'm sure you've seen it in the news, there's a really important tension between how do you safeguard, for instance, US intellectual property versus make sure that folks who would otherwise go and be part of a different computing ecosystem are part of a US computing ecosystem.

 

And so there's a lot of interesting dynamics there. And then there's questions also about can you be, if you get this right, does it make sense for partners, so for places like Japan and others, whether you have some prices that are higher to actually send their workloads to the US where there can be cheap energy versus trying to build data centers over in those places.

 

So this is just say there's a lot of, there's a lot of complex interaction. But I'll start, I'll leave on a, I'll end on a note that I think is something to think about from the kind of broader with my kind of security hat on. If we don't build the power necessary to power these data centers in the US I think it is almost certainly the case.

 

They will be built somewhere, it will just not be here. If I were thinking about what belt and Road 2.0 would look like, if I were sitting in Xi Jinping shoes right now, I would be thinking, gosh, I might be a little better than the US building nuclear reactors right now.

 

And if I can get my, and if I can, my chips aren't like the best in the world, but like, they're functional certainly for inference. If I walk up to a third party country and say, let me sell you a data center reactor pair, and then, you know, that's, that's great for you.

 

And then you're relying on me for both your power and your AI. That's not really a geopolitical situation you wanna be in as the US. And so there, I think they're not just interesting academic questions about can we get this right? And how do we get this right?

 

There's actually a pretty substantial security imperative to get this right for the US and our broader set of partners and allies. So I'll shut up there, but that's, that's kind of the thing that I think about on the horizon.

>> Admiral James Ellis: Great, great insights, Jared. Thank you, I expected a visionary, and I got one.

 

So thank you so much. We're going to go one more cycle around our panelists here, and I'd like to, to perhaps change the direction of thought here and get your ideas on how do we make all this happen. I mean, we've described a very complex set of interrelated challenges, from generation to efficiencies to transmission to siting to regulatory policies and they're like who's gonna do this?

 

Normally in these kind of conversations, you get a hand wave and somebody can't resist mentioning the Manhattan Project. Do we really wanna turn this over to government? I don't think so. I mean, but how do you see Paul and Jen and Jared, perhaps a new set of relationships involving both regulation and more collaboration and consolidation and integration of efforts, rather than the flat out competitive environment that we're used to seeing in our economic model here in the United States.

 

You know, who needs to talk to who and who needs to get together? My grandfather used to tell me, Jimmy, it's easy to be a clown if you don't have to run the circus. Well, now we've got to figure out who's gonna run this circus, and then what do those relationships need to look like?

 

One of the things that I'm model, I like to use is, if you build an organization, it ought to take in. You do it like you. Where you put the sidewalks on the college campus, it's where the paths are worn in the grass. It's people that are already talking to each other out of necessity.

 

And how do you formalize that, speed it up and make it work to our shared efficiency? So this is your opportunity. If you were a king or queen for a day, how would you recraft, if at all the current players in this, in this demanding ecosystem to ensure that, as Jared said, we come out leading and not following?

 

Paul, I'll start with you.

>> Paul M. Dabbar: Well, given I'm a bit in the business of scraping President Trump's speeches and trying to think about how to turn it into execution. And he's focused, he said very clearly we need to double the amount of power plants in 10 years, right?

 

Which is dominion and others. And so how do you do that? I would argue there's plenty of private sector capital notwithstanding 200 billion a site. The private sector is absolutely, I thought I was just down at some friends with one of the major Wall street banks. There's money, okay?

 

There's money between the hyperscalers and the debt markets that's doable. It's really about siting. It's really about clearing the way for locations. And if you go around the country, there are already ones being proposed. Right. People are discussing ones in West Virginia and Tennessee and Indiana and Illinois because of nukes.

 

So, Admiral, to your point, you know, I think we already, we already have the lead about where people, where the private sector thinks in the country it is best located. But still, it's not good enough in terms of actually executing on the siting and government allowing that to happen.

 

And I think. One thing was discussing with Eric Schmidt was pick out ten sites, regions in the country and just wipe clear accelerated reviews by EPA. If interior needs to come in for land by FERC and there's different mechanisms in the government to go to FERC and say focus on this, put this highest on your priority to the president could actually do that.

 

And so it's really about paving the way towards all those approvals, all those sightings for all that infrastructure and have several in the country. They're already identified and I think that's a possibility.

>> Admiral James Ellis: Great, thank you, Jen I used to work for the guy who would tell me, Jim, you're in a crisis when they waive all the rules.

 

We haven't waived any rules yet, in fact, we still keep piling them on. And as Paul noted, from your perspective, who are the key players and who needs to be a part of this? Not the coalition of the willing, but a coalition of the ready, willing, and able in order to get this done for America.

 

 

>> Jennifer Huffstetler: Yeah, well, I think even if we just look over to across the pond in Europe, they had an energy crisis. So I was at a conference and the EU Energy Commissioner wiped the permitting process, right? Accelerated deployment so that after the war on Ukraine, they were able to have energy independence and that worked for them.

 

And I think that concept of a crisis I've just been thinking about since she said that a year and a half ago, what is it going to look like for the US to have a crisis? And I think Jared really touched upon it, that this is really about our national security, our energy security for the future, but our nation's security.

 

And is that something that everybody is willing to lean into? Can we craft that story so cleanly that everybody understands it? Similar to what Paul said, I think the private markets, I've been in many conversations with different firms on the financial side around the globe. The money's there, the developers are there, this is happening globally and definitely in the US the area where I think everyone is struggling is within getting the transmission, the permitting for their sites.

 

That's the area where there's just too much sand in the cogs, right? And they're struggling to move forward. So I think if we can find ways to create incentives or to, you know, fast track those things, that's what is most needed because you've got enough hungry and willing participants.

 

So I mean, where I actually think we're gonna help accelerate the energy transition by this moment and with the hyperscalers pushing for their 24/7 low carbon energy that they've committed to. Now, how do we as a nation step up to help make that happen? And it's got to be in some policy changes or incentives, even if it's for the short term buildout.

 

And it has to be beyond just a dominion. You know, I think also for our national security, we need a more resilient solution with geographical diversity as well.

>> Admiral James Ellis: Great, well, thank you, great insights. Jared, we're gonna give you the last word before we wrap up the panel, but you've been a part of an organization that attempted the Caspian task of trying to reform the Department of Defense Procurement System, and.

 

You know better than most people in this room the challenges that come with bureaucracies and the like if you were in charge for a day, what would you change? How would you restructure it? Everything from an AI Czar, heaven forbid. Don't ever let anybody call you a Czar because the nobles ignore you and the peasants kill you.

 

All right, so what would you do to change the structure of this effort going forward?

>> Jared Dunnmon: So to answer that, let me preface it with a little bit of framing. I completely agree with the points on capital, the hyperscalers have voted with their feet. You've seen the amount of money that they've been willing to spend on, particularly on nuclear is a great example.

 

They're doing that because they need the power and they in many ways need it pretty quickly and they're willing to pay a premium for it. I think what Paul said about the capital markets is exactly right. You've even seeing folks collateralize loans with these chips. So, like, there are ways to make these things work.

 

I'm not actually as that concerned about the capital. And that's a new thing. To be clear, we used to be concerned about the capital for the energy transition. Like, I actually think that in many ways that's like an exciting thing that I can say. I'm actually not that concerned about that.

 

The things that I'm concerned about are actually very much the items that were that fall under, you know, where could we trip ourselves up? Where can we shoot ourselves in the foot? And those are things like letting bureaucracy like permitting and environmental reviews and things like mandatory hearings that don't need to happen cuz everybody agrees, but they still have to happen and the NRC still holds them, right?

 

Like these are things that don't need to happen and could elongate these timelines to a point where actually Capital says, like, actually, no thanks. And we can't have that happen. We also from a storytelling perspective have to get this right. So the way that I would, I would really focus on this is a couple things.

 

First of all, you gotta hold rate pairs harmless. And you saw that with Amazon. You saw that in Pennsylvania. If you don't hold rate pairs harmless, you are gonna get blowback and it is going to stop you in your tracks. And there was literally a control experiment where you look at Three Mile Island versus the Amazon pencil project in Pennsylvania.

 

One of those things is kinda moving forward because it wasn't harming ratepayers. And the other one, FERC said, wait, hold on. Like, don't do that. So I think that's a pretty clear thing that one has to do. I think we also need to make sure that from a, from a, from a basic perspective, there are some fundamentals that we need to be investing in.

 

So I'll call out DOE, DOE issued a multibillion dollar kinda ceiling program relatively recently for putting money in the nuclear fuel cycle. Like to have some of the nuclear options, like you have to have the fuel cycle. And so those are particularly for Haley, right? So you have to do that.

 

So some of the, some of the places where government I think really can help is making sure that some of those foundational, some of those, some of those foundational investments continue to continue to move forward. But the thing that I think is actually maybe the most important determinant of this particular speaking from kinda time, and I'll give you an anecdote from some of my time in government that I'll relate back to this.

 

When I was at DIU, I was directing the AI portfolio, helping to stand up the Energy portfolio. One of the Energy Portfolio's first projects ended up being hybridizing most of the Army's vehicle fleet. And if you go and ask some folks in the army who you know are great Americans, like, why you say, hey, I want to hybridize your vehicle?

 

They'd say, like, why, why do I need that? And you can tell them, well, it's going to help the climate, and they'll say, okay, great, that's great. But I have other things I'm thinking about, like getting shot. And what you then go and say is like, okay, well, hold on.

 

All this other stuff aside, when you're sitting in the desert. In a vehicle, you're gonna run the air conditioner and you're gonna be running the diesel engine, and it's gonna make a lot of noise. Would you prefer that and someone find you and shoot at you, or use the battery and run a silent air conditioner?

 

And they're like, I want the latter. And that accomplishes some of the broader kind of energy transition objectives, but purely by a security argument. I think that's what has to happen here for a wide variety of reasons that includes the fact that various folks are gonna have to pay for things.

 

Capital is gonna have to pay for stuff. I think the story here is AI leadership for America and its partners and allies around the world depends on energy. And if you don't get the energy part right, you have a national security problem. And so, from a bipartisan perspective, if you ask anyone in D.C. they will say, we do not want America to lose leadership in AI.

 

It's like, okay, great. That means that we have to fix some of these underlying issues on the permitting side, etc, so that we can have the energy to power that future. And I mean, I'll end on a hopeful note, which is, I think that that imperative, if communicated correctly across the aisle, across a bunch of different kind of segments of Washington, you actually have a chance to use that moment to get the energy transition off in a very, very productive way by using these of capital, this kind of national security imperative that's driven by some of the AI pieces.

 

So I think we have a really big opportunity, and we just need to not mess it up. Great, well, I wanna thank our panelists, Paul Dabbar, Jennifer Huffstettler, and Jared Dunnmon. They have done exactly what we wanted. We could not have had better contributions to tee up the issues and the possibilities.

 

I would only go back to something that David Fetter and I wrote eight years ago for Secretary Schultz. In addressing our energy circumstances, we as Americans find ourselves in a place we have never been, with opportunities we have never had and choices we have never before been offered.

 

It is a transcendent opportunity to redefine energy security in terms beyond our own domestic needs and embrace the leadership possibilities and imperatives in creating a global framework within which we and our partners can prosper economically, progress societally, and participate in shared energy security. So I hope that sums up our positions going forward, and I thank you all for your participation.

 

Thank you.

>> Kharis Templeman: All right, in the interest of time, we're running a little bit late. I think everybody's aware we've got some refreshments in back. If you need a break, I invite you to do that on your own time. What I'm gonna do actually, is go ahead and transition us over to the second panel where we're gonna shift gears a little bit and talk about Taiwan.

 

And I wanted to just make a couple of notes here. When we were putting this event together, we wanted to accomplish a couple of goals. One was to highlight for our Taiwanese colleagues and friends the really rapidly evolving conversations around energy and AI in the United States. And I think you've done that with aplomb on the first panel.

 

I've learned a lot in the last hour. I've taken in a lot and I hope our audience online has as well. The other goal we wanted to highlight was to make our American friends and colleagues aware of the conversations around energy and AI in Taiwan. And Taiwan is in an interesting, unusual, really unique position in that it's the source of most of the cutting edge chips that are required for a lot of these large learning models.

 

And it also faces some very steep challenges in terms of energy production going forward. And so, a lot of what we've discussed in the first hour has to do with steep energy demand increases, the challenges of trying to increase renewables versus fossil fuels, the impact on the grid, all of these challenges Taiwan and the United States share.

 

But I wanna highlight one thing that is different in Taiwan. Taiwan, unfortunately, has another existential challenge, which is the threat from the People's Republic of China right across the strait. And so when we talk about energy security, it's not just about winning the tech race with the People's Republic of China, it's also about keeping Taiwan free of PRC control.

 

And Taiwan is actually quite vulnerable in many ways on the energy front. Over 80% of its energy needs are imported fossil fuels. Over 40% of that, we'll have a presentation in just a moment. You can correct me if I get these numbers wrong. Roughly 40% of that is just LNG reserves.

 

LNG and the reserve on the island is somewhere between 8 to 12 days only. And so if there were a disruption for whatever reason of LNG imports, very quickly, within two weeks, Taiwan would start to feel the pinch. And that is an alarming place to be for a place that is so critical, so central to the development of AI in the world.

 

So we have, and actually, Jim, can I steal back my bio sheet? We have three wonderful panelists joining us today. We've got one, there we go. Thank you. We've got one with us in person, Zhang Yusheng, good friend and colleague, friend of our program, a great partner during our visit in September when we had a Hoover delegation in Taiwan.

 

We're delighted to have him with us in-person tonight. And among the many hats he wears, he has been an independent consultant within the energy industry in Taiwan, and he's also very knowledgeable about Taiwan's state owned energy monopoly, Thai power. So we're gonna proceed in order. Let me introduce the other two panelists though, first.

 

So we'll start with Dr. Zhang. Second, we have Tsaiying Lu who joins us virtually. It's great to see you, Tsaiying. Thanks for joining us. She's in Taipei right now and she is with a fairly new research institute called the Democracy Society and Emerging Technology Research Institute. She holds a PhD from the Maastricht University in the Netherlands, and she and her team at DSET have actually done some analysis of Taiwan's transition to low carbon energy future and she's going to speak about that second.

 

And then, third, another friend and colleague, Gwen Wong Reeves joins us also in Taipei this morning. Actually, you're in Manila, aren't you?

>> Tsaiying Lu: Well, actually I flew back to Taipei early this morning, arriving at 6 AM because there was a typhoon coming this way. So apologies and mumble nonsense later, but excited to be here.

 

Thank you all.

>> Kharis Templeman: Okay, well, thank you for your commitment to this. We really appreciate it. So Kuor is currently the Engagement Director for GE Vernova in Taiwan where she is responsible for establishing and driving GE's advocacy initiatives. Engaging with local and central governments and other stakeholders on important public policy challenges, as well as advising GE on a broad range of regulatory issues.

 

I should note prior to that she held a role as the Senior Director of Government and public affairs at AmCham, the American Chamber of Commerce in Taipei. And so she's not just an expert on GE issues, she actually knows the Taiwan business community quite well and so we're very glad to have her with us today as well.

 

So without further ado, let me turn the floor over to Guo Xin.

>> Kuor-hsin Chang: Thank you Karis and Elmir and all the gentlemen here and it always my pleasure to be back in your institute and thank you for the invitation. So as Karis said, actually I was chairman of subsidiary company of Taipei.

 

That's why I know Taipei well and inside out. So basically I'd like to share my view about Taiwan's Energy Security today, okay? There are four areas I would like to cover, one is Taiwan's energy mix and the policy and also the projected power consumption looking forward from 2024 to 2030 and the challenges of meeting net zero and the energy security in Taiwan.

 

And also measures are able to address these challenges. So Taiwan launched the energy transformation parish in 2016, one, there are four major areas. One is expedite renewable energy deployment. The other one is to phase out a nuclear power plant by 2025. Also increase gas fire power plant and reduce coal fired power plant.

 

And the goal, original goal was to reach 20% of the renewable energy by 2025, however, that is pushed back to November 2016. So Taiwan's energy policy and mix after 2025 as you all know, Taiwan has a new administration this year in May, but currently there are no major policy changes for the new administration yet.

 

So for last year the electricity consumed in Taiwan like Karis mentioned, coal is more than about 42%, natural gas about 40%, nuclear about 6%, renewable about 10%. So the power consumed in Taiwan last year more than 80% are from fossil fuels which is a lot and the energy mix currently is projection to be in 2030 is 50% from natural gas and 30% from renewable, 20% from coal.

 

For nuclear energy policy for past administration they want to phase out by 2025. For the new administration recently said they have an open mind regarding nuclear power. However, for Taiwan to adopt new nuclear power again, three conditions need to meet, one is nuclear safety and waste management needs to be well planned and also the society need to be able to reach a consensus to use nuclear power.

 

And for renewable energy policy, Taiwan is going to continue the deployment solar and the wind. In addition it will add renewables such as geothermal, hydrogen, small hydro, biomass and this. And the government also emphasize they want to make a lot of effort in the demand control through power saving, demand response and energy management.

 

So the projected electricity demand from 2024 to 2028 will grow on average of 2.5% just like Adam mentioned earlier. And from 2024 to 2033 the growth will be 2.8%, this is from government data I guess. So from this data you can see the power demand increase much more after 2028 compared to before 2028.

 

And this IT demand increase includes expansion of data center, semiconductor and server industry due to AI and also include the increase in the electrical vehicles. Taipei estimated the power demand from data center will grow from 240 megawatt today to 2,340 megawatt in 2030. And Jared mentioned in the US you know in 2030 there will be 30 gigawatt from power demand for data center.

 

So this is less than 1/10 of the US but compare Taiwan's economic size, this is dramatic, it's not small either. On the surprise side, if all the new energy development is on schedule, the government project, the ninth reserve margin ratio can be about 8% from year 2024 to year 2030.

 

Here we only talk about night peak, we don't talk about day peak because in Taiwan in the daytime they already have enough solar energy to cover, so Pao Yu Xi in daytime is Is not as critical as in the night time. Challenges with net-zero effort and energy security, so if in 2030, 70% of power generated from Taiwan oil from fossil fuel, then common reduction could be a big issue.

 

So there are measures need to be taken to reduce the common reduction. For Shiva carbon capture and the storage need to be adopted. However, for existing power plants, this might not easy to do due to space limitation. And currently in Taiwan, they are trying cofiring hydrogen with natural gas to reduce the emission.

 

And also cofiring ammonia with coal to reduce the carbon emission. However, for this cofiring effort, the ammonia or the hydrogen, they can burn only have a certain percent, maybe below 30%. So they were still common emission exists. There are challenges with renewable energy development. For solar, Taiwan have space limitation.

 

The other thing is, like Gerald mentioned earlier, any big development need approved by the local government and it could be time consuming and oftentimes very complicated. So this is a big issue in Taiwan right now. For wind because of the cost increase, some of the developers, it's not easy to find off takers, and that will affect them to get financed for their project.

 

So the schedule also delayed. Geothermal sounds good. But currently, in Taiwan, they only have a very limited power used from geothermal hydrogen if it is not green, how good it is for carbon reduction. And in addition, as Ellis mentioned earlier, Taiwan need to enhance its grid so that it can support more renewable energy connection, at the same time maintain the grid stability.

 

And also, as Ellis mentioned earlier, more than 98% of natural gas used in Taiwan are imported. If 50% of electricity will come in from natural gas, this will pose a security threat to Taiwan. Well, net-zero security will be concerned, especially we have a an unfriendly neighbor. And so some of the measures to address the challenges we just mentioned earlier.

 

We believe the energy mix need to be rebuilt to address net-zero effort and the energy security. And we also think nuclear power plant need to be reopened and the Maanshan nuclear power plant. The license of Maanshan nuclear power plant needs to be extended after thorough inspection and evaluation.

 

This is for the purpose to reduce carbon emission and also improve energy security. And for the current Taiwan's environment, I think Taiwan, in terms of energy security, Taiwan has a bigger problem compared to the United States. Because in the United States, using nuclear power is not a problem.

 

In Taiwan, it will create a lot of social debate. So looking forward, I don't think it is possible for Taiwan to build large nuclear power plant. So for small modular reactor, since the size is small, it is also flexible. I think this could be a good option for Taiwan to provide stable, common free electricity.

 

So I think Taiwan should engage with SMR development at this point of time and evaluate possible deployment in the future. And currently, Taiwan is also evaluating the possibility of purchasing renewable energies from Philippines. If this is feasible, then it can ensure Taiwan have enough renewable energy for high tech industry.

 

And as I mentioned earlier, Taiwan have a new government now and oftentimes Taiwan's energy policy is affected by politics. I think it is a good time for the new administration to evaluate Taiwan's energy policy through scientific evaluation, public hearings in the social debate. So that's all I have to share, and I would like to welcome any questions later.

 

 

>> Kharis Templeman: Excellent, thank you, Kuor Hsin, you've really teed us up nicely. And as our audience can see, there's some major contradictions in Taiwan's energy mix and its future, right? And so I hope to have a robust discussion after we get through the presentations here about what is actually possible in Taiwan politically over the next few years.

 

Next, we'll turn to Tsai-Ying Lu. Tsai-Ying, the floor is yours.

>> Tsaiying Lu: Okay, hi everyone, thank you so much for having me today. So I think I'm going to give a rather different approach from Mr. Zhang and then maybe we can discuss more later. And I mean, it is really an honor to be invited to this, and I will discuss Taiwan's journey toward establishing a resilient power system in details because Mr. Chang has already mentioned a bit.

 

And then I will give you a certain outlook of this and then we can discuss this further, and especially related on how Taiwan could support green and stable electricity demand for data centers specifically. So to start with, yeah, let me see if I can change to next slide.

 

Yes, so where and what kind of DCs the data centers are in Taiwan? This will just give you an outlook of this. So now, we have a range of different data centers from regional to local centers by companies like Chunghwa Telecom. This is a domestic company and we also have eASPNet Taiwan to hyper scale daya centers built by global tech giants like Google which established that in 2013.

 

And then we also have recent expansions by Microsoft Azure, and AWS, and more recently Wistron and Zettabyte. They are just Established to build Taiwan's first hypescale AI data center, which will be in Kaohsiung because of the problem that I will later on explain. So data centers actually can vary significantly in terms of their service types and also the benefits they provide.

 

So this also regarded to their energy usage apparently, so this slide shows the main types of data centers I think many of you might already know. And you can see how this is important for understanding different power usage for each type. For example, the red square that is highlighted here suggests that Google's data center in Taiwan right now, which is a hyper-scale data center.

 

And provide multi-cloud services to customers in Taiwan, and which suggested they will have higher power usage. And also their energy efficiency will tend to be higher as well because they're higher demand. And how serious is the problem in Taiwan? If it's not about the energy efficiency, because first, according to Google's 2024 data center impact reports, you can see that data centers in Taiwan actually have achieved an impressive power use effectiveness rating of 1.12, which is considerably better than the global average of 1.55.

 

So this shows that Taiwan is on the right track regarding data center energy efficiency. Because of you can see how many power they were used in different sectors. So for operating the IDE equipment, it will take 50% and other than that, you would need 12% for air management and also for cooling for 25%.

 

And this is also why people are so concerned about this, because you are afraid that this might cost a lot of energy by using the cooling systems too much. But according to the report, Taiwan's energy efficiency apparently is better, and so if it's not about energy efficiency or it's possible impact during peak hours.

 

Because what Mrs. Zhang just mentioned, that there is a variety of peak loads during a day, right, 24 hours. So you can see that from the data, we also see a single hyper-scale data center Taiwan can consume as much as 8.5 megawatts of electricity annually. Which only represent a very small fraction of Taiwan's total power generation, so only zero 0.021%.

 

And at the regional scale, it only claims 0.09% of central Taiwan's power generation. Because you can see during peak hours, we are afraid that maybe the data center will use, it will consume many power. That will decrease the electricity that generated in the central region of Taiwan and that it might hinder or decreases ability to transmit its energy and electricity to other regions.

 

But apparently, this seems not a very serious problem for now, or this is not a very serious problem for its future growth. So this is a prediction made by TaiPower which is a national utility. This is from 2024 to 2028. You can see in the orange bars, this suggests the AI data centers power usage prediction in the next five years.

 

And you can see there's not many increase annually compared to the blue bars. The blue bars suggested it's the chip industries demand. Actually, the future growth amounts might not be a very serious problem for TaiPower perspective, from Taiwan's perspective. So what are the challenges? So let's look closely to the challenges by using Google's hyper-scale DC in Changhua county and to see what's the challenges shared by Taiwan and also between Taiwan and the US.

 

So first, I think you can see the Google's existing hyper-scale center which is located in Changhua county which is in the central area of Taiwan. And this facility designed to operate around the clock, which is you will need 24/7 electricity, and also it will draw a lot of local renewable energy resources because it will need a green supply, right?

 

So it benefits from being near the Changhua Coastal Industrial park in the central area which is a clean energy hub in Taiwan right now. And the solar power facilities here have a 100 megawatt capacity, but we will see that even this robust setup will face challenges. Because apparently, according to Google's report again, it's just that Taiwan's DC basin only claims 80% of these electricity data by renewables.

 

So essentially, this is not enough yet, and due to renewable energy's inherent variability that Mr. Zhang also mentioned. That additional support through battery storage or hydroelectric resources will be essential to cover the gap. During the nighttime, you don't have the solar energy anymore. And battery storage and hydroelectric will matter in this scenario.

 

And another challenge is that there are transmission constraints when moving electricity from central to Northern Taiwan right now where demand is high in the Northern Taiwan, but local power generation is limited in central. So these kind of bottleneck restricts where the hyper-scale data center can be built. So recently, the TaiPower just announced that the DCs should not be built Northern than Taoyuan.

 

So you cannot build in the Northern center, because there has already a limited capacity to transmit enough electricity from the south to the north. So our team tried to give you a sense of how Taiwan's power flow looks like and this is a average pattern of this. So in average day, you can see how many power that was transmitted from the central, which is central Taiwan to the north.

 

And only during night time, you can see a certain amount of electricity will be transmitted to central area. Because there was to use the electricity to pump the water from the pump storage hydropower plants and to store in the upper end, what's that called? Yeah, the tanks and so that it can also generate electricity during peak time hours in daytime.

 

So you can see how Taiwan's Northern region relied substantially on the Southern Taiwan generating electricity. And we also shows a picture like this, so you can see how the transmission capacity in Northern Taiwan, such as the Xing Hongi Xing substations, they are already at high utilization levels. So Taiwanese to address these kind of transmission bottlenecks in order to support increasing demand Demand, particularly for the ascent, so these are the central issues.

 

So what are the adaptation strategies? I think some of John's presentation has already shown. And this just give you a sense of what this looks like in terms of the green enhancement program, and also the regional ICE integration program. So right now Taiwan has led out several strategies.

 

So you can see Reinhardt, this was planning for integrating more offshore wind energy and also solar power in different areas. And then it will directly send to stats parks and also low sensors and to separate it from the civil use electricity so that a certain malfunction won't cause a large scale blackout.

 

So this is the thing that the test that the Taipower is trying to do to decrease the risk of two centralized power system power grid. And this is also very important for the data centers to have original and stable power output. So and in terms of green electricity that the data center will need, is that we are also trying to increase our renewables, our capacity in the next five years.

 

So this is just a picture to show you how the timeline would be, and then finally we also need to cover the nighttime supply, right? So I just talk about how the battery matters, and also the hydrogen fuel cells matters in this scenario. So Taiwan also plans to launch regional hydrogen tech parts to promote its application for energy.

 

So right now, the data centers that could be located in Kaohsiung, which is in southern Taiwan, also in Taichung, the central Taiwan, they can benefit from these hydrogen fuel cells that will support your load during peak hours or even in the nighttime. These are the most efficient way to do that and also to secure a stable regional supply in the meantime, which I think is very important for Taiwan.

 

Voice way that towards a more disparity and also distributed power units and to decrease the risk of a large scale blackout. And to have the resilience to come back to have a stable power outputs in terms of even during the critical time. So there are certain malfunctions happening.

 

So these are my short presentation and also if you want to explore, you're welcome to email me, and also this is a QR code for our institution. And so I will stop here, thank you.

>> Kharis Templeman: Terrific, thank you, Tsai-Ying, I did want to ask a clarifying question. You had a table up there that showed the projected energy demand of the chip industry actually potentially falling over the next three years.

 

So declining relative to the demand in 2023, I think it was about 2/3 of the way through the presentation actually.

>> Tsaiying Lu: Yeah, that's right.

>> Kharis Templeman: You've got it right there. Yeah, what is that projection based on? Because what we hear outside of Taiwan is that there's just a surge in demand for Taiwanese chips right now and that that's going to lead you know a massive increase in electricity demand.

 

So is that backloaded or are we missing something here?

>> Tsaiying Lu: I think it is very important to suggest this is in our another report which will come out by the end of this year talk about the chip industries power usage. And what we think of is that Thai power is based on the what is the TSMC and other important chip manufacturers, they will submit their power usage proposal for them.

 

So they are actually based on a certain data and evidence, and this kind of decrease I would expect is because of recently many people talk about a higher power usage of the EUV tool, right? So there have been a lot of efforts that's been done by TSMC and its partner SMIC, they're trying to reduce the power usage of UV tools, for example, introducing the sleep mode.

 

So it's very similar to what AI industry and AI chief trying to do, is that you can just shut it down for a while, because if you are not using it in the processing flow and then only by that action you can already save 30% of the total power usage.

 

So this kind of efforts will be increased in the following year. So I think that's one of the reasons why you can see the bull bars are a little bit decreasing in the following two years, and the future growing.

>> Kuor-hsin Chang: It still increase though.

>> Kharis Templeman: It's increased from.

 

 

>> Tsaiying Lu: It's annual increase amount, so overall. That's what it is. Yeah, overall it's increasing.

>> Kharis Templeman: I see, so we've just got it.

>> Tsaiying Lu: Yeah, it's annual increase amount, yeah.

>> Kharis Templeman: Thank you.

>> Tsaiying Lu: Fads in a lot of other places.

>> Kuor-hsin Chang: Yeah, all right, so overall it's increasing.

>> Kharis Templeman: Okay, thank you for that, okay, got it.

 

Thank you, Tsaiying, thanks very much. No problem. So our last presenter for the panel is Gwyneth Wang-Reeves. Gwen, the floor is yours.

>> Gwenyth Wang-Reeves: Thank you, Ellis. And then so hopefully the energy consumption suggestion would be as Dr. Lu mentioned, that in the coming two years the rate would not be as aggressive as recently.

 

So typewriter can have a little bit less pressure. So I reservoir and you might have heard of GE General Electric. Just so give you a short briefing about GV Nova. GE Vernova spun off from General Electric in April this year. And we are a purpose built energy company and our portfolio focus on gas, nuclear, renewables including wind, hydro, and many, many solutions that we want to provide to our customers and markets and countries around the world.

 

And many fellow panelists today, I also wear meanwhile, I am also a board member at Amgen Taiwan and co chair of Amgen Taiwan's Infrastructure Committee. So I very like today's setting because I think my sharing would be pretty aligned with Dr. Chang's presentation, but from an industry perspective.

 

So like in the first session, people talk about AI and data center, how they drive energy consumption globally. Right now, altogether data center, they consume around 1 to 2% of total power, and by the end of this decade the number can increase to 3 to 4%. That means the carbon emission can double between 2022 and 2030.

 

So now we have two major pillars, one is the increasing demand of electricity, the other is increasing emission of carbon. And now look at the case of Taiwan. I think TPC in general, correct me if I'm wrong, they predict that the power demand would increase by 2% annually.

 

But I think recently there are data shows that number could be slightly higher. So that's why I say I hope that Dr. Lu Xiarang could be accurate so that we could have a little bit of pressure. But just take PMC alone, this company, they consume around 5% of Taiwan's total power, which is equivalent to 3 million households in Taiwan.

 

And now I think Taiwan government did a lot of good policies, especially dating back to 2016. They decided to bring Taiwan away from coal to gas bear in mind that unlike some countries, Taiwan electricity market is highly regulated. A lot of policy are spearheaded by the central government and then exercised by the high power companies.

 

So that's why starting 2016, many companies, including Jeep, Renova, we are installing gas turbines across Taiwan, like the map shown by Dr. Lu from North, Central and the South, mostly on the west coast. And now I want to identify one key challenge for this deployment. Environmental impact assessment, not just for the power plant itself, but also for the storage and terminal receiving terminals for natural gas.

 

Years back, if you might remember, there had been a lot of opposition against the third terminal establishment. Luckily the government addressed the issue successfully but then I think Taipower still has a few number of terminals in the pipeline. And if such public perception issue cannot be addressed and that would certainly delay this deployment.

 

And that would pose, I think the first challenge for Taiwan, especially when it comes to the issue of energy security and resiliency. Meanwhile, also around that time, the government decided to welcome a lot of offshore wind developers to Taiwan. And I think Taiwan was a beneficiary from COVID Ironically, I know Covid caused a lot of travel globally, but Taiwan as one of the few markets around the world where business can still be ongoing as usual pretty much.

 

So that's why a lot of foreign investors, they decided to come to Taiwan to take advantage of the wind farm around Taiwan Strait. And also taking the advantage funded incentives provided by the Taiwan government. I wanted to highlight ones in Taiwan government did really good was the support for labor issues.

 

So virtually speaking, in major construction projects around the world, the shortage of skilled labor has been a huge concern for many companies. But I think back then Taiwan government did a really good job, they helped those companies to address these issues. And bear in mind for offshore wind projects, we cannot be offshore wing every month, every day.

 

There is only a certain period of time, a couple months in a year where the construction work can be conducted, however, there are challenges and mainly the localization issues. Recently, Taiwan government had relaxed the localization content requirement but I just wanted to just highlight that previously because the requirement was around 70% and that was quite a lot.

 

I remember a European wind turbine maker in order for them to work with local suppliers, they spend all the efforts so that the local company they can reduce the hours making turbines from 1000 hours to 48 hours. So now you see the gap, how much efforts and investments both sides have to put in to increase and to meet the capacity building.

 

But it's still not enough because of the localization requirement really increase the cost to make wind turbine. And it led to two challenges, one is what if the power generated by the wind power is not affordable enough? Then how can we convince government sorry customers to buy the green power?

 

If there's no sufficient pool of customers to commit to purchasing the wind power, then the investor would be likely to walk away saying that then that doesn't make sense for me. Now Covid is gone, I will work somewhere else to do my my projects and so that's why I think another reason why one of the reasons that Taiwan government relaxed the localization content requirement.

 

Some people might say okay, now town has finally is booming ecological ecology supply chains to produce wind powered wind turbines. But still compared to turbines made elsewhere around the world, for instance in Europe, the wind turbines made in Europe are still around 20% cheaper than the Taiwan made wind turbines.

 

So that's why it's still not easy or possible for Taiwan local companies to really fully join the global supply chain. It might get there one day, but I think for now it would be quite challenging. And so that's the overall observation that we have in gas and renewables but another big pillar is decarbonization.

 

Now Taiwan government just recently announced that they would introduce this new carbon fee which would take effect in 2025 at the rate of around US$9. So US $9 or Taiwan $300 per metric ton of carbon. But they also introduced this preferential rate standing like 100 Taiwanese dollar or 50 Taiwanese dollar.

 

How to get that rate depending on the companies whether they can provide a solid proposal to commit to net zero to reduce reducing carbon emission. Of course the rate might increase in the future, so a lot of companies, especially foreign companies investing here in Taiwan are watching how the fee is going to be increased in the future.

 

And another is now we talk about fees, I wanted to slightly touch upon the prices, as we all know electricity prices is heavily subsidized by the Taiwan government. And every year they have this committee to review the rate and how to ensure Thai power can operate sustainably and also effectively.

 

So there have been a few meetings increasing the rate and I think what the foreign companies would like to have is having a voice in the committee. Right, now the committee have meetings every year around March and September with 12 members discussing how to ensure the rate is reasonable, but at the same time would not damage the financing structure of Thai power.

 

I think a lot of time foreign companies of course can voice our concerns through some local commercial organizations. But I think it would be helpful to have more engagement with stakeholders, especially major investors like semiconductor companies or data center investors, investor companies here in Taiwan. Another thing is before looking is the technologies that enable decarbonization.

 

Dr. Jen had mentioned that Taipei has several plans to utilize hydrogen, ammonia or CCS technologies to achieve that decarbonization. But still these technologies aren't cheap, so it really requires some incentives or regulations from the government so that more companies from the private sector can support high power or even other industries that generate a lot of carbon across Taiwan.

 

But I wanted to also highlight that some good things Taiwan government has been doing. For instance, Taiwan's high power and also petroleum company, they are doing some pilot studies already to find out whether it makes sense for Taiwan to produce great hydrogen, for instance, or how do we do carbon storage.

 

Nearby the sea around Taiwan. And if I could just give a few bullet points about the challenges that we have observed in Taiwan in addition to what I just mentioned. I think Taiwan is a very, very strategic spot, we all know that. But at the same time, it's so difficult to come up with the best energy policy, it's virtually impossible.

 

For instance, the public perception not just for gas power plants or terminal receiving terminals, but also say nuclear. I think this is a really sensitive issue in Taiwan still, especially Fukushima incident in 2011 I think still remembered by some people. But of course the discussion of CMR is rising, so I think it might require further efforts from the government to communicate with the public to understand how to pacify the concerns before the government can move forward.

 

And the same goes with offshore wind project or onshore wind project as well, so public concern really is a key issue. I wouldn't say it's a negative issue because it shows that Taiwan as a democracy, people are really, really enabled to voice their voices over this important project.

 

But I think because of the democratic system as well, usually the discussion can take a bit longer than what everyone wants. And at the same time, Taiwan is building so many new fabs or data centers. How do we ensure that the discussions will not delay or cripple the TPC from providing sufficient energy to those tech companies and the rest of the markets would be a big concern.

 

Another is market, especially in battery, so now we have talked about power supply and also potential technology deployment of decarbonization, how about battery? We need battery to ensure that when the sun is not in the sky, when the wind stops blowing, we have electricity to be stored, and then when Taiwanese power, then we have sufficient battery.

 

There has been a lot of discussion over the imbalance market, which means when the government decided to open up the market for people to build more batteries, the bidders or the developers was way bigger, the number outweighed the actual demand. So now some developers, investors are arguing that, well, now it doesn't make sense for me to invest here in the battery market because I can't make money.

 

You have so many projects going on and can you help us to negotiate better rate or can you introduce a better mechanism for us to offset our loss? I think this is a market issue and I think in Amchan we have a lot of discussions and we are really gracious that we could have a lot of communication with the current government.

 

This issue cannot be addressed instantly, but at least I think it has been recognized and discussed. I think also permitting is a big issue. I think it was Dr. Chang or Dr. Lu, I forgot, maybe Dr. Chen mentioning that for a lot of projects, even after the central government endorsing the direction.

 

It still requires local governments to respond by issuing permits for construction or by issuing permits, in terms of environmental impact. And occasionally we see different voices between the central government and the local government. A recent example is Taipao recently just announced that they planned plan to retire a few more coal fired power plants in Taichung in central Taiwan.

 

And then it instantly drew criticism or comments from the Taiju government that well, I will not want to see so many power plants crammed in one city. They hope Thai power can reduce the power plants in Taichung, so the planning of our plans could to take a more balanced approach but not overly crowded in Taichung.

 

So I think that's also another development that we have been noticing that of course I think central government priority is to ensure the happiness and the quality of life for their citizens. Whereas the central government has to take a higher level of concern and approach to manage such policies, but still, I think it's back to my very first point, public perception.

 

How to ensure that public perception and complications would enable Taiwan's energy transition instead of disabling, that's a really key point here. So I quickly went through my talking points here, but I look forward to the interaction with my fellow two panelists, that will make the discussion even more effective.

 

So I'll stop here, thanks.

>> Kharis Templeman: Okay, great. Thank you very much, Gwyn, that was a wonderful overview of the market challenges that a company like yours faces in Taiwan. I wanna do one quick round with this panel before we shift to a broader conversation with our participants in the room and online.

 

And I think it probably help if I'm just very blunt about the political challenges that I'm hearing in Taiwan. On current trends, I'm having trouble seeing how this all adds up to a solution for Taiwan's energy mix going forward. We've got, as we just heard in the last hour, a potentially huge surge in demand for electricity driven by AI and semiconductor manufacturing.

 

And those are critically important to Taiwan's long-term economic fatality. And yet we've got a political system that hasn't treated this like a security crisis yet, and it by and large seems to be business as usual. There's some tinkering around the margins, there's gradual introduction of new renewable energy sources, but it's nowhere near enough to meet the challenge that Taiwan faces.

 

And there's a decision that was made several years ago to end nuclear power production in Taiwan and that still today has not been effectively reversed. And so I don't know how Taiwan makes this work over the next 10 years, how do you meet this surging energy demand while still maintaining some semblance of energy independence and security?

 

And let me just ask. So that's the observation, and then the question is TSMC is at the heart of this, right? TSMC is a really critically, strategically important company globally, and in Taiwan, it has almost taken on almost this kind of sacred status. The Hugo Shan Shan, right?

 

The sacred protector of Taiwan's security. So it has tremendous political influence right now in Taiwan, and so I'm curious whether TSMC, in their, I would characterize it as pretty desperate efforts to secure future energy supplies, whether they are trying to weigh in on kind of the political conundrum that Taiwan faces right now on energy.

 

And whether they're able to put their hand on the scale and tilt things towards a solution to this dilemma. So let's do a quick round, and then we'll open it up. Yeah, Kuor Hsin?

>> Kuor-hsin Chang: Yeah, I think you just mentioned about the nuclear power, right? Original the previous administration want to phase out in 2025, right?

 

And now you have a new government comes in, however, they are all the same party, right? It is, politically, if you want to have a dramatic change in the energy policy, you need to do it in a very dedicated way. So in addition, within the DPP, the current ruling party, there are still a lot of sentiment against the nuclear power.

 

However, as time goes by, especially globally, nuclear energy, it's kind of suddenly wake up, everyone adopting nuclear energy. And Taiwan aware of this. So for example, I have friends around me, originally they were against nuclear power plant, and some of them are gradually changing their view. So, I think the government is waiting to build the consensus, like I mentioned earlier, right?

 

And I think it is not far away, I think things are changing in the right direction. When that will happen? Hard to say, but I think maybe sometime before, in the middle of next year, they might have changed, but it's an issue Taiwan need to face. You need the stability power supply, you need the energy security, you need the carbon reduction, what else can you do?

 

Nuclear power, and you have two power plants, has been running smoothly for 40 years, okay? If there is a problem, the problem surface that is before already, your assumptions are available there, you can extend another 10 to 20 years. Why not? So I think this, for this, I think I must say, more than 50% of people in Taiwan already agree on this one, from the public poll.

 

Even DPP supporters is changing in this direction. For TSMC, as you said, safeguard Taiwan. So they have a, in Taiwan, they are treated very special as well. In Taiwan, the power supply, they have different label, right? The first one, if there is a power outage, they have a four categories.

 

If there is a power outage, the fourth label will be out of power right away. TSMC is the number one label. In addition to that, they also have some new power plant built, they supply power directly, just needs a supply power for TSMC. So I think in TSMC in Taiwan, in terms of power usage, I don't think, it's probably a concern, but not a big concern for them.

 

However, they do have a concern, they want to use renewable energy because of their customers' request. For that regard, I think that's why I mentioned earlier, it is very critical for Taiwan, the economy continue to grow, and the Taiwan need have reduce the carbon emission. And at the same time, need to continue to develop renewable energy.

 

Okay, Tsaiying.

>> Tsaiying Lu: Okay, yeah, I'll try to answer two questions. So, regarding the nuclear one, I'm currently not 40% of the supporters, what Mr. Dun said. So because, there are a few reasons of that because, one is that if you are going back to use nuclear in Taiwan right now it will cost a lot.

 

It was not just about the restarting cost of using the first and the second nuclear power plant. It also regarding the administration time that you will need to have the new licensing to go to restart the, or to extend the usage of the the Taiwan nuclear power plant.

 

So considering only its cost, and especially about the new political phenomenon about Trump's reelection, they're talking about to put more taxes on protecting Taiwan. That could add more pressure to Taiwan's financial burden, in terms of when the president trying to keep pushing forward agreeing the renewable energy spending, right?

 

So if we are going back to nuclear again, that will apparently, will eclipse what certain amounts of these spending that could spend on the expanding of renewable energy. So if we're really going to that direction, I think this is, it is a more responsible way to lay out all these kind of costs, in terms of social costs, financial costs, and also the administration costs.

 

To actually think about whether we are going to that direction or not. Because once you are restarting the process of, like, not phasing out nuclear, that will take around five to seven years. So that's we are saying that, maybe 2032 or 2033, that's the possible time for we to start nuclear power again.

 

So this kind of time framing is that feasible for Taiwan, and to sustain its strategic position against the multiple threats around. So this is one question for that. And about the TSMC, I think we have to really clarify what kind of electricity that TSMC needs. So that we can see, is that really a challenges or not for Taiwan?

 

Because in our reports by the end of this year, we are also clarifying that how TSMC is requesting a stable regional output instead of a total amount. Because the total amount you will see, it's a huge growth, right? But in terms of their specific factories, these fabs will be located in Xinju, Taizong, and also Gaozhong.

 

So these are three different regions in Taiwan. And so, how to support a stable regional output, that would be more important. And especially in terms of resilience, during a very serious earthquake in Taiwan this year. You can see how TSMC actually have the capacity to sustain its production even after the earthquake that shut down many of the factories.

 

And that even they have the power to support the city use in Taipei. That Taiwan city could use part of their electricity from their fabs, because they have these kind of resilient power units and everything. So, I think this kind of development is very important for Taiwan, no matter it is about the critical Weather condition, or it's about its contingency, I think toward the resilience power system that is the most impact important also the realistic way for Taiwan to go through.

 

I'll stop here.

>> Gwenyth Wang-Reeves: Okay. Okay, I think. Yes, thank you. So I think for nuclear. Yes, maybe like Dr Lu said, even if the government decides to restart the existing nuclear power plants, whether going to be decommissioned one where the one decommissioned already, it would take minimum six, seven years.

 

But I think for the government point of view they are looking long term, right? They are not just looking at the next eight years, they're looking at the next 20 years, 30 years whether Taiwan has sufficient baseload. So, I actually believe that this is still a topic word to be further explored.

 

Especially public perception requires time to be addressed. And Dr Zhen rightly so touched from the generational gap. I think globally speaking, younger generation people especially have a slightly more positive view of nuclear power. So I think this is something maybe Taiwan government can address in the future. But I think the key issue in addition to what we just discussed today, another important one is storage.

 

How can Taiwan government resolve the storage issue is not something the government can single handedly decide. It also requires to some extent, I individually believe that international support so that Taiwan can find a way to address this storage topic successfully effectively. On TSMC, Yes, Dr Lu is right that I think they have the capacity.

 

However, I also want to point out that TSMC alone does not represent every company in Taiwan. I think ICT Semiconductor is a very big supply chain and I don't think every company is capable of doing such big scale of green power purchase plan like TSMC. So, in order to support the further development of AI semiconductor here in Taiwan, I believe that the majority of tech companies here, especially the local ones.

 

They would like to have more support from Taiwan government so they could purchase affordable carbon free electricity. One thing I forgot to mention is there are some initiatives in Taiwan, they want to resolve this problem. For instance a local think tank, they want to introduce this carbon-free electricity certificate, which means if the power plants which are running on gas.

 

If they could have the right technologies to reduce carbon emission, or by carbon capture, or by doing more carbon storage then these power plants could be certified as carbon free. I think this is a really innovative approach. And of course a lot of companies are watching closely to see if this idea CFEC, carbon-free electricity certificate could be moved forward.

 

However, for this certificate to be accepted widely, not just in Taiwan, but also globally to make sense that Taiwan companies, especially those tech companies, can benefit from CFEC. I think there is still a long way to go. So my comments are, yeah.

>> Kharis Templeman: Okay, great, thank you. So, now I'd like to kind of transition us over to a broader conversation, and I wanna start by number one, seeing if our previous panelists, are you still online and can we bring them up on screen?

 

Fantastic, okay, I see Jen there. And second, I wanted actually to throw it over to my two colleagues who know this space better than anybody else in the room, I bet David Federer and Jim Ellis. If you'd like to start us off with some broader questions for the group, David, please.

 

 

>> David Fedor: Thanks, Ellis and to our panelists and appreciate folks staying online. Just a quick observation and some of my takeaways from some of Admirali's initial set of questions. What I'm hearing on the US and the Taiwan side, and then an open question for the group on what more could be done.

 

Jim asked about costs and capital, and how we assign those costs against different users. I heard from Paul, from Jared, and Jen in the US. We feel like the capital is there, and part of that is perhaps due to the entry into the power system of non traditional buyers like the hyper-scalers, who are cash-rich and who have scale in addition to other financial players.

 

In Taiwan, I'm not so sure I heard that from our panelists, the capitalist there. We heard about the sort of government setting of rates for Thai power. And I look at their balance sheets and they're losing billions of dollars a year based on those rates. And so that gives me some pause about, particularly with their sort of traditionally vertically integrated status about their ability to put some of this investment to both drive down emissions and increase capacity and security at the same time.

 

Perhaps that speaks to drawing from the US experience, the wisdom and the utility of bringing in some of these non traditional players into the system. And finding creative ways to do that through direct access through allowing those, the hyperscalers or the manufacturers who have a real incentive to get that right.

 

To make sure they have the supply they need to also be participating in the furnishing of that supply in a more direct way without having to be mediated through utilities. Which actually we're trying to sort of program around ourselves in the US. The second question of technologies, do we have the technologies?

 

And I heard from Paul in the US, the answer is all the above, given the scale of the growth and the speed. And I think you can talk about which technologies can serve need at which different timeframes, we had a discussion of nuclear in Taiwan and I guess I just observe.

 

Jared mentioned, compute is gonna go where there's power and where there's cheap power. And I do have concerns, especially whether it's in the mean term or the longer term. If Taiwan is keeping options off the table, they will always be less competitive than the next player. And so, as we think about compressing for six, seven years, think about compressing some of those timelines, just observe whether it's for traditional plants in the US.

 

We're working now in a very sort of unprecedented effort to restart plants. The NRC is surprisingly moving a sort of flexible and fast way to do that in a very constructive way. They're a collaborative organization by their nature. The NRC has embedded regulators from other countries who observe their processes.

 

I think this is a great opportunity for Taiwan to send people just to observe that process so that it's not waiting to see how it works in the US. Then deciding we're gonna do it in Taiwan, but we're doing that in parallel. So people are getting smart at the same time and learning from each other mistakes.

 

Similarly with SMRs, if that's gonna be an option in Taiwan, we're not gonna have SMRs in best case in the US until 29, 2030. I wouldn't be in a position where Taiwan say, we'll wait for 20, 30 and see how that worked. And then we'll do our own sort of ten year process to build SMRs embed now, learn alongside.

 

Especially with these huge buys that we're seeing from the hyperscalers that are really have the potential to get through first cost down to INF costs. So, embed now to keep the technology portfolios open. Permitting, okay, I'm hearing permitting is perhaps the limiting factor in both areas we're democracies and maybe this is sort of something that's.

 

We're just gonna deal with this in the geopolitical context of where compute goes and where you're able to build things and we'll have to learn from each other. I'm struck by this point of narrative and the opportunity that is presented by sort of a broad social interest in.

 

It's not just power and it's not just climate, which has sort of been the dominant narrative, but there's this other sort of competitive element with the growth of AI. And I think people understand that. I think the next big energy bill in the US is not gonna be an energy bill, it's gonna be an AI bill, but it's gonna do everything about energy.

 

I think that's an opportunity to break through some of these permitting questions that have become quite politicized. Grid is blue and pipelines are red and perhaps that's an opportunity in Taiwan's highly politicized energy environment as well to cut through some of these decades old sort of factions about how you cite.

 

And don't as you're expanding the pie, and so now I'll sort of turn to my question and there's a US Taiwan shared interest and potential for collaboration here. I think chips has been sort of the broader semiconductor question has been, you can sort of disagree with the quality of certain commentary on that.

 

But more has been written and said about Taiwan and the US press in the last few years, in probably the last decade or two, just because that became so salient to Americans, to policymakers, to businesses. I think there's a potential for that in AI and the compute needs as well to bring US and Taiwan closer together.

 

And my question for any of the panelists would be, are we seeing that level of collaboration now across US and Taiwan on some of these shared issues in the AI space. Whether it's government to government, where sort of we have some limitations. We have infrastructure to do that or is that business to business conversation of the sort that Gwyneth is right at the center of where I think that is alive and well.

 

Jensen's in Taiwan sort of every month giving speeches about this. But our governments can't talk about it. Is there a way to break through this? And I just observe also on the AI side US Europe, AI is becoming this thing for other democracies to talk about what their strategy is around the world.

 

And I worry that Taiwan is not at the table for those conversations as those conversations are being had from the US and their partners in Europe. So, if there's more that can be done, I would love to hear that because I think this is a platform for closer collaboration which would be good for security, economic prosperity of both of our countries.

 

 

>> Kharis Templeman: Okay, great, and I've just realized we promised to get you out of here at 7:15 and we are at 7:08. So I'm gonna use moderator's prerogative. If there's anybody in the room who has a question, keep it short and direct. We'll just take several questions and then we'll do a speed round through all of the panelists.

 

So any other questions from the audience in person here, yes sir?

>> Student 1: I saw one of the, one of the charts that Taiwan would use hydrogen to supplant other fossil fuels for electricity generation. However, hydrogen can be very expensive energy wise to create especially with ammonia and then later hydrolyzing it, right?

 

Does Taiwan have sources of green hydrogen or is it just gray?

>> Kharis Templeman: Other questions? David, are there any questions? Let's go Wesley first.

>> Wesley: Based on my report I read the TSMC is a huge electrode consumption of Taiwan like 8 to 9% currently and these consumption rate will increase to 2030 to like 11 to 12.

 

So it is a huge energy problem. Taiwan and also the like karis mentioned like 80% energy should import from the other country and also the renewable energy is unstable right now. So, I think I agree with Professor Chung's viewpoint that nuclear energy still very important to Taiwan and the SMR.

 

We should think more about it. And I think yeah, these issues raise a huge debate in Taiwan. But from my personal perspective, I think nuclear energy is still very important, so.

>> David Fedor: All right Tiffany, last question.

>> Tiffany: I'll keep it very short. So when I say that if you lag behind some key technological progress.

 

For example development, small module, AI efficient technology, then it is gonna take exponentially longer to actually incorporate it back to existing infrastructure. So I think my question for all the panelists is that I just wanted to hear that what are some. I think probably already mentioned that what are some policy changes.

 

And incentives especially for Taiwan that you would like to see that the government can provide both the public and private sector to keep up with these energy related tech.

>> Speaker 2: I want to follow up with the hydrogen question. I saw the hydrogen production on the northern part of Taiwan, then your storage on central part of Taiwan, then utilization in northern part of Taiwan.

 

I mean, yeah, northern south. Sorry, mixed northern south. But who is responsible for the transportation infrastructure for the hydrogen in Taiwan and if so, what's your consideration for the machine leakage? Because hydrogen is very light. I mean what's the strategy behind? You have three different locations for the production, storage and the utilization for hydro and if I'm allowed, the offshore wind.

 

What's the role for the offshore wind in Taiwan's energy mix?

>> Kharis Templeman: Okay, so we've got a lot on the table. I'd like to return to our first panel and we'll just have you react to anything that you've heard in this session so far, give us a couple of takeaways and we'll go in the order that we presented with.

 

So Paul, you're up.

>> Paul M. Dabbar: Yeah, I mean I would comment and maybe it's a little bit of the zeitgeist of what's happened last week, but I do think that all the above, I'm hearing a lot about hydrogen, I hear a lot about net zero and so on. I do think what we're seeing in Europe in terms of energy security, forgetting about energy security, forgetting about energy prices can be a fragile environment and availability.

 

We talked about that a little bit earlier. But I do think, at least from my point of view, while those are all kind of good points and good questions, I do think balance and energy policy drivers should be important for any country.

>> Kharis Templeman: Okay, great, Jen, your thoughts?

 

 

>> Jennifer Huffstetler: Yeah, one thing I just want to expand upon, I think one of the gentlemen just touched upon it is what does that secure and resilient supply chain look like for AI overall? So we know that the CHIPS Act was a first effort in that. Is there another pillar that is some type of energy act or a nuclear energy act that really looks at building out that supply chain so that it's secure and resilient?

 

We know we won't get more semis. Semifabs, as has been discussed, are very high energy consumers. So is there something that could be put together there for both of our national competitiveness? And then two of the other pieces to think about are education of the populace. It's not just in Taiwan that people are nervous about nuclear, I think that's globally, in the US as well.

 

And then the last piece is workforce training. How are we thinking about the labor that will be needed to maintain these solutions moving forward?

>> Kharis Templeman: Okay, great. And Jared, parting thoughts?

>> Jared Dunnmon: Yeah, I'll be very brief. I mean, I think on the policy side, what would I want to see on the US side?

 

I very much agree with what Jen just said. There's a complementarity to, all right, you need to get CHIPS figured out. You then need to get the power figured out. And I think that's a good way of putting it. I don't want to say that I'm. I feel I have a very strong opinion about this, but I think it's worth thinking about.

 

There is an argument that, hey, we need to get the grid figured out, we need the conductor sizes figured out, we need to get permitting and siding and a bunch of stuff figured out. Or we could just lean into behind the meter on really big loads. That's a thing we could choose to do and even like lightly incentivize them and say, look, get things repaired.

 

Get good at building them together. I'm not saying that's the answer, but that's a thing that's probably worth thinking about. And I would, I would like to see someone, I would like to see folks think about that before we automatically assume that we have to have, you know, large network generation to effectively power point loads.

 

The second thing, and this is actually more of a question for, I think for kind of follow on and perhaps for some of my colleagues. Is all the above mix, I think is absolutely right, particularly given the time frames that we have to work with. Which you have to solve it in both the short and the long term.

 

In the US, there's a reasonable hypothesis that advanced geothermal particularly can be part of this. Cuz you are looking at firm power and you are looking, particularly for the US, a workforce that is pretty substantially experienced in the same type of operations that you need to do that come from the oil and gas industry.

 

I think Dr. Chang, you mentioned, I think in one of your slides. Yeah, you had it there as like, I'm not really sure that Taiwan is going to do a lot of it. If that's true. Is that because it is a. Is that because of, you know, the resources aren't there, like the geothermal resources aren't there or is it because of like workforce or cost?

 

 

>> Kuor-hsin Chang: Okay, cost is the main issue. So right now they have it below 3,000 meters, that's what currently Taiwan is developing. And the capacity is about 8 megawatts, but the power that it can generate only about 2 to 3 megawatt. So there are a lot of problems. That's why.

 

So right now Taiwan is is going to develop the deeper one, so will be above 3,000 meters. So I think that's where people believe it has huge potential. But the issue is you need to identify the area. The resources. The resources that the government is helping out on that one, okay?

 

 

>> Kharis Templeman: Other parting thoughts question and then we'll- Yeah, actually, yeah, so Tsai-Ying, do you have any parting thoughts for us, any pearls of wisdom you'd like to leave us with after this conversation?

>> Tsaiying Lu: Yeah, well, first of all, I want to thank everyone that I really learned a lot from today's discussion.

 

And about the questions about hydrogen, because one of our current project is also doing about hydrogen about Taiwan. Because Taiwan does not have the capacity to produce enough green hydrogen for now. So the strategy is that we're gonna import them from Australia. So right now we're doing a research on how Taiwan and Australia can build on their certain.

 

They have established a certain partnership in terms of energy imports for now. So how to expand that and how the Taiwanese manufacturers could play a role in this Indo-Pacific hydrogen supply chain. I think that is very important, essential thing for Taiwan to do in the next few years.

 

And in terms of the different location of the hydrogen parks in Taiwan that they are planning to do the major issue correctly, yes, it is about the transportation. So transportation cost of green hydrogen is very high. So right now Thailand has a certain capacity because they have been importing ammonia or exporting ammonia for a long time, like 50 years.

 

So we have the technology and also the tanks to have these green ammonia to be imported from Australia or other places to Taiwan. That is for sure. There's no problem. And now from my former understanding, I think there's a project that's doing a feasibility study that we will open a receiving terminal for green hydrogen in the southern part of Kaohsiung Port.

 

So we started with where the petrochemical industry gathered together in the in Kaohsiung in the southern Taiwan. And we'll use that as a demonstration site for our hydrogen terminal. So that we can step by step establish a local hydrogen supply chain and also different adaptations and industrial adaptations started from petrochemical industries first and then we'll expand them to other sectors.

 

Yeah, that's my short comment and explanation to that. And if you want to see more, you can see our reports in the following years. Thank you so much.

>> Kharis Templeman: Okay, wonderful, again, DSET is the organization that's putting out these reports. Gwen, you will have the last word.

>> Gwenyth Wang-Reeves: Thank you.

 

Very quick, two points. Number one, hopefully that after January next year whenever opportunity for bilateral discussions that there will be opportunity for private sector to be invited to share our experiences with our counterparts in Taiwan. For instance, we could certainly share Jay Benovar's SMR construction projects in Canada which would be expected to generate power around 2029.

 

So I think that would be a good story to share with anyone interested in in from Taiwan. Second is, I highly agree with Jan's suggestion on workforce. I think talent over the past few years have been spending a lot of resources to develop talent in semiconductor. What about talent in the energy sector, especially nuclear?

 

There's only I think if I could be wrong but I think there's only one university, Tsinghua University has a specific department addressing nuclear science. So if Taiwan is serious about diversifying its nuclear, sorry, energy portfolio then I think it's about time to make educational investment to ensure that Taiwan has the talent in the pipeline to support its future energy transition today and tomorrow.

 

So my comments for now, thank you. And it's wonderful to be in this panel, thank you all.

>> Kharis Templeman: Wonderful. Well, thank you all for sticking with us and pulling a late night here. Laura, any final comments for us?

>> Laura Smoliar: I just want to thank everybody for coming in person and online and I hope this will be the first of many conversations.

 

I think this is a complex topic that is rapidly evolving that it is and we could clearly talk about it for quite a bit longer actually.

>> Kharis Templeman: I'm sorry we have to cut us off tonight before people leave. I just want to give a shout out to our excellent events team who helped make this possible.

 

I especially wanna thank Ali Kubota, who's provided virtually all of the logistical support for this event and who is staying late like the rest of us on a weeknight here. So thank you again for coming, and we hope to see many of you at Future Taiwan Project and S and T Hub collaborative events in the future.

 

Thank you very much.