Chinese policymakers are looking to diffuse AI into key industrial, commercial, and governance fields as deeply and quickly as possible.
The overarching framework for their approach is captured in the AI+ plus initiative.
Officials published the overarching policy architecture for AI+ in late August.
And more granular AI+X plans (e.g. AI+Energy, AI+Transportation) are now starting to roll out.
In this week’s Trivium podcast, host Andrew Polk is joined by colleagues Kendra Schaefer (head of tech policy research) and Cory Combs (head of critical mineral and supply chain research) to explain China’s AI+ approach.
They delve into:
The overarching AI+ framework – and how it differs from previous AI policies in China
The six key fields that China is targeting for AI diffusion
What to expect as the AI+ policy ecosystem takes shape in the months ahead
China’s AI+Energy plan – what it seeks to achieve and why it is so fundamental to the wider AI+ vision
How officials will approach more foundational aspects of AI+ – including foundational models, compute, data supply, developer ecosystems, and talent
The gang then wrap up the discussion by zeroing in on why the AI race is different from previous global competitions to produce technological breakthroughs.
The key point: Achieving AI breakthroughs will look more like past pushes into electrification and the internet – as opposed to the nuclear weapon or space races
You’ll definitely want to stick around till the end for this one – enjoy!
Transcript:
Andrew Polk:
Hi, everybody, and welcome to the latest Trivium China Podcast, a proud member of the Sinica Podcast Network. I’m your host, Trivium Cofounder Andrew Polk. And I’m joined today by two of my colleagues. First, our Head of Tech Policy Research, Kendra Schaefer. Kendra, how are you doing? Good to have you.
Kendra Schaefer:
Good. Good to be here.
Andrew:
Yeah. Glad to have you back on the pod. And, secondly, we’ve also got Trivium’s Head of Critical Minerals and Supply Chain Research, Cory Combs. Cory, how are you, man?
Cory Combs:
Good to see you. Good to be on.
Andrew:
Yeah. It’s great to have you both. I’ve got you both here today so that we can discuss China’s latest policy announcements on AI, and specifically the recently released directive on the AI Plus policy initiative. So, this is something we’ve written about but haven’t seen covered extensively, at least not very well, at least in our view. So, we’re going to go through that today. We’re going to talk about what exactly China means by AI Plus. Then we’ll look at the various ways that China is looking to integrate AI into different aspects of the economy and society. And one of those, specifically, we’ll look at several areas, but they recent release the AI Plus Energy plan, which we’ll have Cory sort of lay out for us as well.
That’s why we’ve got Cory to focus on that part in particular. But before we get into all of that, I have to start, as usual, with the customary vibe check. Kendra, it’s been a little while since you’ve been on the pod. How’s your vibe?
Kendra:
I have a really basic bit to answer to that question, which is that I just moved-
Andrew:
Careful. I don’t know if this is a R-rated podcast. My mom listens to this. I’m going to get in trouble.
Kendra:
Well, your mom should brace herself. I just moved to the East Coast, and it’s almost fall and the leaves are turning colors, and so I’m here for it. I’m so ready to go for a bunch of sweater walks with lattes. It’s going to be excellent.
Andrew:
Nice. Okay. That’s cool. I like that. Cory, how about yourself?
Cory:
I’m a little less pumpkin spice latte enthused. So, I’m also just back on the East Coast, which is lovely, back in LA where there are no seasons, except it’s beautiful. Simply that. But now I know it’s been good. It’s been good to be back and forth to the East Coast, and yeah, vibes are pretty good.
Andrew:
Good. Glad to hear it. It was good to have you out in D.C. for a while. Good to see you briefly, and we’ll look forward in future pods to talking about some of the takeaways from your trip, but we thought this AI Plus sort of content was ripe for getting out into the world. And also, we had been looking to do this for a while, but we’ve got so many different schedules going on. It was hard to get us all together. So, we’re glad to finally do that. My vibe is, I don’t know, depressed. I’m turning 44 tomorrow and it just seems like, I don’t know… just you can see the years ticking by. I read something recently. There, at least, was one study that said the two points in your life on average where the aging process accelerates as 44 and 66. So, I’m just assuming that I’m going to age terribly over this next year. So, that’s my vibe that I’m bringing into this. Maybe we can solve that with AI.
Kendra:
Great place to start. That’s excellent.
Andrew:
Yeah, really, really cool vibe both. Okay. But that’s where my headspace is. So sorry, listeners. Vibe’s all over the place today, so we’ll see if we can unify throughout the course of the pod. But before we get into the meat of the discussion, I have to do the usual housekeeping as well. First, a quick reminder, we’re not just a podcast here. Trivium China is a strategic advisory firm that helps businesses and investors navigate the China policy landscape. That, of course, relates to domestic policy in China, whether that be the tech space, as Kendra covers closely, the critical minerals space and supply chains, where Cory covers closely — autos, etc., — we cover it all. But it also relates to policy towards China out of western capitals like D.C., London, Brussels, and others.
So, if you need any help on any of that stuff, please reach out to us at hq@triviumchina.com. We’d love to have a conversation about how we can support your business or your fund. Otherwise, if you’re interested more generally in receiving Trivium content, please check out our website, again, www.triviumchina.com, where we’ve got a range of different subscription options that you can look at, both free and paid. You’ll definitely find the China policy intel option that you need on our website. So, check that out. And finally, as always, tell your friends and colleagues about the podcast and about Trivium as a company — really helps us grow our business, grow our listenership, and help to continue putting out great content like this podcast. So, with that, let’s get into it.
We’re going to talk about AI Plus. The big kind of anchor development that we’re going to look at was that on August 26th, the State Council, so China’s government, published a high-level directive on implementing the AI Plus policy initiative. AI Plus represents China’s most comprehensive blueprint yet for how it plans to develop and deploy AI domestically and compete on AI internationally. That’s about as far as I’m going to go in terms of introductions. I’ve got the experts on here to talk us through the rest. So, Kendra, why don’t you jump in, give us a little bit more context and detail on what exactly AI Plus is, and also maybe how it might differ from previous AI policies that China’s articulated.
Kendra:
Yeah, sure. Well, you could think of AI Plus, I’m probably going to get some flak for this, but you could think of AI Plus is China’s second major national AI policy. the first major national AI policy that China released was the New Generation Artificial Intelligence Development Plan, which came out in 2017. And there have been plenty of other policies before that and after that that mentioned AI. But, typically, the entire national policy wasn’t focused on AI exclusively. AI was mentioned in the context of other sectors or as part of a broad basket of research technologies. But the new Generation Artificial Intelligence Development Plan, which is a huge mouthful, was really the kind of first policy that focused on AI exclusively. But that plan was primarily focused on frontier research on AI. It was a big list of technologies, AI adjacent technologies that China wanted to support its labs and companies in pursuing as a research direction and as a research goal — machine learning, computer vision, that kind of thing. And it didn’t really focus on the ecosystem of ingredients necessary to support AI development, like computing power or data, developer ecosystems, or anything like that.
So, the first policy was really a sort of push to get Chinese researchers focused on AI as a development direction. AI Plus, coming out this year, is extremely different. It is focused on AI adoption in six specific industries, and those specific industries are AI Plus science and technology, AI and industrial production, AI and consumption, AI and quality of life, governance and global cooperation. So, sci-tech industry, consumption, quality of life, governance, global cooperation. And AI Plus also focuses on supporting the broader ecosystem of components that underpin AI development in a way that earlier AI policies did not. That means that it discusses how China is going to support computing power, right? Build out its computing power base, create data sets that can be used for AI training, support developer platforms and ecosystems, open-source communities, and then, of course, the talent pipeline to develop sort of Stem talent that underpins AI.
And so this new policy direction is reflective of two things. One, that AI has developed to the point that it is ready for commercialization. And two, it also speaks to a stronger policymaker understanding of the ingredients necessary for AI, particularly generative AI, actually to be successful.
Andrew:
It’s interesting, the way you lay it out, it reminds me just about an issue we’ve talked about before when it comes to kind of the Chinese government, China policy apparatus’s thinking around AI, which is very much about use cases in the real world, right? About adoption into industrial processes and things like that, which is very different from, again, the U.S. approach, which is really like we’ll go after artificial general intelligence in particular, and we’ll sort of figure out the use cases later. But China very much is like, how are we going to adopt this now that it’s pretty much gotten a certain level of capability now? And what are we going to do with it? So, I think it just really, again, reflects sort of the different approaches between at least China and the U.S. on this. But why, speaking of the adoption, is China focusing on those six fields you outlined in particular would you say?
Kendra:
That’s a good question. There are some fields you would expect to be in there that aren’t, but ultimately, each of the areas earmarked for AI integration under AI Plus are areas where the state is facing some kind of sticky economic, industrial or social development problem that past and, frankly, current policy efforts haven’t really been able to solve. I think that’s probably better illustrated with an example. So, I’ll take the first one on the list, which is AI Plus science and technology. And I guess a little bit of background on the problem. Right? It’s become clear to us at Trivium, based on what we’ve seen in the last couple of years, that the state feels like it’s not getting enough bang for its buck in terms of what it’s invested in scientific research versus what China is getting out of that investment in the form of scientific breakthroughs.
The state has thrown crazy resources at developing its sci-tech capabilities. Hundreds of labs, state labs have been built over the last 20 years, with multiples of those being huge national laboratories, hundreds of industrial parks and special development zones have been established, most of that on the state dime. National sci-tech budgets have generally, with a couple of exceptions, gone up every year. They’ve gone up 10% over the last two years, and 2025 national research budget sitting at 398 billion renminbi. Universities are heavily subsidized. China graduates more STEM talent than anywhere else on the planet. And with all of this expenditure, that begs the question, where are all of the earth shattering, world leading scientific breakthroughs that the state really wants to see?
China’s still great at scaling, but not quite as good at leading the pack on lab breakthroughs. Right? So, a couple of years ago, we really saw a shift in the policy conversation. It seemed like Beijing went, “Hey, wait a minute. We have enough resources, we have enough money, we have enough equipment, we have enough scientists. Why aren’t we really seeing the results that we expect?” And the answer that Beijing appears to have settled on is that those resources are being wasted and they’re being used very inefficiently. So, we’ve been reporting on this for a while, the state has sort of undertaken dozens of sweeping reform efforts to identify and break down systemic problems in the basic research pipeline, restructured the Ministry of Science and Technology.
It’s tried to reduce corruption and the research community. It’s changed the funding structure for national research projects, etc. All of that is ongoing. But in the middle of all of this, here comes AI, a commercialized form of AI which can not only help researchers make discoveries by assisting them. Theoretically, AI can make its own scientific discoveries. It can discover its own drugs, right? To a certain extent and develop new proteins, this kind of thing. So, the AI Plus plan essentially calls to use AI as another solution to this problem of scientific and research efficiency to bolster its research capabilities. And, specifically, the plan says that China wants to use AI to enhance scientific discovery and develop new basic research paradigms by creating scientific large models, by upgrading national research infrastructure with AI technologies, and by creating high quality, shared, open scientific data sets.
And AI is the solution to that problem. And you can sort of identify something similar in each of the other areas in which China wants to apply AI under this AI Plus plan. You’re probably, Andrew, familiar with the third one. The third one on the list is AI Plus consumption, right. China is having a really hard time changing the consumption environment, right? I mean, that’s your wheelhouse.
Andrew:
Yeah. So, I’ll dive into that a little bit. I mean, this is an interesting one. I’d say the problem that China is trying to solve, at least in Chinese policymakers eyes, is that, as they see it, people aren’t consuming because the things that they want to consume don’t exist, or China doesn’t produce kind of the quality or type of consumer goods or consumer services that people want, right? And so, it’s a very classic Chinese approach, which is if we invest in the supply side of things and ensure that the types of services and products that people want now and in the future are available, consumption will naturally increase as that availability comes online. So, it fundamentally kind of gets to the question of how do you support consumption?
Most people in the West would say, “Actually, you know, you should support the demand side, basically put more cash in people’s pockets to get consumption up.” The issue is people don’t have enough disposable income to spend, not that the items don’t exist. So, there’s a question as to whether China has diagnosed the problem correctly on this one. Like, this is a consistent approach to consumption policy over years and years and years. So, no one’s going to convince them that they’re wrong. So, the bigger question as to whether they’ve diagnosed the question properly. But either way, the approach is going to be through this classic desire by Chinese policymakers, or not desirable, but framework that the state’s role is to channel resources into essentially the investment side of the economy or overall supply side of the economy, and that if the state invest properly, then the demand side will sort of match it or kind of build it and they will come mentality.
So, that’s where they are. Whether or not that’s right, we’ll see. But here’s some examples from the plan of specifically the types of things that they’re trying to do on creating the types of things that people want to consume, things and services. So, goods and services. They break it into those two categories. So, there’s the idea that they want to expand new service consumption scenarios. And that is a lot of things like strengthening the development of smart consumption infrastructure, improving the quality of life services such as entertainment, e-commerce, housekeeping, property management, travel, elderly care, child care. So, a lot of the areas where the Chinese government and the Chinese populace wants to spend more money, so better elderly care services, better child care services, more opportunities to travel, more opportunities to travel more efficiently.
And so, that’s kind of the general takeaway is we want to improve services through AI so that people will consume those services on an ongoing basis. On the product side, it’s a little bit more ambitious, I guess I would say. I mean, it’s basically like putting the internet of Things on steroids, if you will, and with the idea of that will drive people to buy entire new ecosystems of products, sort of. So, they talk about promoting the intelligent connection of everything, including smart terminals, cultivating a smart production ecosystem, and vigorously developing a new generation of this term ‘smart terminals,’ which by that they mean a range of different things — smart connected cars, AI powered phones and computers, smart robots, smart homes, smart wearables. So, a range of different smart ecosystems, connected ecosystems is what they mean by that smart terminals idea.
And then the last piece on the product side is they want to accelerate the integration of AI with technologies such as the metaverse, low altitude flight, additive manufacturing, and brain computer interfaces. So, very ambitious on the product side in terms of wanting to connect to everything. And the idea that these AI driven consumer product ecosystems will then be much more desirable to the population. So, we’ll see if they diagnose the problem. But at least in the AI Plus consumption side, that’s sort of what they are trying to do, as I see it.
Kendra:
Yeah.
Andrew:
Go ahead.
Kendra:
Exactly. I mean, it’s pretty interesting. I mean, on the consumption of services, they also specifically target the idea of using AI to be a companion for kids and the elderly. So, you’re looking at companion robots or chat LLMs. You know, this is something that’ll be particularly important in light of China’s aging population. If you don’t have enough kids to support the expected explosion of retirees, then those retirees are going to need the assistance of AI-powered tools or automated tools, either software or hardware, which is both a product and a service.
I mean, that seems like a bleak future to me. I, frankly, can just imagine myself sadly, speaking of aging, at 44, like sadly shuffling around in my house as I’m 80 years old, followed by my Roomba which is talking to me. That doesn’t...
Andrew:
Well. I mean, that brings up this bigger question that people talk about is like, is the future of humanity disconnected from each other in person and more connected to these AI companions? It’s interesting to me. Oh, we won’t answer that question. I think most people our age and above would say that’s a bad thing, but younger people are already interacting with AI in different, more sophisticated ways, I think, than a lot of us thought they would. But, I mean, it’s a bigger philosophical question of what we want the world to look like, but it’s interesting whatever the answer to that question is that China is sort of leaning into that future and saying, “Oh yeah, we want AI to be able to do all this stuff.”
Kendra:
Yeah, exactly. I mean, ultimately, it’s a problem, which is that we don’t have enough kids to support the elderly / we want there to be more services that people will need and want to buy. Plus, we also want new products and hardware that people want to buy and spend their money on, you know, AI powered phones, AI powered autonomous flying cars, etc. And if we build those things, people will spend on them, even if they’re bored with their options today or they don’t have the latest options today. And we develop these things so people will pull out their wallet. We actually see a similar pattern there across all of the six sectors that are identified in AI Plus. The AI Plus industry section seeks to use AI to solve China’s supply chain and food security problems, which are also massive. And we probably don’t have time to go into.
AI and the quality of life talks about using AI to boost workplace productivity, which will of course be an issue also considering China’s shrinking population. Right? Also talks about improving health care, which is a perpetual problem. The AI Plus governance section is about making civil service and law enforcement better at their jobs. AI Plus Global Cooperation outlines how China is going to make a space for itself on the international stage in terms of AI governance, and seeks to solve the problem of China getting shunted out or pushed out of global tech development and global supply chains. So, ultimately, the overview is that it is about solving problems with AI, using AI across a range of sectors, pushing adoption across a range of sectors so that AI can boost efficiency in those sectors, help make breakthroughs in science, help push forward consumption, help strengthen supply chains, etc.
Andrew:
Yeah, and this is all early days, right? Which is why it’s important to understand this AI Plus policy initiative at the highest level, which is kind of setting out the framework, to some extent the philosophy and motivation behind really the long term goals that they’re trying to achieve with AI and to put clear frameworks like breaking them down into these six specific fields. But my understanding is that the state will release separate, more detailed policies for each one of these fields, and probably subfields. And we’ll get into one of those with energy just here in a second. But the text of the AI Plus policy, as it currently exists, really just gives a high level overview of what exactly the state wants to integrate AI with in each of those areas, although not a lot of specific at the moment. Is that right? Is that how you see it, Kendra?
Kendra:
Yeah, yeah. AI Plus should basically be seen as a tone-setting preview on the types of industrial policies that China intends to release related to AI going forward. This is just a taste. It’s just the starting line. Now, what is expected to happen is that different ministries and regulators are going to start issuing policies like AI Plus consumption, AI Plus health care, AI Plus education, right? Policy specific to their sector that give many more details about what types of problems AI is going to solve and how AI is going to contribute to economic efficiency in that very particular sector. And this has already started. Just a couple of days after the AI Plus plan came out, the National Energy Administration and the NDRC released just another big AI Plus policy called AI Plus Energy, which outlines how China wants to use AI to tackle systemic issues in the power grid. That’s what Cory is here to talk about today. So we’ll go to him in a second.
But I also just saw yesterday that the Ministry of Transportation announced that AI Plus transportation is going to come out shortly, right? So, we’re going to see sector specific plans. We’ll also probably see local governments release a bunch of AI Plus for just Beijing, AI Plus just for Shanghai, AI Plus just for Shenzhen. We’ll probably see those sort of targeted policies where local policymakers identify important industries in their areas and then seek to affect economic upgrading using AI in those specific industries.
Andrew:
Yeah, well, that’s the perfect place to segue. Cory, you’ve been very patient here. I appreciate that as we kind of lay out the high-level stuff. But good spot to dig into this AI Plus energy plan that was released recently. Give us some of the details on what’s in there and how you’re thinking about it, why it matters what the government’s trying to achieve here.
Cory:
Absolutely. So, first thing, this is a big plan. It’s called implementation opinions. But the plan is, in my view, more collection of ambitions than a concrete roadmap in many ways. So, it’s much more specific, say, than the AI Plus framing guidelines. But as Kendra is getting at, that we’re going to see sectoral plans like AI Plus energy proliferate. That roadmap is pretty set. But we’re also going to see a lot more sub plans beneath this. That’s sort of implementing each of the 30 odd specific targets within this plan, or specific to sub-areas that Beijing wants to invest in AI vicinity. By saying it’s a collection of ambitions, that’s just not a critique at all. It’s to say that this is a truly complex endeavor, and this is a really nice way to pull together.
It’s the first time I’ve seen this comprehensive of a blueprint. So, the very simple question, what can I do for energy? That is a very simple, straightforward question that if you ask a thousand people, you’ll get 2000 answers. Right? And that’s what part of what this plan does is to set the framework of here’s what we’re thinking about. So in that context, I would like to focus on what I think is, and I think we agree, is the most important piece which is the grid. Now, I should note when we talk about the grid, we’re talking about electricity as opposed to all the other parts of energy in the energy systems, so oil and gas and all that stuff.
That is, in some way, it’s a smaller part of China’s energy systems as some will realize. Electricity is about 30% of final energy consumption in China, about 70%,68% is not actually electricity. But at the same time, the grid is by far the most important part for development because electrification is happening very quickly, not only with electric vehicles but also industrial applications. It’s also the most complex to manage given the role of renewables and the fact that the grid is so dispersed across the country. It’s not commodifiable. You can’t ship electricity that way you can ship a barrel of oil. And so, the actual management of the grid is incredibly complex. So, it raises the question how can I help the system.
So, there’s really two sides of the picture I will talk about. We’ll start with that. The other side of the picture is though it’s not really the point of the plan, but I think is a perfect opportunity to talk about, which is the role of energy for AI. I mean, the plan is all about AI for energy, what’s the role of energy in AI? And part of the reason I think AI Plus energy plan is so important is because energy is the lifeblood of inference in particular, but also training. And so, long term, there are lot of grid challenges that will have to be solved for the AI data center buildout at its current pace to survive. There’s going to be bottlenecks if something isn’t done, and this plan is effectively what is to be done. So, there’s that angle of it too. So, diving into what the plan says for the value of AI for energy, where is Beijing looking?
I mean long and short, energy has just been an incredibly, incredibly complex, incredibly high stakes balancing act. The interesting piece about this is that almost all the components are actually really highly digitalized already, which is great because it makes the combination of high stakes, complex, and very digitalized, makes energy the grid, specifically electricity, a perfect target for AI driven optimization. So, why is it complex? I’m not going to go into too much detail, but at the end of the day, what we have is an equation where you have thousands of supply nodes of varying sizes and characteristics and millions of demands nodes — everything from people going home and cooking dinner to industrial power plants to medical centers, to now AI and data centers.
That supply side goes up and down, increasingly with more renewables, but always has been up and down for various reasons. And then the demand side is varying constantly as well. And the stressors are even more significant as we electrify more things and have more power demand. And all of this has to be perfectly balanced, not just on the day ahead market, but second to second, sometimes sub-second level, there’s trading happening. How do you actually manage something with that degree of complexity? And the fact that it works at all, frankly, I think is one of the less appreciated modern miracles. The grid has always been a technological triumph, in my view, but it’s getting more and more complicated over time. And that’s something that the plan takes very seriously. And frankly, I think every government should be taken very seriously is there’s no silver bullet.
We do need an increased, you know, renewables, not just for climate, for decreasing costs. There are system requirements. You have to make that manageable, you have to manage that variability. And that’s going to cost money. There’s also the fact that AI is a different type of demand than the grid is designed to build. I’ll touch on that in a bit more detail. But, basically, we need to plan across the board, not for let’s have virtual power plants and that’s it. Or let’s build transmission infrastructure and that’s it. There has to be a coordinated approach to optimizing how the system works overall. What’s interesting about this plan is that it pulls on like every lever I can think of for how you could possibly try to optimize the grid.
And typically, those are things that are too complex to do manually, hence the role of AI. That’s the long and short of it. So, when we come to the, I guess, relevant facets of this, this is an incredibly high-stakes. You can’t fail at the grid. If you have any grid instability, you end up with brownouts of blackouts, which means, one, beyond a very unhappy populace, you also end up with industrial shutdowns, medical incidents, you know, a medical center goes down because a blackout. Grid is not something anywhere in the world that is treated lightly. And so, trying to change how it works is very sensitive. At the same time, not changing it and letting it stay the way it has been for 100 years when the world is changing around it is also not safe.
So, it’s a very sensitive area. It has to be treated very cautiously. And this is kind of one of the key issues we see across the world. Everyone’s facing this issue. Every country that has a data center will face the same issues as China. This is not unique. What is going to be differentiated across governments is the degree to which they leave it to the market to figure out which is the role to which governments are involved in trying to figure this out with smarter companies.
But anyway, so that’s another piece of the plan. Beijing is effectively saying — we don’t think the markets can handle this. We’re going to build an ecosystem. We’re going to try to encourage optimization of those things so that data center and digital economy can build out without constraint. And so the rest of our economy doesn’t get screwed by the increased demands of the digital economy, etc.
The last piece there was, how is this even possible? I mean, this is one of the first questions I got when we started posting about percentages. Okay, this is just a pipe dream. Is this even doable? And it’s a good question because when it comes to AI applications, a lot of the time it’s like, you know, Kendra has great work, and we can talk for hours about this, but there are a lot of things that it makes sense to try to optimize in some AI system, but there’s no data.
If it’s not digitalized and there’s no data, and you can’t control it. What is the role of AI? There’s so much work to be done before you can actually use AI. That’s not the case with the energy system. With the electricity system in particular, it’s highly, highly digitalized, it’s how it runs in the first place. And so, in that way, it’s kind of a perfect starting point, which means it’s already primed for AI applications. That’s in contrast to, say, a lot of industrial manufacturing, where some nodes are highly digitalized and some have a long way to catch up. There’s a lot of inconsistency between different parts of the industrial ecosystem, and many of the systems already that digitalized aren’t compatible with each other.
Kendra:
Yeah. I mean, to my understanding is not only is the power data digitalized, but that data is to some degree centralized. There are only a couple of state-owned power companies. And so those power companies hold that data and that’s fairly easy for Beijing to access on a unilateral basis. While in a sector like manufacturing, there are thousands of manufacturing companies, many of them are not necessarily state owned. They do not necessarily want to share their data with the state. It’s all different types of data. It’s all held in different databases. I can understand why power data would be an attractive target for AI upgrading, given the state of the data as it stands.
Cory:
Yeah, absolutely. It’s a great point. Yeah. So, the two utility companies, State grid, it’s about 80% of China, and Southern grid is about the other 20 in Guangdong in the southern region. And they are really responsible for making sure the grid functions. There’s a ton of power actors below them. But at the end of the day, there are two utilities that manage everything. And so that’s exactly right. So, with that said, there is the flip side of the question that I actually think is really interesting. For those who want a bit more detail, I’m keeping this at a more strategic level, but we’ve actually done a lot more work on the specific applications that China will encourage to support electricity grid optimization and managing loads and stuff.
But I think a lot of the importance of that becomes more evident when we actually talk about energy is important for AI and other applications. It’s just a little bit easier to see why this matters so much. So we turn now to energy’s importance for AI. And again, I think this is a real reason that energy is an early focus. As much as I can help energy, energy is vital for the digital economy overall. So, first thing first, one of the questions we get a lot is, is energy bottleneck or a competitive driver for AI in the digital economy? And the short answer, I think, and this is why you’re going to get two views is kind of extremist yes and extremist no is not a problem. The reality of course is in the middle. Right now, no, energy is not a competitive driver in AI. That’s really down to chips and talent and deployment to make models of people use, etc. But energy is the lifeblood of both training and inference. And the problem is the demand is growing so much faster than the grid can handle that here in a short while, we will have a big problem.
And whether that manifests is really a question of foresight. And I’m going to get ahead of myself a little bit here, but it’s worth noting that in the U.S., for example, companies generally bear the brunt of ensuring energy supply for their own AI data centers. The problem with that is, I’ll get into it in a minute, is that there are certain grid constraints that they’re not in control. They’re going to run up against walls. And either those walls get expanded, and so there’s more room to grow or they don’t. That’s not up to the companies. China is doing the opposite. China is saying — we want to make sure this economy, this digital economy expands as it wants to. So, we’re just going to go ahead and make sure that there’s space for them to grow into.
Kendra:
Well, that makes a ton of sense actually, because I mean, let me just sum up what you’re saying really quick to make sure I understand it, which is that AI requires data centers to function, data centers require electricity to function, and the demand for, and the expected demand for data processing, and thus electricity is going to boom. And it’s especially going to boom in light of the fact that China’s AI Plus initiative intends to ensure that it does by shoving AI into every single sector. And so, you’re going to see a massive uptick in demand for inference, the process of asking AI a question and getting an answer, essentially using an AI tool. And so, China’s doing two things in this AI Plus energy plan.
Number one, that we use AI to ensure there’s enough energy to use AI. And then additionally, to just use AI to make sure there’s enough energy. Period. Paragraph. Okay, so we’re talking about now this issue of needing an expected boom in data processing power required to run AI services nationally. How much power are we talking about?
Cory:
Yeah, absolutely. First, that’s exactly right. And the key issue here is that, just to put a fine point on it, nowhere in the world has ever developed a grid to manage something like the modern AI data center. It is just not the kind of thing that grids are designed to handle. And I’ll explain why — So, with grids, the issue is not really about generating power so much as actually getting it where it needs to go. And this is where the data center issue becomes so tricky. The three issues. First is AI data centers require so much power, just incredible amounts of power.
The way that inference is being run is basically really, really dense clusters of very power-hungry chips. Right? And for context, a traditional data center that was, you know, before AI might have acquired 10, 20, 30MW. Don’t worry about, you know, how this relates, just 10, 20, 30 kind of that range. Today, inferences, about 200-megawatt data centers typically, and we have plans on the books for up to 1000 or 2000 megawatt data centers. The scale is huge, frankly. And the problem with that is it’s not just making that power. China has a bunch of solar and wind and coal, and it can make that power. But grids have transmission and distribution. It’s not completely unlike highways and streets. You cannot just dump Beijing levels of traffic into a small city like shown Xianning, Anhui, and not expect there to be backup.
We literally call it congestion on the grid too. You can’t necessarily tap so much power in a grid that is not handled to dispatch that. And the problem with data centers is all of that demand is in a literally one facility plopped somewhere, and the grid around it has to adapt to accommodate it. Another issue here is that inference in particular has very inflexible demand. No one is like, “Oh, we’ll answer your question six hours from now when we have cheaper or more available power.” It’s like pretty much on demand. And so basically what happens is inference data centers run constantly. So, there have been big power demand, power hungry projects on the grid before obviously. But one of the ways they’re usually handled is, well, they’re not always online.
And so you manage which ones are online until a sudden shut down. And so, for example, one of the biggest power-hungry demand centers in China and globally beneath AI data is aluminum smelting. And you see two things. One, they don’t run all the time, and two, they are shut down whenever there’s not enough power. Just actively are just called to shut down. You can’t really do that with an AI data center. When someone you know wants to be competitive, you can’t say like, “Oh yeah, sometimes we just don’t serve our clients.” That’s is not going to work with them, right? And so, the other thing actually, the third thing you see with aluminum smelting that’s interesting is, as an example, they’re only built in certain geographies. Like you see in Yunnan where there’s a bunch of hydropower that isn’t being used for much else.
And you see them in Inner Mongolia and Shandong, where there’s just tons of coal being burned on site to fuel these. So, you’re kind of getting around these issues. If you have a data center in Guangzhou or something, you don’t have the flexibility there. And third is, of course, the growth outlook. And that’s the piece that rightfully gets the most attention. But, at the end of the day, there are other big things that the grid manages in various ways. But none of those things are growing 1,000% a year. Things like alumina smelters, electrified rail networks, medical centers. You don’t build hundreds of new ones of those every year. We’re going to build hundreds of data centers every year, and they’re going to get bigger.
And they’re not just going to be in the desert. That’s the issue we’re facing. Now, we’re not currently in a crisis. I don’t want to be in panic mode on the podcast, be like, “Trivium says…” No. But it’s the kind of thing that you will eventually run into, very hard limits on what the grid can handle. And the issue is that you can’t wait until the issue materializes to take care of it because data centers come up very quickly. Infrastructure doesn’t. China can build infrastructure, especially grid infrastructure, much more quickly than the U.S., and maybe a few years as opposed to 7 to 10. But it’s still a matter of years to build a lot of this stuff. And often it takes longer than that to get it to functional. Meanwhile, you might have 100 new data centers pop up that year and a three year development pipeline for the stuff to help you actually supply.
You have to get ahead of the problem, and that’s kind of what we’re seeing as part of not only this plan, but other by the utility companies.
Kendra:
I mean, that’s really interesting as well because we’ve been watching without that context. We’ve seen policies over the last ten years, frankly, where the national government tells local governments to, please, God, please stop randomly building data centers in the middle of grid infrastructure that cannot support it on some kind of land that is prone to disaster. So, there’s been a lot of data centers that were developed in a way that they do not integrate well with the grid. So, I could understand trying to use AI to rationalize and optimize some of that poorly planned capacity.
Cory:
Absolutely.
Andrew:
Well, and I’ll also just throw in, in a way, it sort of raises the stakes here, meaning inefficiencies become much more impactful and potentially catastrophic, sort of, as you said, like if you build a bridge to nowhere or you build an extra apartment building or you build a government office space that’s unnecessary, yes, the asset has to get written off, but it’s not like proactively sucking energy resources and electricity resources from competing demand centers. Right? So, it just strikes me that the scale of this and the speed of it, when I say this, I mean the AI buildout and integration of AI into everything, and the infrastructure under that is so large that those inefficiency, again, inefficiency seems like too small of a work. If you don’t plan it out properly, it could be catastrophic and totally kneecap or undercut your AI ambitions from the get-go.
Kendra:
Nationally.
Andrew:
Yeah, exactly.
Kendra:
That puts great context around why the AI Plus energy plan came out first of all.
Cory:
Yeah. And I mean, it’s funny, the plan, I mean, when you look at the text, it’s almost all about like, here’s how AI will apply to hydropower versus renewables versus this versus that. But why does that all matter? It’s because we’re trying to extract the most that we can out of all of these systems, but most importantly allows them to work together efficiently in a couple dimensions. One is when you have massive, inflexible demand that putting major stresses on the grid. That’s the reason any of this matters. And AI data centers are one of the top, the reason that that’s such an issue now. And I’ll say, you know, there’s a vision here. It’s not a complete roadmap, but we’re going to see more plans that come out to, yes, let’s increase demand and flexibility and all that stuff. There’s more plans on how to do that yet to come.
But the basic target of the AI Plus energy plan is by 2027, build out the basic systems, kind of the usual policy language. Get a basic system in place for coordinating development of computing power and all the energy that will supply it. So, again, even though the text is all about this particular energy source, that particular energy source, the top target there is actually on coordinating computing power with energy to supply it. And I think that really lies, you know, it shows what the real motivation there is. And to that end, just to fly quickly, it wants replicable demonstration projects. And that’s recognizing that there is the grid and there are two grid companies in China. But the grid, so to speak, is always inherently local.
Again, because of these transmission and distribution bottlenecks. Long distance trading is still very nascent in China. Even within provinces, trading can be very messy and difficult, and they all have different characteristics. Every province basically has its own little subsystem. And so there needs to be kind of localized, adapted approaches to this. You can’t have a one size fits all that’s going to fix all of your local energy issues because your local energy supply and demand sites, neither of them look like the rest of the country. And so, we’re looking at replicable demonstration projects for various systems, technical standards and financing. All the usual stuff. I mean, just for the technical folks, yeah, for my money, I do think the demand side portability issue is going to be number one.
Like, you’re not going to tell AI companies that they have to stop using inference for periods of time the way you can tell an aluminum smelter to stop functioning for a few days when the hydropower reserves or the basin gets low. That’s not going to happen. Training, you know, there’s flexibility in training, saying, “Hey, maybe train, you know, whatever you’re doing, model workload tonight instead of during the peak load of the day.” Maybe, but the other issue is my understanding training is rapidly becoming a much smaller share of total workload with AI. And so, your space with flexibility on the AI side, it’s really small and shrinking. And so, it’s the rest of the system that’s going to have to become more flexible to allow for AI to be sustainable, not sustainable in a high climate sense, but in a kind of energy consumption sense.
That’s not a new project, but it’s a project that’s been going on so long because it’s so difficult to tackle. There’s a lot of market reforms going on this infrastructure developments going on, all this stuff that’s happening, and it hasn’t solved the problem of demand flexibility in China. It hasn’t solved a lot of the other grid issues. And now we dump AI in, and it does make sense to me. It’s like the only hope, maybe this is overstating the case, but certainly a big piece of the picture is like, what haven’t we already done that could help manage this? And the answer is smarter, more efficient systems. And, again, we can talk more detail later on about specifics, but that’s what this plan is really about. Making a smarter system that can let your AI make everything else smarter.
Kendra:
Yeah, I mean, that falls right in line with the rest of the AI Plus plan, which is essentially about — hey, look, we’ve already tried to boost consumption and all of these different ways. It didn’t work. Hey, we already tried to galvanize sci-tech research in all of these different ways, and it didn’t work. Well, it didn’t work is a strong term, right? That it hasn’t been as effective as we wanted. And frankly, we’re kind of out of ideas. And so now it’s about we can add this
Cory:
And we’ll confirm, it’s not a pipe dream. I mean, for what it’s worth, like virtual power plants, they do work. We’ve already seen a number of local applications. Do they work at scale? Do they work at certain cost levels? That’s all to be decided. There are a lot of AI applications that are proven, at least in principle or small scale. So, this isn’t magical thinking, but it’s still there’s a long, hard road ahead to make this functional at the scale required.
Andrew:
Yeah. Well, that’s a really excellent kind of dive into the energy specific piece of it. I think it’s obvious probably to listeners at this point why it is likely that they put this energy plan first. Right? We haven’t seen anything written specifically from the Chinese government that says we did AI energy first because it’s so foundational, but it just strikes as it’s quite obvious. I do want to back out, though, a little bit and speak a little bit more about the broader AI Plus policy initiative, again, now that we have kind of seen how detailed the new AI Plus X plans are going to get. The first half of the policy document that we talked about at the beginning, Kendra, really focuses in on those six fields and the rest talks about more foundational areas like foundational models, compute, data supply, data security, open source developer ecosystems, and talent, and what the plan is for those things. So, talk us through that second aspect of this bigger AI Plus policy document.
Kendra:
Yeah. So, each one of those areas probably deserves its own podcast. I mean, we keep saying we’re going to do one on open source. We should just do that at some point. But for this podcast, I’ll probably just stick to talking about two of those, specifically how China says it’s going to develop compute and foundational models. And the reason these are interesting is because the directives here shed a lot of light on how China’s strategy for pushing back against U.S. export controls on semiconductors is intended to work. In short, it sounds like AI Plus makes it sound like China is going to attempt an end run around export controls by leveraging technologies that make per chip computing power less important. Right? So, okay, Chinese chips are not as powerful as Nvidia chips on a per chip basis. It’s going to be a while until China compete on per GPU power.
And so, the question is, what can China do to stay competitive in AI applications in the intervening years while it frantically tries to sort of catch up on the manufacturing side? And paraphrasing from AI Plus, some of those strategies include one developing AI models that consume less processing power, that are designed smartly. And so, they are just not as big of an energy hog. Two, accelerating the development of ultra-large-scale intelligent computing clusters. In other words, using more but crappier chips instead of less, more powerful chips to ideally get the same result. And then three, supporting innovation in chip software ecosystem. So, that’s basically a three point strategy for not necessarily needing to be bleeding edge on the GPU manufacturing side.
Andrew:
Interesting. And I mean, you can see sort of inherently, in that three prong strategy, why the energy piece of this is so, that Cory just talked us through, is so fundamental to their approach. It’s kind of a lot of what they’re trying to do is boost efficiency so that it requires less overall resources or fewer overall resources and energy. But do you think that three pronged strategy will work?
Kendra:
I mean, well, you’re absolutely right. Particularly interesting in that is the development of this ultra large scale intelligent computing clusters. Those are going to require massive, massive amounts of energy. Those are the energy hogs that Cory was talking about that are sitting right on the power grid sucking out massive amounts of mega wattage. In terms of whether or not the strategy is going to work, I mean, I’m going to be a little bit of a coward and punt on that question because I’m not a chip engineer. But I do think that what’s interesting here is we’ve actually already started to see, like, I think China thinks it’s going to work. We’ve already started to see this, those three strategies playing out in China’s AI industry, deep seek, as we have heard ad nauseum, is a model that attempts to use less energy to achieve effects similar to more energy intensive models.
And, interestingly, on the chip clustering piece, just last week, we saw Huawei announced that over the next two years, it intends to launch a series of hardware stacks that will cluster tens of thousands, potentially millions of domestically developed chips into basically huge servers that work like Legos. Right? These sort of powerful server blocks that can be networked together to form a more powerful servers, like a transformer or something. There’s been a lot of controversy about that announcement on whether or not we can actually pull that off, given its limitations and chip manufacturing capacity is a matter of great and furious debate at the moment. But that is the plan. And that mirrors the AI Plus plan, right? That is a reality that Huawei has announced that has mirrored what is in the AI Plus plan in terms of export control strategy.
In terms of software ecosystems, same thing. Just a couple of weeks ago, we saw Huawei announce quite clearly that it is open sourcing its chip development software CANN, which is Huawei’s alternative to Nvidia’s CUDA, which is the development environment for basically a package of software libraries for chip development. And the reason that Huawei has open sourced that tool is explicitly in order to catch up to Nvidia’s capabilities. Huawei doesn’t think of itself as a good software company. It’s a great networking company, it’s a great hardware company, but it’s a pretty garbage software development company, and that software is a really critical piece of the technology stack. If the software isn’t good, chipmakers can’t design on it. If they can’t design on it, then Huawei’s whole stack is kind of dead in the water.
If they can’t use the software, they’re not really going to want to use the hardware either. So, what’s interesting to me about this is that it seems to me that the realities of what Huawei and deep secret doing actually defined this policy, and not the other way around. I don’t think policymakers had enough know-how to sit down in a room and say, “You know what we’re going to do? Ultra-supercomputing clusters. That is going to solve this problem.” Like a bunch of officials did not figure that out. What happened is they talked to Huawei and said, “What are you guys working on? What is the most promising solution to this problem?” And Huawei said, you know, we’re working on this clustering system. And then policymakers went back and, you know, as part of the drafting for the plan and said, “Okay, this is what’s going to go in here.”
I mean, this plan came out I think, days before the clustering servers were announced. I mean, that’s not different from the United States, which, in the best of times, goes and asks its industries, “What are you guys doing?” Bureaucrats don’t know what future technology is.
Andrew:
Definitely. I wanted to bring it back to that. Exactly. Which is it is interesting to see the interplay of government policy and industry. I was going to say the private sector, but people would scream at me and say, “Well, Huawei is not a private sector company,” and all that stuff. We know? Okay. But industry players and government, that happens in China. People really don’t think about that process all that much. It also happens in the U.S., right? But they’re playing out in two very, very different ways. And so maybe a good kind of wrap up is to bring it back to sort of the U.S.-China competition on this front. What is AI Plus, as we know about it now, or our current thinking on it, what does it tell us about how China intends to compete more broadly with the U.S. in the AI race?
Kendra:
I mean, I actually think this policy really underscores how differently, I mean, you mentioned it a bit in the beginning of the podcast, but how differently the U.S. and China framed the so-called AI race. As we’ve mentioned before on the pod, the framing in DC.. is that the US and China are in a race to AGI, artificial general intelligence. The problem with that is there’s no single universally recognized scientific definition of artificial general intelligence. I mean, OpenAI defines it as an AI system that is generally smarter than a human and capable of outperforming humans at most economically valuable work. Right? So, it’s a machine that can learn faster and more efficiently than a human. And then, like a human, apply that learning to domains it didn’t have any familiarity with before.
And that’s to differentiate it from an AI system that’s only good at doing one thing like image generation, for example. So, we’ve basically defined, in a way, we’ve defined a specific targeted technological outcome. But in another way, that outcome is very ill defined. And this framing of the U.S.-China AI competition in terms of a race to AGI, I think is pretty reminiscent, and probably informed by Cold War era hard science races of yesteryear. Right? The race to the atom bomb, the space race. I’ve actually been to a couple of talks in D.C. about the so-called AI race in which this speaker’s slides were kind of covered in pictures of mushroom clouds and rockets. So, I think our framing of the great power sort of sci-tech competition is still, to some extent, stuck in this decades-old framework.
Andrew:
Can I just jump in and say that every time I talk to somebody in the U.S. government, they’re talking about how they are doing a “Manhattan Project for XYZ.” Like that’s the language we have to talk about. And it’s very much how they frame it. So, that’s a really good point.
Kendra:
It is. The thing with that is those hard science races, I think the difference between the actual Manhattan Project and the race to AGI is that those races had a very clear finish line. The race to the atom bomb is over when the bomb goes boom. The finish line of the space race is when somebody lands a man on the moon. The promise of winning those races was also pretty obvious. It was a clear military advantage, first and foremost. Secondarily, those technologies, the development of those technologies would lead to more scientific breakthroughs in a couple of other key sectors. Right? Nuclear bombs gave us a major military advantage. They also led to advances in civilian nuclear power and early computing and medical imaging, blah, blah, blah.
But what we get from developing foundational digital technologies like AI is really different than what we get for making a scientific breakthrough in a discrete sector. A digital technology is not just about military advantage. Of course, that’s part of the equation. But the promise of digital technologies like these is what we’ve been talking about, that they can boost economic efficiency across every sector of the economy and society, and that they’re going to fundamentally change the way that we live, everything about how we live.
So, what China is doing in the AI Plus plan and what they AI Plus plan really says about China’s mentality is that China is treating AI less like the atom bomb and more like electricity. It sounds really dumb to say the race to electricity. So what happened? The whole world went through a multi-decade process of electrification, which, as Cory mentioned, it’s still going through that involved every single company in every single country changing the way they produce and built, and every single individual changing the way they do every task in their lives — the way they cook, clean, work, entertain themselves. And it involved every government body participating in building the infrastructure necessary to support that change.
We had the same thing happen with the internet, to the extent that there was a competition. The competition wasn’t about who got the internet first, it was about the rapid adoption of the internet. There are, of course, countries that did a garbage job of adopting electricity and the internet, and those countries are in a very bad economic condition today. They’re egregiously behind the global curve. And, of course, the countries that did the best at adopting and diffusing those tools are comparatively doing economically much better. So, that’s really how China’s thinking about things. I think that’s the key difference that we have sort of put this, the stake in the ground of we’re moving towards AGI, and China’s kind of thinking about this as a sort of foundational shift in human development.
Now, to be clear, I just want to underscore, because someone is going to yell at me later — I’m not saying that China is not pursuing frontier AI or AGI, right? We know that DeepSeek has stated that its goal is to reach AGI. We know several state labs are pursuing frontier AI tools. I wouldn’t even be surprised to see an AGI policy emerge at some point. Of course, well, if a technology is driving such massive change in the way that humans live, of course, you’re going to chase the most advanced version of that technology, right? As far as the line of research will go, you’ll chase it. And since AI is going to be applied in every domain, of course, that includes national defense. This will lead to military development. So, I’m also not arguing the U.S. AI or U.S.-China AI competition isn’t a NATSEC issue. It is. But the U.S. framing of AGI as the end state really doesn’t track with the type of technology we’re dealing with. And I think that’s the critical breakdown.
Cory, I just want to wrap up quickly, when it comes to how the U.S. is approaching the energy piece of the AI race and how you’ve outlined how China is, what do you see as the differences there? I mean, my guess is that the U.S. isn’t really approaching this holistically at all. It doesn’t seem like. But tell me if I’m wrong.
Cory:
Yeah, absolutely. And I love Kendra’s framing and the irony of using Manhattan Project kind of language around AI, and this ties in directly to the energy side as well. One additional piece of note is tell me if I’m overthinking this, but I think it’s ironic that as much as the U.S. government has more of a Manhattan project, the kind of sense of AI, it’s also, if you really wanted to do a very government heavy kind of project, Manhattan Project type AI development, you would act more like China than like the U.S.
So, in some ways, it’s having opposite views. The views are kind of opposite the actions we’re seeing. And on the energy side, what I mean by that is, well, actually jumping back, the project analogy is really useful because the other piece I think is this is commercial first or at least commercial simultaneous. And that’s the completely opposite from a lot of like the Manhattan Project space race. That was stuff that the private sector industry could not do by itself or in many cases, even if it could have done, shouldn’t have done by itself versus things like the internet. That is obviously the DARPA origins, but what it was, was an opening of an ecosystem in which a lot of commercial applications flooded in at the same time.
And so this kind of simultaneity, and AI is more like that. AI is not like a bomb. And is not in terms of the commercial participation of prospects and what’s actually driving it’s development. Frankly, the government would not be investing in AI the way that the commercial sector is because it’s not as much of a driver at this stage, until it gets good enough and suddenly NATSEC wants to catch up. But, anyway, very, very different drivers. And I think the kind of proof in the pudding of how the rhetoric in the U.S. differs from the actions and the willingness to act in the U.S. is, again, if you really want to support where the government’s kind of centralized push to support AI, even if you’re thumbs up with AGI is a little unfounded or a little ill-formed or whatever, if you really want to promote AI, what should you do?
At the very most basic level, you should make sure that your AI developers and your inference, you know, focused data centers, etc. have everything they need to have low-cost power, have the ability to produce whatever makes the most sense for them to produce in terms of near demand center. So, if it’s in Virginia or Atlanta or Nevada or wherever, you should make sure they have what they need, and that the current prices won’t skyrocket and that their energy demands will not skyrocket the prices of the local ratepayers. So, right now, all of that stuff in the U.S. is basically being left to the companies and to local utilities who are not in a position to handle this.
And I will note there are applications, like a lot of data centers in the U.S., build their own power plants. I mean, their own, you know, we call it captive power or distributor of renewables in some cases. So, they’re not entirely dependent on the grid. So, rather than fix the grid issues, we just kind of make our own-
Kendra:
Privatize. Privatize, privatized, privatize.
Cory:
I kid you not, I’ve had multiple conversations over the last couple weeks of, okay, when will small modular reactors come and save our asses? And the answer is no one knows. And we can’t wait on a 10-year development timeline that might be 20 years. No one knows when this stuff is happening, or if it’s even going to be usable as a baseline, or as a foundation of data center and load support. And yet all these uncertainties, where is the government action if it really wants to support this? I don’t think there’s anything inherently wrong, by the way, with saying the private sector should figure out. But you should be providing your grid companies and local utilities the support needed because that’s a natural monopoly. It’s state governed. It’s state overseen and managed.
So, do your part. And then when Beijing does something to do its part for a natural monopoly such as the grid, it gets wrapped up in this very problematic cold war discourse that it’s very misplaced in my view.
Kendra:
Yeah.
Andrew:
Well, that’s all really interesting. It’s kind of depressing because it does seem like China, as per usual, has a quite sophisticated and well thought out approach here.
Kendra:
It’s a downer today, though. Jeez, downer vibes. Oh, good lord.
Andrew:
I know. I’m turning 44. Well, you guys are the ones cursing all over the place. My mom’s going to kill me. All right. Well, on that unfortunately pessimistic note, we’ve gone long. It’s been a great discussion that we will continue to look at all this stuff in the future. We will wrap it up here, but I just want to say thanks both for these great thoughts. I think, Kendra, the piece on how we should think about this technology, just from a framework point of view and what it is, an atom bomb versus internet or electricity is great.
And that’s one we should really, really kind of try to get out in the world because I think that will be helpful to the discussion. Cory, excellent stuff. Getting us into the energy piece of this couldn’t be more obvious now how critical the energy infrastructure and supply to all this will be in China and elsewhere. So, thanks to you both for joining me today. Appreciate it.
Kendra:
Absolutely.
Cory:
Thanks so much.
Andrew:
All right. See you next time, everybody. Bye.
