Taiwan in the Indo-Pacific Region held Boom or Bust: Can Taiwan Secure the Energy Supplies It Needs to Meet Its High-Tech Aspirations? on Thursday, April 10, 2025 from 3:30-5:30 pm PT at Shultz Auditorium, George P. Shultz Building.
The prowess of Taiwan’s semiconductor industry puts it at the center of the AI boom. Chips made in Taiwan power most of the leading AI platforms, and its data centers are expanding at a rapid pace, driven by tech giants in cloud computing, AI, and the semiconductor industry. But this boom is also straining Taiwan’s energy supplies: the surge in electricity demand is happening while the transition to zero-carbon sources of energy has fallen behind schedule, and its final nuclear plant is scheduled to be shut down this year. Taiwan also faces a rising military threat from the People’s Republic of China, and its heavy reliance on imported energy supplies is a serious security vulnerability.
This event featured several experts with industry experience discussing these two parallel trends in Taiwan – the rapid AI-driven increases in demand for electricity, and the lagging development of new, more secure sources of carbon-free energy.
WATCH THE EVENT
>> Kharis Templeman: I have some framing remarks before we get into our panelists presentations today. And I wanna start with the question why a program on energy and AI? What do these two things have to do with one another? And the answer to that is in part an answer that Taiwan President Lai Chingda gave in a speech last June.
On June 4th of last year, less than a month after taking office, President Lai stood before a large group of tech industry leaders at the Taipei Nangang Exhibition center and pledged to turn Taiwan into a quote, artificial intelligence island. In his words, Taiwan is very blessed to have a technology industry in which leaders spent decades working and catapulting Taiwan into the center of the AI revolution.
And they turned Taiwan from nobody into one of the world's supporting technology pillars to execute on that vision. In February this year, the Lai administration announced that it would boost its sovereign AI program by aiming for collective computing power. So both public and private collectively of 1,200 PETA floating points per second or petaflops, up from 103 petaflops today.
Now if you don't know what a petaflop is, you are not alone. I just learned that term this week. For those like me unfamiliar with the unit, one petaflop is the equivalent of 1 trillion calculations per second, right? So that's a lot. To support this ambitious goal, Taiwan's first publicly owned supercomputer is scheduled to start operations in May next month.
And before the end of the year, a new cloud data center is set to open in the Southern Taiwan Science Park. Right there is also for I see several people in the audience who were at a separate event that we co sponsored with the Taiwan Science and Technology Hub on campus on Monday.
A basic law for AI management that is under consideration in the Levi Yen in Taiwan. And Taiwan's Executive Yuan has now unveiled new management guidelines to regulate the use of generative AI in government business. They have also, on a related move, imposed a comprehensive ban on the use of the China developed AI platform DeepSeek.
So there's a lot of movement in the space over the last year. The Lai administration is both setting ambitious targets and moving on policy quite rapidly. And the reason for that is that Taiwan's long standing advantages in chips design and production and a deep bench of AI researchers, including a couple who are here today with us and will join me on stage in a moment.
They those advantages appear to position Taiwan well to achieve Lai's ambitious goals for Taiwan. I don't think it's a controversial statement to say that if Taiwan falls short of these goals, it will probably not be due to the lack of talent or know how in Taiwan, but instead due to a much more mundane constraint.
And that constraint is energy. Taiwan's data centers and supercomputers will consume massive amounts of electricity. So will their semiconductor foundries. TSMC, for instance, consumes today about 8% of Taiwan's total electricity supply. All by itself, that share is projected to rise sharply in the coming years. And could be as much as a quarter.
25% of all of Taiwan's electricity consumption by the year 2030. That's only five years away. Last year, roughly 5. 95% of Taiwan's energy supplies were imported. If we focus just on electricity supplies, Taiwan was just under 10% of electricity that was generated from renewable sources. And another 5% roughly from nuclear power.
Meaning that the remaining 85% of Taiwan's electricity supply depends on imported fossil fuels. And the bulk of that comes in the form of liquefied natural gas, or lng. This supply, of course, is especially vulnerable to disruption. Whether through natural disasters such as earthquakes or typhoons, through geopolitical stability at the source countries, or from interference by hostile power.
And Taiwan does, as everybody in the room knows, face a hostile power right across the strait. I should note Taiwan's leaders are not blind to this challenge. To the security vulnerabilities that have been created by this dependence on imported energy. President Lai last year pledged. At the same time that he pledged to make Taiwan in AI island, he also pledged to ensure a reliable supply of electricity.
And in particular to prioritize, quote, diverse forms of renewable energy that can be produced on the islands and are much less vulnerable than imports to interruptions. But as we will hear more about today, Taiwan is also falling behind its renewable energy targets. The Ministry of Economic affairs estimates that renewables may by the end of this year increase to 15% of the total mix of electricity generation.
That is well below the 20% target that the government set several years ago. To complicate matters further, Taiwan is also only a month away from the shutdown of its last remaining nuclear plant, the ma Anshan Number 2 reactor in Pingdong County. That will eliminate another historically important source of zero carbon electricity generation in Taiwan.
And so, as a consequence, Taiwan's high tech and AI ambitions and its energy policies appear to be in considerable tension today to face a so-called Mao duo or a contradiction, right? So here today to help us understand this complicated picture are five specific speakers from Taiwan. All are experts in their respective fields.
And all of them will give us their distinct perspectives on the intersection between the AI and high-tech race and the struggle to develop reliable indigenous supplies of energy in Taiwan. We're just a procedural note, we're gonna divide this event into two sessions. First we have two speakers, Peter Wu and Jane Hsu to tell us a bit about where Taiwan's AI and chip manufacturing ambitions are likely to take the the industry over the years.
And then in the second panel we have three experts from the energy industry in Taiwan. We've got Dr. Lynn Lifu who's an advisor to Formosa Heavy Industries. Vincent Chen, a Taiwan native who until recently was investment manager at GSSG Solar and Gwyneth Wong Reeves who is the current engagement director for GE Vernova in Taiwan.
In the interest of time, I think I will go ahead and turn the floor over to Mr. Wu to kick us off.
>> Peter Wu: Thank you, Caris. Good afternoon. My name is Peter from Taiwan, actually from the Asus Group, the laptop company. But the over the past seven years I am in charge of the AI supercomputer business in ASUS.
And the subser is new called Taiwan AI Service Corporation. And it is interesting because it is a joint, it is a PPP without government. It's been spun off from our national Supercomputer Center. So I would like to give you a brief of AI cluster development in Taiwan before we have a panel discussion, okay?
As you know that Taiwan is accelerating high computing power not only semiconductors but also GPU servers. But you may not know that our government had a ambitious goal actually back to eight years ago to build up a AI cluster business in Taiwan, I mean the industry. So it is important because when you're doing an AI training a model, you can only do that thing done if you have a cluster, not just a server, not to mention just a chip.
So a cluster is so important because it is real, it's really run the computing power. So you can see that back to 2017 we have a series of supercomputer progress projects. We call it Taiwan Year 1, 2, 3, 4, and Taiwan Yeah 2 is important because at that time our government decided to allocate about $1,160 million US to develop the first AI supercomputer in Taiwan.
And at that time, Jensen Huang visit our minister and discuss we should investigate it together. Even there are no big models at that time. At that time the model is so small, we don't need as computer. But our minister said just like semiconductor, AI will be bigger and bigger so we should prepare in advance.
So our government invested before the ChatGPT almost five years ago and divided Taiwan Yeah 2 the supercomputer like this. And after two, it's 3, 4, 5. And the Taiwan Year 2 the project team spun up to the company, I'm running the Taiwan AI Service Corporation. And our core business is to help customer to build and run AI supercomputer.
So that's why we have a hands on experience in Taiwan for every AI supercomputer that. I will have a summary for you later. And another thing is that we knew AI superconductor its energy a lot at that time. So the ministry, I mean the minister decided to use cooler, I mean the direct water cooler technology at that time very very early.
And we decided to have energy friendly design. So we have a very good top of green 500 ranking at that time, it's a tenth. And my company, the power and AI Service Conversion, as I mentioned is spun off from the government. And right now there are seven supercomputers in top 500 right now.
Seven out of 500. Of course, majority is the states, but Powerline is growing fast and five of them are developed by us. And,
>> Peter Wu: Okay, and another thing is that before we discuss the energy, I would like to share because Taiwan developed the US supercomputer very early. So we tried to ask us what we can do, what kind of innovation we can leverage that AI supercomputer.
So we are very lucky because we developed very early. So right now we have plenty of use cases right now, foreign language model, foreign to Smart city healthcare. So when we go to other countries in Asia. Yeah, they are not just curious about Taiwan's developing hardware. They are also asking us what's our use case, what kind of applications we can do on such kind of supercomputer.
And Taiwan is also leading in sovereign AI. You may not know, but you know that after not just ChatGPT after DeepSeek many countries want to do sovereign AI. They want to have their own models because it's cheaper, affordable and the model performance is just as good as frontier models.
So there are at least five teams develop Southern AI right now. One is us and another is a team from MediaTek. You may not you knew that. And another is covered coming from our government Tiger. And another is coming from FASCA, another biggest high tech company from Taiwan.
So you can see that compared to other Asian countries we have previous experience of doing tat and we even help our friends in Paraguay to do Spanish language models. So it's another thing that not just technology. We discussed this with Congressman Councilwoman last year. There are lots of policy related things we can discuss.
Okay, and back to the AI cluster. The cluster is just the picture I show you. And when you look deeper into that of course there are hardware, GPU servers, networking servers. But there are also softwares that can manage the GPU resources so that the performance of the cluster can be so good to train a language model.
And after that there should be a portal user interface so that AI developers not just for research search but also from the AI experts in enterprise can do AI. And for the infrastructure there are also PUE thing we should consider. PUE is the efficiency of energy usage. And of course, there are also energy thing we should consider.
Okay, so that's my last deck and this is my quick summary of AI clustering Taiwan. There are four blocks you can easily understand that, right hand is the AI cluster from public. Right on this side, actually, there are three ministries invested AI cluster. One is the Ministry of Science and Technology Council is the first ministry and another is the Ministry of Economic Affairs.
The ministry invested the cluster with NVR together to build a cluster called Taipei-1 in Kaohsiung. Taipei-1 is not in Taipei in Kaohsiung because just like southern Taiwan has sufficient energy right now. And another ministry is the Ministry of the Data Affairs. They are also investing computing powers to serve startups.
Actually three ministries are our clients right now. And another AI clusters from the public mainly for research like Academia Sinica or NHI or big hospitals. They are also invested in AI clusters and the left side is from private sectors like us or another big company called Ubilink. Ubilink, Zheng Wei, Ubilink they invested a big AI supercomputer last year.
And another is for their own private use purpose, Foxcom invested their own AI supercomputer last year is their phase one development also in Kaohsiung and they have an ambitious goal. It's a three year project. We invested maybe 10 billion NT dollars in Kaohsiung to develop their own AI data center in Kaohsiung.
Okay, so that's a brief picture of AI cluster Divine Taiwan. And right now, there are two things we can observe right now. One thing is you can observe that big new AI cluster all will be planned in Southern Taiwan. Not just Foxcom, not just Nvidia but also the National Science and Technology Council, they plan another two big AI supercomputer in Tainan.
So why the big cluster developing southern Taiwan? It is because energy. Because there are sufficient energy right now in southern Taiwan. But another observation is that northern Taiwan is lack of energy. So last year our government announced that northern Taiwan including Taoyuan and New Taipei City cannot develop data center right now if you are going to invest in 5 megawatt above.
But there are of lots of big guys in northern Taiwan. So just like us, ASUS has a building in Taipei. It's a northern Taiping and we have a building. We plan to host our own AI supercomputer. So we need to use new renewable energy technology. So like Bloom Energy that we discuss a lot.
Southern Taiwan has different energy discussion right now. And from my another observation is that, the energy problems in Taiwan may be not lack of energy but the allocation advantage. How many allocation we should reserve for AI data center, may be the right question. Okay, so that's all from my part.
Thank you very much.
>> Jane Hsu: Okay, well, thank you all for coming. And so in fact, when I received the invitation to join the panel, I was a bit panicked because I know AI, but I have no idea. I mean, very little idea about energy consumption. So the question in the title is a big question to me.
Okay, next, let's see. Okay, by the way, I just want to point out on the website, the information about me is a bit outdated. I moved to Chang Gung University in Taoyuan where there's many data centers located. But I also took up the new responsibility of being the chair for Taiwan AI center of Excellence.
So I'm currently leading some of the discussions on some policy for AI research development in Taiwan. Okay, so this is kind of the pretext from Karis. So I decided I'm going to do some exploration and found out that not just about local companies like ASUS and Foxcom and TSMC.
Of course, there are many international high tech big guys coming to Taiwan set up data centers or as pointed out, the Taipei-1 GPU clusters. And as you can see, some of them, they do pay quite a bit of attention to clean energy, commitment to renewables and PUE and so on.
However, as was pointed out, the renewable supply in Taiwan is way behind schedule. Okay, so even if the companies committed to buy clean energy, they may not have that much energy to be able to acquire. And also TSMC is still building up, okay, not just worldwide, but also in Taiwan.
The projection I saw was less than Karis' number, but it's already 12% in a few years. TSMC is the crown jewel of Taiwan in terms of our AI competitiveness. However, that much energy consumption is scary, okay? So can Taiwan secure the energy supplied it needs to meet its high-tech aspirations to be a leader in AI worldwide?
ChatGPT told me, that's a million-dollar question. Well, actually I should say that's a multi billion-dollar question, not just a million-dollar question. And also it pointed out, we won't be able to do it without compromising its energy security or geopolitical stability. It went on to give me a lot of analysis about the short term and medium term or long term risk in terms of a lot of things pointed out by Karis.
And so that's kind of a pessimistic, okay? And what's even more important is our geopolitical risk is so high with Earthquake, typhoon, and China, and everything. And so liquid liquefied natural gas is one big source for our energy right now. But that's very fragile. Okay, so we definitely need to diversify.
However, given that I'm not an energy expert, I'm going to let the second panel to talk about the solutions I think I should focus on in terms of AI development. What should we do? So for one thing, personally, I am more or less against building up too many data centers in Taiwan.
Okay, or even too many server farms, GPU clusters. So the question is we definitely need to explore why, you know, we need to build up those facilities for what purpose. So we need to be very clear about for Taiwan to be an AI island. What are the important technologies within AI that we should develop.
So those are energy challenges. And so I'm talking about from the perspective of Taiwan, AI, kind of a national strategy. So in fact, Taiwan started putting AI into the national policy area more than 10 years ago. So we are already in the end of the second phase of the so called Taiwan Action Plan, Phase two.
And so in that, I think in phase two, the main thing was trying to answer the important questions facing Taiwan society, meaning the lack of workforce, the reduced reduction in workforce and the aging society, as well as the goal of trying to reach net zero by 2050. So those are very big problems and ambitious goals.
And also last year, they started planning the next phase, and it's mostly based on our leadership in CHIP and manufacturing and design. So it's called the cbi. So Chip Based Industrial Innovation Program and AI. A lot of the AI effort will be kind of part of that big initiative at this point.
So there are multiple research projects funding or research institutions, universities in Taiwan. Since Taiwan, AICOE is under the National Science and Technology Council, so we fund universities in other industries. But so initially the project were called for those to answer those questions and also trying to kind of promote more attention to AI governance as well as data management.
So those are the things that happening right now. And the centers are distributed throughout the island, so on multiple universities from Tai Da to the south in Chengong and so on. And also within these research centers, there are many collaboration with international research institutes. And I'm not going to go through the details, but we have signed collaboration agreement, umbrella agreement with the us, with France, with Germany.
So there are many ongoing international collaboration going on. And the center hosted multiple international events for the past few years. And going forward, we will continue this effort in identifying the right partners and, and working on the right research topics. So with the new Taiwan Center of Excellence, I tried to assemble a big group of faculty from multiple universities.
So we are grouped into different functions at this point. The major enhancement from previous was the addition of a new AI safety group. So previously, I think the ongoing projects focus a lot on applications. However, I think for the new project, we want to be able to focus on some of the core technology and security is a very highly emphasized role.
So here are the key terms we will be focusing on. Sovereignty AI, as was pointed out, sustainable AI, safe AI as well as agentic AI. So the agentic AI is meant to be a gateway to connect to the applications. But the other ones were not heavily emphasized in the ongoing projects and that's where we want to push forward in what's coming.
So we are releasing the call in May, early May, and the proposals are due in mid June. So one thing I do want to point out in talking about sovereignty AI and sustainable AI and safe AI is, you know, for a long time many of the development in AI has been kind of affected by the notion of the scaling law for the transformer architecture.
So the idea, simply put, is that in order to improve the performance of the large models, you need to have the proportionally improved computing power as well as proportionally improved data quantity. And so that has been kind of, you know, shown by all the big, larger language model, growing bigger and bigger, consuming more and more data, even to the point where some companies are claiming that they use up all the data they ran out of data.
And in fact, in Taiwan, we rent out the data way early because simply do not have enough trustworthy data in traditional Chinese. And that has to be one important push for the maybe legislative regulational point of view And so here, so I'm not going to go into the detail.
You can ask ChatGPT, it will give you the different models and the size and maybe the how many parameters and their energy consumption will grow proportionally to the size of the model. And there are many big players, and I think earlier this year, the big shock came from deep sea.
Okay. And as a person who believed in Sovereignty of Taiwan, I did not explicitly test Deep SEQ on my machine, but I did try to look up everything about Deep SEQ and try to get somebody else to do the test. And so in fact, there is a startup doing safety in the valley that they have examined closely, DeepSeek in comparison with OpenAI Mini.
And so they identify the risk of using Deepseek is high or medium. And even though they do provide good performances. So I would advise people who want to take advantage of freely available models to really evaluate safety issues before you plunge in. Although I think they do serve a very good purpose where they promoted the idea of open model.
And that's very important for Taiwan because Taiwan do not have the resources to develop our own model from scratch. And so we are an active participant in the community within the open model community. And however, on the other hand, I do wanna point out Deepseek do have many good advantages.
In particular, they are very good in terms of energy efficiency, and they improve energy efficiency on multiple fronts and that's very impressive. So I think, you know, in Taiwan, so one thing Peter mentioned, that there are at least five different efforts going, going on. I highly encourage that we need to consolidate because after Deepseek, even China consolidate into one giant effort.
Taiwan cannot afford to do more than one. So we need to consolidate and we need to take advantage of the good technologies that have been explored and proven to be useful. So in terms of algorithm and hardware aware improvement, hardware aware meaning that when you are designing the software, you consider the information about the hardware so that you optimize your algorithm.
And also they are very good with the infrastructure in terms of dynamic energy aware scheduling. So they would try to use whenever the lowest energy demanding machine is available. So that's kind of reduce their energy demand significantly. And also the fact that it's an end to end integrated system design really put them on the map for that reason, okay?
And so on the other hand, you know, in, in not just for AI needs to address energy efficiency in the AI algorithm and training and so on, but also AI can be very helpful for improving energy. For example, for grid management, we can try to use AI to optimize energy distribution, predictive maintenance or even integrating different types of renewables, okay?
So those are all very difficult, complex problems that may use the health of AI. Okay, so I'm sorry, the video does not play, but for tight effort, we produce a small documentary. And so the idea is that Taiwan needs to develop our own model Instead of just using LLAMA or Deepseek.
Mostly because we need to preserve our sovereignty in terms of our culture and political information. But also sovereignty. AI means not just the content, but also the infrastructure. And the ecosystem for supporting AI development needs to be self-sustained. You know, for Taiwan, it is very important that we will have our AI working even during the time of war or, you know, other major natural disasters.
Okay. So we are establishing more regional collaborations, the international collaboration on related issues. And so with Germany, there is some effort with the semiconductor and battery technology. Okay, and so it is important to point out that energy is no longer just the environment issue. It is a national security pillar for any nation that wants to be a leader in AI.
So Taiwan must do multiple things. And I think we will have time to talk about the details. And so that concludes my presentation. Thank you.
>> Kharis Templeman: All right, great. Mics are working, we've got a very knowledgeable audience on these issues as well. So I just have a couple quick questions that I'll throw at you and then we'll turn it open to the audience.
The first question is I wonder if actually both of you, I'd like to hear you tell us what sovereign AI means to you. What does it mean to have a sovereign AI platform? And does that just being trained on its own on data from your own country? Or does it also mean the kind of hardware, the infrastructure and how can you have a sovereign AI platform that's also built on open source or open model?
An open open model architecture.
>> Peter Wu: Soft AI is very hot after Deepseek, not just Taiwan, we discussed with many countries. And from my perspective, Soft AI should conclude with four aspects. Everything is about control and support the innovation in the society. The first thing is infrastructure and the second thing is the AI model itself.
And the third thing is the mechanism to do AI innovations. Mechanism maybe some accelerator programs, some subsidies from the government, everything that can accelerate innovations from startups to enterprises. And the fourth thing is the AI policy itself that tailored for this country to divide its own AI. I want to share a re-experience we discussed with a president of Paraguay last year.
It's a not big nation but they have 100% green energy because of the water. Yeah, they have a lots of water energies, so they are small in high-tech. But the president want to be somebody in Latin America with the help of Taiwan. But they cannot copy everything from Taiwan because we have different environment, history, resources.
So the president access if we can develop AI policies for them to help them to integrate their resources, not just infrastructures, not just model. Not just the innovation mechanism, but the AI policy may be the most important thing. Thank you very much.
>> Jane Hsu: Okay, so in fact this terminology is relatively new.
I have to say that more than two years ago when we started working on Tide, which is the trustworthy AI engine, the term did not exist. So we started working on, on just the, the pure idea of we want a model that's better at traditional Chinese because at that time most of the models were not very good with traditional Chinese.
And we realized that the model was mostly trained with less than 1% in Chinese and mostly in simplified, okay? And so that's why we started the effort on collecting corpus in traditional Chinese. So that was the initial motivation. But during that time there was a big kind of incident.
Somebody released a AI model, and that AI model answered many of the sensitive questions incorrectly, and that created a big raw from the society. And our leader was called to the Legislative Yuan to answer many questions. Okay, you know, how do you ensure that the language model is not going to give us the wrong answers and so on.
So that was kind of a shock. And so, although personally I think it's not just about those politically sensitive questions, it's about the model being able to align with the value and the objectives of the society, of the people. So what we care and what we can benefit from.
Okay, so that's why, you know, it is important to train on our own model and also use all kinds of alignment methodology to align with our values. And also there are things that nobody else would do for us. For example, one project that I promoted last year was to build a large model for Taiwan's climate, and in particular the weather prediction system.
So the fact that there are global climate systems available, but they are at the resolution, that's not fine brain enough for Taiwan to be able to do a very good prediction on typhoon. Because I think for anyone who has lived through a rapidly changing typhoon in Taiwan, you will know that every time there are people criticizing the government prediction, why didn't you give us the right thing?
And you know, whenever they call the day off, typhoon did not come. And when they called the day to be on, then typhoon becomes terrible, right? So it is important for us to be able to take advantage of our own data that has been collected over time in Taiwan to help Taiwan to deal with these natural disasters.
So those are the kind of motivations for building sovereignty AI and the fact that as pointed out by Peter, the infrastructure and the ecosystem, you know, we need to have a resilient AI ecosystem In Taiwan.
>> Kharis Templeman: Okay, so one more question. Both of you mentioned there are big players in, in the United States actually building data centers in Taiwan and trying to expand their footprint.
And I noticed you had Microsoft up there. Microsoft recently committed to buying an entire, the entire generational capacity of the, what used to be known as the Three Mile island nuclear plant for 20 years. And so in the US at least they are securing private sources of supply that they have exclusive access to.
And I wonder if there's any effort by private companies in Taiwan to bypass Taipower or other electricity generation companies and basically build their own private supply on the island.
>> Jane Hsu: Well, there might be more experts down there, but my understanding is that many companies in the Hsinchu Science park they have their own power generators and some people, so for example Foxconn is building up their own power supply infrastructure so that they will have their own power system, okay?
So maybe Wes, we can share a little bit more on that, okay? But also I, I think in Taiwan there are people who are looking into the small scale nuclear plants that's not governed by the current, you know, Thai power and, or the bigger nuclear management thing. So but I don't know about you know, in terms of regulation and business side, you know how viable that's going to be in the short term, yeah, yeah.
>> Peter Wu: I want to add more information about that. Cloud and AI is a little bit different from Cloud's perspective. There are mega data center approach, but for AI they are more diversified small and medium clusters around the world. And that's one thing. And another thing is, if you look up the catalog from the big three CSP, you cannot find GPU resource in there or Taiwan region, because their GPU resources reserve for the states first.
Some of them reserve for Europe customer, many. Some of them allocate in Tokyo but not in Taiwan. So the data center you mentioned, Google, Amazon, Microsoft is for cloud business. It's mega data center, mega data center. It's no 1 megawatt, even 1 gigawatt per site. But for AI the cluster is not that big after deep seq, it's more energy efficient, more you know, efficient technology.
So the cluster need not be so big if enterprise or government want to do their own models. And, and the AI is so important. So many of big enterprise and government want to own their clustering, they're controllable probably. So it's quite different from the cloud business. And the third thing, it's because it's a small to medium cluster, so it's maybe a more suitable for new energy technology, like automation, again, like balloon.
It's maybe one megawatt Watt 2 megawatt per site, it's big enough for medium sized AI clusters to do a closed loop AI consumption. Yeah, thank you.
>> Kharis Templeman: Thanks for that clarification. That's actually very helpful. I want to open it up to the audience. If you have a question, I'll just come out with the mic one interested in.
Yes, okay,
>> Speaker 4: we haven't.
>> Kharis Templeman: And we've got a mic to mic. All right.
>> Speaker 5: Thank you, that was just so interesting and informative. This is going to reflect, I'm afraid, the non-technical side of this event in terms of probably ignorance and naivete. But aren't there gains being made in the efficiency of the operation of these models that could reduce the need for energy?
What is the scope to use the same amount of energy five, seven years down, two years down the road and have much more AI computing power because we're getting more bang for our buck.
>> Jane Hsu: Okay, so that's a very good point. And in fact, you know, many models are claiming that they are becoming more efficient.
You know, for example, Deep Seq is a significant one and also there are many companies like Nvidia is claiming that their newer chips are more energy efficient as well. So that happens both at the software and hardware level. I think yesterday Google just announced their new TPU architecture which is 20 times more efficient in terms of.
So that has not been tested but that's their new release, okay? But they are not following the same architecture as the Nvidia GPU. So that's kind of a competing architecture. So we can see that per instruction the energy cost is going to come down. However, because of the power of the functionalities is so evident now that more and more people are trying to use AI and many people, when they use AI they are not, they do not have the energy consideration in mind.
So many applications are designed in a very energy non efficient way. So I can envision, on the one hand we are getting more efficient, and on the other hand we are getting more wasteful. So there is a tug of war, yeah.
>> Speaker 6: Both of you alluded to the importance of value alignment for creating the training and the models that Taiwan uses.
Could you share a little bit more on what the kind of practical governance structure of that might look like? Who in the society is, is weighing in on how we define and implement those values? Is that something that is decided by the government or statutorily? What are the kind of guide rails there or in the civic society?
What efforts are we seeing?
>> Jane Hsu: Okay, I think in terms of AI governance, the most prominent leader is probably EU. You know, they are way ahead in terms of defining all the different governance policies, principles and regulations. And they are the first formally enacted AI Basic Law. And as was pointed out, Taiwan is in the process of defining the AI Basic Law.
So, in fact, after I go back to Taiwan I will be at the Legislative UN attending one of the Gong Tinghui hearing, open hearing. But it's still very much in the discussion and debate mode, although I think the panel organized by Wesley a few days ago, there was a kind of a common sentiment that we should not over regulate AI because that may impede the progress.
So it is kind of still a very fluid situation. But in Taiwan, the AI Basic Law now will be under the Minister of Digital affairs, and it will be facilitated by digital, the Ministry of Economic affairs, and also the National Science and Technology Council.
>> Kharis Templeman: I want to be cognizant of the time.
We've got time maybe for one more question before we transition to the next panel. So thanks, thanks.
>> David Fetter: Let me jump in. Really appreciate the presentations. David Fetter from the Hoover Institution. I learned more about sort of how Taiwan is thinking about AI Compute than I've seen before.
So that's excellent. I have an energy background here in the US So think a lot about what we're doing on the US Side and thinking about powering AI, which is if you look at sort of the priorities of our Secretary of Energy, our secretary of Interior, our president for the country overall, they have AI dominance as one of their top five priorities.
And it's really, it's bipartisan. You know, it's really an amazing public private partnership in terms of what the US Needs to do to become an AI superpower. And on the energy side, it's changing everything. I mean, we have forecasts for about 4% of electricity demand today coming from compute needs going to about 10% of all US electricity demand by 2030 coming from compute needs mostly for AI compute as opposed to cloud.
And that's for training and it's for actual application needs. We're talking about sort of 10 cluster sites of compute of 5 to 7 gigawatts a piece from some of the hyperscalers. I mean, so it's multiples or nuclear reactor unit sizes for each of these clusters, and huge deregulatory effort to try to support that, get behind the meter build.
Trump talks about a doubling of US Power demand for AI. That's probably too much, but maybe in Northern Virginia we could see a doubling of something like power demand on, on the back of the AI needs. So it's, it's, it's really completely rethinking how we're thinking about our entire energy system in the US in order to enable this.
But I, a question for the both of you. You know one, one thing that comes up when we think about the semiconductor side and you know, Peter pointed out that 100% of the chips for sort of AI accelerators are made in Taiwan. 9% of the servers know that the Biden administration's December 2024 AI diffusion rules, which seek to in a pretty heavy-handed way control where the top AI accelerator chips can go around the world where Nvidia can sell.
Those chips of course are made in Taiwan, but Nvidia gets to sell them. And it divides the world into three tiers of countries. There's tier one, 18 US plus 18 countries. Not even all of Europe makes that makes that makes that cut. Israel does not make that cut.
Singapore does not make that cut. Taiwan does make that cut. Everyone else in India does not make that cut. They're in tier two. And then Tier three is sort of restricted countries like China. But tier one is basically the United States government I read that saying we want these 18 countries to have unfettered access to the best AI chips.
We want you all to be successful and active in AI and sort of be AI dominant with us sort of team west on AI. Is that perceived on the other side in Taiwan, this sort of sense that the United States wants Taiwan to be one of the 18 major players in AI.
Is there a sense of sort of responsibility or opportunity there or is that just something that's in the background and you're focused on your own agenda?
>> Peter Wu: Yeah, actually we are happy to be in one of the Taiwan countries and as far as know it's Taiwan, Japan, Korea, right.
So actually Taiwan can be a play a very important role in regional based AI computing. Not just have to build but most important is to manage a network across countries. I mean the whole world, but maybe the Southeast Asia together with Japan or some countries in Asia to be a AI computing network to serve further AI inference needs.
So I hope Taiwan can leverage tier one policy so we can do that better.
>> Jane Hsu: I think for Taiwan, Team America is our choice. So I think that's without a doubt also the fact that we will have very close ties with Japan and Korea, not just in terms, in terms of the regional alliance, but also culture wise and language wise.
There are lots of commonalities so in fact the Japanese government is doing this CJK large language models and Taiwan is participating in that. And so hopefully the training effort of building up the large language models will arrive at the point where it's mature enough that we can focus more on the inference side, inference and application.
And so for Taiwan we are very strong, not just at the server side but also on the edge devices. So we do want to be able to have a very good role in the scene of defining the next generation of AI applications.
>> Kharis Templeman: Well, that's all the time we have for the first panel, I like to give our speakers a round of applause, thanks.
Yeah. And we'll go ahead and transition straight to the next panel which is on energy security in Taiwan and I believe online with us today. Today we have Dr Lin Lifu who is an advisor to Formosa Heavy Industries. But more importantly, he previously served as the vice chairman of the Atomic Energy Commission in Taiwan, which was recently renamed the Nuclear Safety Commission.
This commission supervises Taiwan's nuclear power plants, its nuclear facilities and the use of radioactive material in commercial and research activities. From 2009 to 2020 13, he was the program managers of Taiwan's National Energy Program and he led the National Science and Technology Council's clean coal projects. Before that he spent more than 30 years as a researcher at the Institute of Nuclear Research, including serving as general manager from 2004 to 2007.
So, Dr. Lin, if you are online, the floor is yours, please go ahead.
>> Li-Fu Lin: Yes, okay, distinguished guest, it's my honor to join this workshop. I would like to discuss the role of nuclear energy for improving the energy independence and security of Taiwan under the potential blockade threat.
Specifically focus on the power sector where is the nuclear energy exclusively utilized in Taiwan? My presentation will cover four parts Part A Definition of energy independence and security or the power sector. Second, the background information of Taiwan power sector up to 2050 part C comparison of scenario with and without nuclear options.
Part four recommendation for restoring nuclear oxygen to national energy portfolio the definition of energy independence or power sector is defined as the ratio of domestic resources produced. Electricity and total generation electricity and security we were concerned on stable and affordable electricity. Energy security index of fuel will define as the ratio between the actual reserve capacity and two months reserve capacity is considered an under threat scenario.
And the total energy security index of power sector is a summarize of the energy security index of fuel multiplied by the fraction of electricity of fuels in the generation. And for affordability, we reexamine annual generation cost of each power technology and amortization cost of reserve capacity. Now I will give a brief in a nuclear power developed in Taiwan.
Taiwan belong to early adopter of nuclear energy in 1985 with 5.1 gigawatt installed which benefited for energy independence and economic development after the 1973 energy crisis. However, follow the Chernobyl and the Fukushima disaster coupling with the nuclear free homeland policy of Korean ruling party. 2017's Electricity Act amendment mandated completely phase out the nuclear option by 2025 but recent poll shows the majority favor keeping nuclear option beyond 2025.
So legislative representative are working to replace nuclear free homeland policy by carbon free homeland approach. This slide summarizes the status and future development possibility of nuclear power in Taiwan. Taiwan power on the four nuclear power plants size, each has a capacity for expansion, in total up to 15 gigawatt is possible.
Currently, Nuclear Power Plant 1 and Nuclear Power Plant 2 is under decommissioning but their spin fuel stay in retro coal due to the delayed Springfield storage program. The NPP 3 unit 1 now is entered decommissioning in last June and unit 2 will be followed in this May. The NPP 4 has completed construction but not yet got the operating license.
If the nuclear energy will restore under regulatory compliance, NPP2 and NPP3 could potentially return to the service within 18 to 24 months. If we consider the recovery of NPP1 and the restart NPP4 would require 3 to 4 years. In addition to add new capacity, the small modular reactor technology as well as the large system with inherent safety feature should be considered.
>> Li-Fu Lin: To mention the challenge of zero emissions by 2050, Taiwan's strategic energy plan involved to deploy over 100 gigawatts of non dispatchable renewable energy by 2050. And around 20 gigawatt dispatchable in all emerging renewable technology like enhanced geothermal and ocean thermal energy, a conversion technology not without nuclear option.
>> Li-Fu Lin: This slide summarizes the trend of energy independence from power sector from 1983 to 2024. Here you will see in 1983 the energy independent index is around 65% whereas the nuclear energy and renewable energy play a major role. And later with the economic growth accelerated increased dependence on imported fossil fuel so the index is declined to around 70% in 2024.
The similar trend as happened with the energy security index in 1983 is 94% and the decline to 2024 is around 49%. Although the last decade the renewable energy was strongly deployed but the effectiveness of the renewable energy expansion undermined by the phase out nuclear program.
>> Li-Fu Lin: And here is the comparison of the reserve capacity cost among the nuclear fuel, LNG and coal.
And for the investment billion per SGO, you will see the LNG is the most expensive and amortization of the safety the reserve capacity you will see the nuclear fuel is still the cheapest one. So nuclear engines is the most cost effective option for Taiwan to expand reserve capacity and also enhance the energy security.
And this slide summarizes the latest two years the annual average generation cost of all power technology. And the red line you will see that is representative the annual averaging generation cost in 2024. The nuclear generation cost is far below the average cost outperforming the oauth power technology. In Arbor State you would see the non dispatchable renewable technology like PV and wind power.
As well as dispatchable renewable technology geothermal are significantly higher than the average. Now coming to part C the comparison scenario with and without nuclear option, we assumed 2.3 annual demand electricity growth rate from the 2025 to 2050. So in that scenario the energy independence index without nuclear in 2030 is around 31%.
With the nuclear will improve to 40 to 50%, in 2050 without nuclear is around 54 to 84% and with nuclear with up to 88 and over 100%. This slide is just representing the numbers, you will see that without nuclear the 31% in 2030 and around 54 to 84% in in 2050, but with the nuclear we're up to 40 to 50% and 88 to over 100%.
And similarly trend will happen with the energy security index in 2030 without nuclear it's 57%. And with the nuclear with up to 63 to 68% and in 2050 without nuclear 72% to 88 and with the nucleus 90 to over 100.
>> Li-Fu Lin: And this slide is intended to comparison between the dispatchable power technology from 2025 to 2050.
The renewable technology we pick up the enhanced geothermal and ocean thermal energy conversion. With the nuclear option we pick up the printlife extension and the one module reactor technology. Here you will see we compare in the aspect of the affordability and reliability. Availability or reliability, in availability we compare the technical redis level for not yet mature technology like the small module reactor in LTO thermal and ocean thermal Some more energy and print life extension belong to the mature technology and also the generation cost for the mature technology.
You will see the general generation cost for LTO is far below to the hundred dollars per megawatt hour. And then if we compare the innovative technology for technical maturity, we pick up the technique reduced label. And we will see the engineer geothermal is less behind the nuclear scale and ocean thermal.
But if we compare to the levelized cost of electricity in 2030 and 2040, you will see the nuclear option is superior to the ocean thermal. And if we compare the market size potential in 2040, you will see the nuclear option is better than the engineer geothermal. So in short, the nuclear power with dominant advantage by 2050 the dispatchable renewable technology.
So come to the recommendation for the action item Considering the demand for electricity in the AI era, strong growth and also under the threat of the geopolitical blockade and also concerning timely availability and affordability of dispatchable renewable technology, I'm convinced that the nuclear option should be very beneficial to enhance the energy independence and security of the power sector in Taiwan.
And for the Taiwan's nuclear power program there is a weakness in the back end. Although the world practice has been popular in international, but availability of the site for low level waste and high level waste still is problematic in Taiwan. So without a public awareness of the energy, nuclear role for enhanced energy independence and security of Taiwan, the political leadership and the determination will not be very useful for restoring the nuclear power to the national energy portfolio.
So the most important action step is to raise the public awareness of the nuclear energy role in energy independence and security. So that's my conclusion. So before I close, I would take this opportunity to thank Zhang Guoxin Bose, Independent Board Member of Satanic Inc. Mr. Jian Futian, former deputy general manager of Tai-power company.
And Dr. Ruey-yi Lee, Executive Secretary, Advisory Committee, NARI National Atomic Research Institute for their valuable comments. Thank you for your attention across my talk.
>> Kharis Templeman: Thank you, Dr. Lin. That was a very informative presentation and I would be remiss if I didn't tell people it is not yet 8am in Taiwan right now.
So he got up very early to join us this morning and for that we're very grateful for you. Thank you. Our next speaker will be Vincent Chen, who's a Taiwan native and an energy investment and policy specialist with a decade of experience in the private sector. From 20 to 232020 to 2023 he served as an investment manager at GSSG Solar, which is a US based renewable energy private equity firm where he led the development of its power generation portfolio in Taiwan.
His work there included building Taiwan's first hybrid solar energy and aquaculture project backed by a foreign investor. I'll leave it there and give the floor to Vincent.
>> Vincent Chen: Thank you Karis for the invitation, and thank you everyone for being here today. So as Charis mentioned, I grew up in Taiwan and had the fortune to come here to study at Stanford with my my advisor Larry Diamond here today.
Very thankful. So really quickly before I start my presentation here, I just want to show the kind of projects that we're working on in Taiwan. K mentioned earlier that one of the major focus that we had was building what we called co located or hybrid types of solar energy development.
And this is really a response to the reality that real estate is short in Taiwan. We are very densely populated. We don't have a lot of space. So a lot of the renewable energy development in Taiwan has to coexist with whatever land use that is already happening in our island.
What you see here is solar panels, elevated almost like the parking lots that we see in the US here, sometimes times on top of fishponds or shrimp ponds. So this allows the activity, the economic activity to go on continuously underneath and solar generation to happen on top. I think this is a relatively new concept both for Taiwan but also in the world.
In the current government's energy transition plan, around 50% of the planned solar energy capacity add in Taiwan will be some kind of hybrid land use solution like this. So now to refocus on the topic of discussion today I think this part of the panel, we're talking about the supply side of energy resources in Taiwan.
And I really want to focus on the binding constraints that is limiting the increase of supply that we're talking about here. And I'm going to bring primarily a practitioner's view to this. And something Keras mentioned earlier really resonated with me. He said that Taiwan's AI ambition will not be bounded by kind of the talent or the tech development.
I will venture to make a similarly audacious claim that Taiwan's renewable energy supply will not be constrained by capital or the demand of such clean energy resources or the policy support. In fact, the binding constraints in my opinion are something that are less directly related to energy or energy policy, it is about Taiwan's land use policy.
It is about Taiwan's local governance, especially at the very local level where these projects are being approved and built. And also, more importantly is Taiwan's sometimes conflicting industrial policy. It's not unimportant to focus on building local capacity to build hardware to supply the nuclear, sorry, the, the renewable energy industry in Taiwan.
But it does come at a, a cost, whether it's, it's monetary cost or in terms of time. It's not a free option there. So before I go further, let me contextualize this for everyone a little bit more and look at the big picture here. I think this data is showed, alluded to a little bit in some of the presentations earlier.
What you're seeing here is essentially the energy mix that Taiwan's energy system uses for the past 20 or so years. You will see the majority of it is imported fossil fuel. The pink and the gray is what you see as natural gas and also coal. You will see the dwindling part is of yellow is nuclear.
You will see the phase out happening pretty soon and then the, the lining on the top that is slowly but steadily growing is renewable energy. As of this year, Taiwan, Taiwan's plan was to have Approximately I think 20% of energy generation from renewable energy. We're currently, I think as of the end of last year we're at about 12.
So we're like a little bit more than halfway there. If we project this forward. What you see kind of the faded part of this diagram represents where we think we need to be. So this is the government, government's forecast of a load growth. Unlike the US we, we haven't seen load growth in a while.
Taiwan has been planning for load growth for quite a bit. And on average we're growing about 2 and a half to 3% on an annual basis and going forward. So that's the, the steep slope that you're seeing there. And you see that, you know, 2025 is when we phase out the nuclear, so there's no more yellow on the graph.
And then in the future the mix is, roughly speaking, especially in the long run, 50% imported LNG, which is the pink part, 20% imported coal, and then finally 30% of renewables. This includes solar, wind and also hydro. I think that the most obvious takeaway here is that the delta between where we need to be and where we are today, the green gap is probably the largest between today and, and, and 10 years down.
I think this is close to a threefold increase that we need to see. So how, how are we building to address this renewable energy gap? You will see in this graph that Taiwan actually made very good progress for the past one years, the increase is quite steep. And I think particularly since 2016, the DPP Administration's 3 so far has has been doing quite a good job ramping up both solar, in particular, but also offshore wind generation in Taiwan.
However, I think the story is familiar here. We are still well short of where we need to be in terms of our renewable energy goals. We are about less than halfway for our offshore wind target for this year and cumulatively that is and we are about only 3/4 there in terms of installed capacity for solar energy in Taiwan overall in terms of the, the accumulative renewable energy resource capacity that we have in Taiwan, we're about 30% short overall up until this year.
And I would also point out it's a little bit hard to see in this particular graph, but the stock slope of increase in fact slowed for the first time in 2024. We are about 20% less on a year on year basis compared to 2023 when it comes to renewable energy capacity added into the Taiwanese grid.
Some will argue that this is kind of the lagging effects of the COVID supply chain challenges, which makes sense. It usually takes a few years for projects to actually come online. But also I would argue that there are also fundamental issues that needs to be addressed for us to really recoup that loss in momentum.
So now I'll focus on my kind of three main guesses that explains why Taiwan is falling behind in its renewable energy capacity. The first is the cost of the cost of land. As I mentioned earlier, real estate is at a premium here and it is really expensive to lease land for renewable energy.
You will see here on the left hand side, this graph is essentially, it's O& M costs for renewable, for utility scale solar power plants. And the green portion of the cost bar, sorry, the orange portion of the cost bar, is the shear that could be attributed to land leases.
So you will see on the, on the, on this graph, Taiwan not only has one of the most, the most expensive O and M costs to operate a solar power plant, it also has one of the most expensive land leases almost across the world In I think US and EU that's about what, 10 to 20%.
In Taiwan it's whipping 40%. So think about this, the operation of a silver power plant. Almost half of it goes to landowners. And how did this come about? So this, this isn't always the case. On the, on the graph on the right hand side you'll see two different lines.
The blue line on top is the guaranteed government purchase price for solar energy in Taiwan. This is what we call the feed in tariff, which is essentially the government coming up with a favorable subsidy for the industry and encouraging investors to come in and build solar energy, knowing that they will have the ability to recoup their costs over the years for 20 years.
In fact, you will see the red line that is increasing sharply is again showing the cost of land leases as a percentage of the annual revenue of a solar. Solar power plant in Taiwan. It went from, I think, 2% back in year 2016 up onto almost over 10% today.
So not only is land very expensive in Taiwan, it seems to be the case that as. As a share of the revenue of a solar power plant, it's going up in Taiwan. So what is it that gives landowners this market power to continue to charge more? I think that is the, the real question here.
I don't have the answer, but I have two guesses. First, it really is a seller's market on the ground. When we're developing projects, the permitting process really gives a lot of leverage and influence to the landowners, which I think it's one way to protect the rights of the landowners.
But I think there's also a time where that becomes overplayed and becomes a impediment to, to, to energy development. And especially when the government is faced with delays and delays, delays after year after year, they continue to bump up the financial rewards, as evident here as the feed entire freight.
That never really decreases because their policy design concept behind a feed and tariff rate is that it should decrease over time and eventually go into a merchant market. We have not seen that decline. And the landowners are able to extract Extract more and more out of the share of the profits here.
So I think there is some kind of a cycle here that there's more policy support, there's more demand into developing and investing, and that gives more power to, to the landowners as the sellers in this particular market. I will also highlight the second reason here. A lot of the project developments in Taiwan rely heavily on brokers for land aggregation.
These brokers come in early in the project and they exit early in the project as well. They are not financially aligned with the overall outcome of the development, let alone the 20 year kind of revenues prospect for these projects. So they, they basically help push the price upwards because they can go out and commit to these landowners knowing that the commitment really is not going to be coming out of their pocket.
So that that kind of creates an upward pressure to keep the land costs high as well. So the net result of this really is, in my personal view, I really think the current policy landscape enables a wealth transfer from Thai power customers, which is the taxpayers of the Taiwanese country, to landowners and brokers that are operating in this particular industry.
The second supply constraint that I want to talk about is about the regulatory landscape and the policy design here. It is incredibly complex, it is uncertain. And when we have the policies in place, it is really rigid. The diagram that you're seeing here is something that a developer like me had to face.
When I think about developing a project. Each box is a individual step, a permitting step in terms of what the developer has to do. Or we have to hire consultants to whip up a report or some kind of application to get the project going. Each color that you see on this diagram represents a different agency or authority that has jurisdiction over a particular permitting process.
So this ends up being an extremely costly, very time consuming process. And because there's so many different agencies involved, there are sometimes very often contradictory guidance that developers get. And whenever there are guidances also we are following very rigid kind of administrative rules. We have to follow very rigid zoning coding regulations that really delays project development.
And because a lot of the permitting happens at the very local level, at the village level or at the county level, it's unfortunate to say that also this is very prone, the system is very prone to corruption. And in fact there are a number of high profile criminal investigations ongoing that, that is related to renewable energy development in Taiwan.
And I think thirdly, perhaps my, my colleague Gwen will talk about this a little bit later. I think there's also kind of a moment that the Taiwanese government need to reflect on what its Priorities are when it comes to industrial policy for for example, in the solar industry world, even though there is no overt kind of discrimination between locally produced PV panels versus foreign ones.
But the government guaranteed purchase price that I had mentioned earlier. The government is willing to give a bonus tariff to power plants that end up using Taiwanese PV panels. It's great in terms of supporting a locally robust PV industry. However, from a cost perspective, from a developer's perspective, that is inefficient it we, we are buying at higher prices, lower technology products quite frankly that we can easily import from other countries, whether it's from Southeast Asia or from or from China.
We can debate the security concerns from importing TV panels from China later, but I think regardless, the economic argument stands in the wind industry something similar. In our windy sales, there are usually local content requirements that are taken into consideration when they decide who to award development rights in our coastal areas on the west coast of Taiwan.
In this diagram here, you see a wind turbine. Everything in yellow, you are not able to import from China. So you have to either find a local supplier. If a local supplier does not exist, you would have to force form a JV somehow and go into it building it locally with them in Taiwan.
And this is again a costly process, does not bode well for ramping up our renewable generation capacity quickly. So I'm going to stop there. I guess the takeaway really I want to leave you with is that again I think the key here to ramping up renewable supply. It's not something about the lack of finance or the lack of policy support.
We have all of that in Taiwan. It really is about subtle changes in land use policy, in how we run local governance and also in some of the higher level industrial policy objectives that may sometimes run encounter of the objectives to quickly wrap this up. Thank you.
>> Kharis Templeman: Thank you, Vincent.
That was fascinating and a bit alarming. Our last speaker today is Gwyneth Wong Reeves. I've known Gwen for quite a while. Gwen is kind of, to use a baseball analogy, a utility infielder. She can cover a lot of different issues because right now she works for GE Vernova in Taiwan.
Prior to that though, she was the Senior Director of Government and Public affairs at AMCHAM Taipei. And so she knows a lot of these issues very well. And before that she actually held several senior policy roles at Taiwan's National Security Council and at the Presidential office. And so Gwen, floor is yours.
>> Gwenneth Wang-Reeves: Hi, I hope you can see my face. Well because I'm quite petite. So I'm going to lean here and standing quite comfortably. Thank you again Caris, for having me here sharing this stage with so many respected experts here. I would say the story of Taiwan's energy transition is one of the most ambitious one in Asia.
But behind ambition lies a very delicate balancing act, that's not just about decarbonization, but also it's about geopolitical realities, energy security, and also economic competitiveness. And I'm sorry I don't have a deck, but if I may this, and can I borrow your slide? I. I would like to use the, the one that shows the, the.
The lagging energy transition if that's possible because that's quite relevant to what I'm going to address today. When we talk about energy mix in Taiwan, especially in the past eight years, people often talk about offshore wind development. But I want to bring in two in additional element, very important elements into the story.
One is gas. The role gas been playing in Taiwan's energy mix has been overlooked. So I would like to bring that up today. The second is the importance of having a closer US Taiwan collaboration. And so let's go back to Taiwan's energy transition. It began in 2016 when then government decided to introduce this really aggressive energy transition policy, four elements.
Number one, to reduce coal. Number two, to face our nuclear. Number three, to ram up gas, and number four, to develop renewables. So today we could quite be comfortable saying that Taiwan has made a lot of progress now. The portion of coal has dropped quite significantly, whereas nuclear is going to be phased out next month in May.
And gas now is accounting for over 40% of the total energy generation now renewables. And I would like to focus a bit about offshore wind because Vincent has done a really good job on solar. Offshore wind, it's one of the first movers in Asia. And I want to give credit to Taiwan because back in Covid time especially Taiwan was the one of few places in the world where businesses and daily life can continue nearly as usual.
So with that, in addition to government policy, lots and lots of foreign investors were attracted to Taiwan to invest, to build, to localize those projects. Offshore wind projects across Taiwan and Taiwan has abundant wind farm, especially on the west coast. And of course there are challenges. Number one is the localization.
I think Vincent just touched upon that briefly. It is strategically important for Taiwan to have its own indigenous supply chain. However, the local industrial capacity does not always meet the timeline of projects. So when that happens, that leads to delay of procurement, that leads to rising cost of construction, and occasionally it creates tensions between local suppliers and also foreign investors.
Number two is regulatory. Again, like Vincent mentioned, the complexity in permitting is a really, really big hurdle for a lot of developers. And for offshore wind, the permits they need to acquire often has to go to local government levels as well as central government levels. And not to mention when you deal with local government levels, you have to engage with local communities, for instance fishery communities.
And sometimes those engagement might come with unexpected outcomes. So really, it creates a good level of certainty that could unsettle for investors. The third one is maritime space. Just give you an example. Taiwan has a very aggressive goal to develop offshore wind. For instance, they want to have at least 18 gigawatt by 2035.
And right now in the west coast of Taiwan, if you count the area that is suitable for fixed base offshore wind, actually they can only accommodate around 12 gigawatts. Yes, there are other suitable areas in the water, but often they are in sensitive waters. And also they are under the superbition of Ministry of Defense, Ministry of Transportation.
So again, for investors, they have to go through multiple discussion and permitting processes oversold by different ministries. And that's again creating a lot of hurdles. Number four is macroeconomy. Because the global inflation, and also rising interest rate, and also the higher cost of commodity and logistic transportation, for instance.
It's not a Taiwan unique challenge, I have to say. It's a global issue. So around the world, a lot of projects are feeling the same pressure. But I want to mention that the story of Taiwan's offshore wind development. It's one of inspiration, but also one of complexity. What it is going through now, I would say is a growing pain, how to maintain the momentum at the same time, or survive this growing pain.
I would say Taiwan needs a more clear regulatory framework. Taiwan needs to have more communication across different ministries, different agencies. In Amchon Taiwan, we are advocating an idea of which is called Energy Czar. Which means that for foreign investors or companies, we could have a single point of contact or one agency that would talk to which could help us navigate this really complex landscape.
And now when we have lower and lower coal in the net energy mix, and with nuclear going to be phased out, renewable is being scaled up, but not there yet. What is holding towers? Energy security. My statement today I want to make is gas. Gas is the backbone of Taiwan's energy security.
It's not by choice, it's by necessity. So gas, I think in the chart here you can see the portion is bigger and bigger. It is actually accounting for over 40% of the total energy generation. And in the past eight years, Taiwan has seen expansion of LNG infrastructure and gas power plants.
For instance, we are seeing more and more terminals. The number three, I believe would be complete soon. And there is another storage being complete being built in around central Taiwan Taichung area. Not to mention there are quite a number of mega gas power plants already completed or going to be completed.
And gas, the value of gas is flexibility. Gas power plants can be ramped up quite quickly. So whenever Taiwan has a huge surge of demand suddenly or the renewable generation is not meeting the goal, then gas can be summoned to support the energy supply. And this is really important, especially base load is is so critical in Taiwan.
For those who are not in the industry, like I four, five years ago, what is base load? Just think of baselo as the minimum electricity you need to have 24,7 in order to keep everything running. In your house. You have a tv, you have refrigerator, you have your Internet router, you have phone charger.
When you go to sleep, even you don't use them, they're still quietly consuming electricity. So that's baseload. And for Taiwan, what baseload is so important because of the demand from semiconductor, from data center, from AI especially for chip makers, they cannot afford even a day of power drop or even 1 second of power drop.
So that's why baseload is so critical. Therefore, the role gas has been playing, like I said at the beginning, has been overlooked. And people might say, yes, but gas is also fuel and it's not clean enough. I will argue that any type of technology, gas is also improving itself.
For instance, for gas turbine, if you have, say, hydrogen, for instance, to be mixed with lng, you could reduce carbon emission. And the goal, of course, for a lot of gas turbine companies OEM around the world is to ensure that the turbines you have today can run run on 100% hydrogen.
Of course, we're not there yet, but I think that day will come soon because the need is growing. Earlier, Larry asked about when AI and also technologies are becoming more efficient. What does that mean? Does that mean the demand will go smaller? I think in general, what we see in the industry is that, yes, those technologies might become more efficient themselves.
However, the users around the world. For instance, one Google search costs way less electricity by one search on ChatGPT. So coupled with that factor, I think overall the consumption and the need for electricity will continue to grow. So that brings me to my last point. I want to bring to this story of Taiwan's energy transition is the importance of US Taiwan collaboration.
Yes, gas has been an enabler stabilizer for Taiwan's energy security. But at the same time because of that, Taiwan attracts a lot of foreign investors to Taiwan. For instance, AWS, Microsoft, the slide we saw earlier that they are all putting a lot of resources and money to use Taiwan as a hub for their R&D, or AI, or cloud based in Asia.
And we see a lot of potentials in addition to these kind of collaboration. LNG, for instance, US is a important US LNG supplier to Taiwan in addition to other foreign LNG suppliers. But at the same time it could also create more geopolitical flexibility for Taiwan when it comes to fuel sourcing.
Other than fuel, we see areas like ccus, carbon capture, power storage grid, or even with so called next generation nuclear power plant that Taiwan and the US can work together. Why is that important? Because Taiwan's energy security is also really relevant to global supply chains, to the global development of AI, cloud and semiconductor.
Even though there are a lot of complexity, especially coming from local level, for instance permitting or landscape that a lot of foreign investors have to go through. But overall, what I want to emphasize today is we still believe that with the right policy direction and also with more investment and more collaboration between public and private companies, for instance Genova, Taiwan could be leading its own transition on its own term.
So I will stop here because I think we should just have more conversation and hopefully I can help illustrate any question you may have. Thank you.
>> Kharis Templeman: Right, so given the time, yeah, please have a seat. I'm going to call an audible here and ask all four of our panelists to come up.
So Peter, if you would come up and Jane and I want to do kind of a round robin here. And in the interest of time, I thought we'd take a handful of questions from the audience for any of the four panelists and then we'll just go down the row and you can answer whichever questions you like.
And so I want to toss one question out to especially our two energy presenters. I want to ask Dr. Lin mentioned this energy security index and I'm wondering, have any of you heard of this? Is this an influential idea in the Kermit administration? Is there any broader recognition that this is an important thing to Consider in Taiwan's energy mix.
And if not, how can we make it more important and more salient in Taiwan? So think about that question and then let me open it up to our audience. If anybody in the audience would like to ask questions, questions of anyone, please raise your hand. Yep, let's go with Patrick first.
>> Patrick: Hey. Hi, I had two quick questions. One is the problem red tape and rent seeking on land, or just that there isn't actually very much appropriate land available? Those are two different things. And then the second question I had was, do you have any sense of whether the LNG stock brings a higher geopolitical risk just because it relies on extremely fragile infrastructure compared even to coal and oil imports, which at the end of the day are relatively easy and commodified supply chains?
Thanks.
>> Kharis Templeman: Questions, okay, hold that thought.
>> Speaker 5: Yeah, so my question is also about geopolitical risk, but I'll just put it more boldly. And we've been trying to make this point in our visits to Taiwan with multiple audiences. Anything you're importing is looking at very serious geopolitical risk.
And I am just speechless at the detachment of the political conversation in Taiwan from the gathering geopolitical realities. You've got a noose tightening around your island and you're closing down the last nuclear power plant that gives you, you know, one source of energy that isn't subject to fairly immediate strangulation of supply.
It just, I'm sorry, it doesn't compute for me. And the other is, of course, the solar and the renewables on whatever land you have in this regard. I just wonder, Vincent, if you've read this new book by Ezra Klein and Derek Thompson, abundance. Because when you were. They have a lot to say about California and I think you know, San Francisco as well.
And when you were talking about Taiwan, I thought I could have kind of closed my eyes and imagined you were talking about California. The difference is we'll just continue to sink into our stagnation, and bleeding away of population, and taxpayers, and investors, and we'll otherwise be okay. We'll just have more inequality and more homelessness.
Taiwan may not survive if it is gonna let greedy landowners and not in my backyard, otherwise understandable expressions of democratic opinion block the necessary energy self-sufficiency.
>> David Fetter: Normally, the issues we discussed in the newsletter, we talked about different force generation. But you could argue there's sort of a structural problem underneath, which is, is there the right regulatory structure in place?
You know, when I, I study a number of Asian energy economies and to be frank, you know, Taiwan is 20 years behind in deregulating its energy system, both on the power and on the hydrocarbon side. I know there are numerous attempts to that, there's political difficulties. The challenges being faced today are obvious to everyone.
Does that change the political calculus in terms of that underlying structural problem, which also arguably starves the whole system of the capital, to make the sorts of changes that you would want to with liberalization, pricing.
>> Kharis Templeman: Okay, great. Well, let's go. We'll start with Vincent and just come down the row.
You can respond to anything that's on the table.
>> Vincent Chen: I'll take on the solar questions and I'll leave the rest, especially the lng, for my colleagues over here. The short answer to whether there is suitable land to start first. I think the short answer is yes. It's a yes in terms of from a mapping exercise perspective, we can actually get to where we need to be being creative with how we use it.
The challenge here is obviously, unlike California, when we can go into the desert and just clear a bunch of land for commercial scale solar, that's not something that we can do in Taiwan. We already are just naturally short on flat space, it's very important to have flat space.
We can't be building solar projects in the mountains, and most flat spaces are occupied by people doing something. So that brings us to the second question is how do we deal with the rent seeking that's happening? I find it hard to kind of blame the stakeholders involved to be doing what they're doing, they're all behaving on their economic interest.
I do think that there are policy designs that needs to take place to make it a little bit smarter. I think the government is starting to think about whether what role they can play in land aggregation. It used to be the wild wild west where the government says, okay, we want to do, you know, 20% of solar by this time.
Everyone go out and like find land and you have all of these local influential people that are tied up with the village chief going out to people. And persuading them to sign pieces of paper committing their land for 20 years and no one knows exactly what that means.
There has to be a role for the public sector to play in this process. Whether it's more intelligent kind of siting decisions, like looking at what are the sensitive areas that has cultural or economic or conservational values. That we want to exclude before we even start this conversation and even after the exclusions, how do we go about this in an orderly manner?
I think there is a huge role for policy here and that is something that I think we are just at the very beginning and we need to be acting much, much faster to make this transition happen.
>> Kharis Templeman: Jane, anything you've heard that you'd like to add to?
>> Jane Hsu: Well, given that I'm not an energy expert, maybe speaking as a resident in Taiwan, I think one of the problem with energy policy in Taiwan has been the politicalization of the issue.
So in fact many of the issues were debated based on emotion rather than rational arguments. So I think that will be something that has to be dealt with kind of directly. Otherwise many of these conclusion is not going to change, yeah.
>> Vincent Chen: If I may add one thing in terms of the security implications of having homegrown renewable energy resources.
Unfortunately I don't think that concept linking national or energy security to clean energy, I don't think that connection has really happened in the masses yet. It's probably clear in the mind of policymakers, but I think that energy security concept is now finally being surfaced through the nuclear debate.
Through the real manifestation of a preparation for a blockade by the PLA. So I think there, perhaps the government or the civil society needs to double down on the fact that not only are we paying for paying a premium for clean energy that is expensive as solar and wind.
These are also independent energy resources that we don't have to rely on foreign supplies in an event of a military conflict.
>> Jane Hsu: Yeah, okay, so one thing I do want to add to the energy index, I think in the analysis of the costs, there is one omission that should be included is.
We should include not just the generation cost, but also the entire life cycle cost, yeah. So and that may kind of change the equation.
>> Kharis Templeman: You're thinking specifically about nuclear.
>> Jane Hsu: Yeah, yeah.
>> Kharis Templeman: Nuclear waste is still a major issue.
>> Gwenneth Wang-Reeves: But I want to use Jane's comments, a segue to talk about maybe nuclear a little bit.
I feel as though, like you said, things about nuclear could be easily become a political issue, not a scientific issue. But at the same time, I've seen more and more open discussions in the civil society. And so to some extent, as someone who is in the energy sector, I'm constantly optimistic that the society is ready, more open to discuss the possibility of having nuclear back to its energy mix.
So that's why in the past 12 months you have seen officials coming out saying testing the water, basically. And I think that of course, for private sector, it's quite a positive development and back to the question of LNG, yes unfortunately, Taiwan does not produce LNG. But I want to quote a really good report from Atlanta Council.
They mentioned for in Asia, for instance, Taiwan, Japan and South Korea, if they procured LNG from the US. This procurement can potentially give them more deterrence capabilities in the event of potential pressure from countries not friendly to them. So I think this is quite a strategic decision to some extent that Taiwan is going to procure maybe more LNG from the US and at the same time, Tai is also developing more LNG infrastructure.
Like I said, more terminals and also bigger storage and at the same time, more gas power plants across the island. And remember, one slide in Peter's deck you mentioned, in the northern part of Taiwan, companies are actually told that you can't build anything there because there isn't sufficient electricity.
And there's a new terminal going to be completed in recent years in Taojuan, so hopefully once that is completed, that could, right. Maybe not completely, but to some extent, yeah, I'll stop here, I would chime in later whenever I think of anything else, thanks.
>> Kharis Templeman: Okay, Peter? Nothing.
Okay. All right, I think we are past time and I don't wanna abuse our audience any further by keeping them there here longer. So I want to thank Dr. Lin especially for joining us at such an early hour in Taipei. I wanna thank Peter, Jane and Gwen for making the long trip here from Taiwan for this event.
I especially wanna thank our wonderful Hoover event team who've been provided the support for this, especially Amy Alonzo and Michelle Arroyo. And I just want to note this has been a production of the project on Taiwan in the Indo Pacific region. I'm Kara Stempelman and I will note that this event has been recorded and will be posted to our webpage if you'd like to revisit it or if you missed anything in the conversation today, thanks.
Good evening and final note, there is free food in the back, especially for the students in the room, you've been very patient. I invite you to enjoy the hors d'oeuvres and the drinks and linger as long as you'd like today. Thank you.
ABOUT THE SPEAKERS
Peter Wu is the CEO of ASUS Cloud and Taiwan AI Service Corporation. He has led the development of AI Foundry Service (AFS), which advances on-premises AI deployment, cloud-based AI applications, and generative AI ecosystem to implement trust-worthy AI 2.0. In 2013, Dr. Wu represented Taiwan at the WTO Business Forum where he shared ASUS's development experience in cloud services, and he was appointed as a member of the Advisory Committee on Bio Taiwan Committee by Taiwan’s Executive Yuan in 2017 and 2019. In this capacity, he provided guidance and advice on the strategic direction of the biotechnology industry in Taiwan. From 2018 to 2020, he also managed the biggest AI supercomputer project in Taiwan, helping it to achieve its best-ever ranking of 20th in the TOP500. The project was then spun off into Taiwan AI Service Corporation, the first commercial AIHPC supercomputer cloud service provider in the Asia-Pacific. Dr. Wu is actively involved in various organizations and committees, including serving as the chairman of the Taiwan AI Alliance, and holds prominent roles in the fields of Smart Medical, AI, cloud computing, and others.
Jane Yung-Jen Hsu is a professor and department chair of Computer Science and Information Engineering at National Taiwan University. Her research interests include multi-agent systems, intelligent data analysis, commonsense knowledge, and context-aware computing. Prof. Hsu is the director of the Intel-NTU Connected Context Computing Center, featuring global research collaboration among NTU, Intel, and the National Science Council of Taiwan. She is actively involved in many key international AI conferences as organizers and members of the program committee. In addition to serving as the President of Taiwanese Association for Artificial Intelligence (2013-2014), Prof. Hsu has been a member of AAAI, IEEE, ACM, Phi Tau Phi, and an executive committee member of the IEEE Technical Committee on E-Commerce (2000) and TAAI (2004-current).
Li-fu Lin is an adviser to Formosa Heavy Industries. He previously served as the vice chairman of the Atomic Energy Commission – recently renamed the Nuclear Safety Commission – which supervises Taiwan’s nuclear power plants, nuclear facilities, and the use of radioactive material in commercial and research activities. From 2009-2013, he was the program manager of Taiwan’s National Energy Program, leading the National Science and Technology Council’s Clean Coal Projects. He spent more than 30 years as a researcher at the Institute of Nuclear Research, including serving as general manager from 2004-2007. He holds a doctorate in mechanical engineering from University Karlsruhe in Germany.
Gwenyth Wang-Reeves is the Engagement Director for GE Vernova in Taiwan. She is responsible for establishing, and driving GE’s advocacy initiatives in Taiwan, engaging with local and central governments and other stakeholders on important public policy challenges, as well as advising the GE businesses on a broad range of regulatory issues. Prior to joining GE Vernova, Gwenyth was the Senior Director of Government and Public Affairs at the American Chamber of Commerce Taiwan. She has also held several senior policy roles at Taiwan’s National Security Council and Presidential Office, as well as the Australian Department of Foreign Affairs and Trade in Taiwan. Gwenyth holds a Bachelor’s degree in Political Science from the National Taiwan University, Master’s degrees in Political Communication at the Royal Holloway, and Democracy and Democratisation at the University of London and University College London, as well as a PhD in Politics and International Relations from the University of Warwick.
Vincent Chen, a Taiwan native, is an energy investment and policy specialist with a decade of experience in the private sector. From 2020 to 2023, he served as an investment manager at GSSG Solar, a U.S.-based renewable energy private equity fund, where he led the development of its power generation portfolio in Taiwan. His work included building Taiwan’s first hybrid solar energy and aquaculture project backed by a foreign investor. Before joining GSSG, Vincent led business development and fundraising at Jupiter Intelligence, a climate risk analytics provider, and Lucid Motors, an electric vehicle manufacturer. His research interests encompass power markets, environmental markets, and carbon border adjustments. He holds a master’s degree in international development economics from the Harvard Kennedy School and a bachelor’s degree in environmental economics from Stanford University.