China builds greener data centers in Xizang

Added: 2026-05-22

[00:00:01][music] This is Climate Watch. [music] I'm Jaoyang. Welcome to the show.

[00:00:13][music] The rapid rise of AI has triggered an unexpected surge in energy demand along with increasing greenhouse gas emissions.

[00:00:26]In China, part of the solution is taking shape far from the country's eastern techups. In the high altitude regions of Xi Jang Autonomous Region, new data centers are harnessing cold temperatures, clean energy, and even waste heat to reduce the carbon footprint of AI computing. These projects are part of China's broader eastern data western computing strategy which shifts energyintensive computing tasks to the country's western regions where renewable energy resources are more abundant. So how green are these high alitude data centers? Can efficiency gains keep pace with AI's rapidly growing energy demand? And what does this digital transformation mean for Shiang itself? For these questions and more, we are joined by Jao Jung, professor of international political economy at the school of international studies at Ping University. Professor Jai, it's great to have you on our show.

[00:01:24]>> Thank you.

[00:01:25]>> So, first of all, what advantages do high alitude regions like offer when it comes to data centers? Well, I'm not an engineer when it comes to such topics, but uh I have been to Gujo on a couple of occasions and uh the data centers hosted there have a at least one decade of history and then you have major Chinese companies including some foreign companies that hold a data center there.

[00:02:01]my understanding of hosting a data center at a high altitude region like Shiang um or for that matter in English they often refer to that as Tibet is that it's energy saving because the weather is more temporate uh and then throughout the year you have regions there specific spots there that do not get overly hot and data centers by nature uh consume a lot of energy and it can become overheated.

[00:02:36]>> The Yajang 1 computing center uses an innovative solar power plus waste heat recovery plus high efficiency cooling system. Can you walk us through how this mechanism works?

[00:02:48]>> Solar power is clean and easier to operate than conventional forms of power. But solar power faces the challenge of transmission to consumption centers. So by hosting a data center in Shang, I would think it makes perfect sense in terms of generating the electricity that's clean and efficient but without having to invest in power transmission and then um waste heat and other uh uh forms of energy that's locally available and the cooling Again, I'm not an engineer, but then this >> as I said earlier, uh combined with the more temporate uh weather or temperature systems there, it would be highly energy saving, more energy saving than other parts of the country. and then was through these cable lines which are you know a uh proven to be efficient in transmitting data back and forth. This is a classic example of uh efficiency maximization when it comes to the use of electricity for powering data centers. China is fortunate in being able to do this.

[00:04:18]>> Okay. But what about the water usage?

[00:04:21]Because we know data centers around the world are increasingly criticized for their water consumptions. But does the cold climate in Sizan um significantly reduce the need for water-based cooling systems?

[00:04:34]>> Cooling systems using water does mean that the same water can be reused again.

[00:04:43]Now when people talk about the competition in consumption of water that usually refers to the competition between diver uh diverting the water that's often needed or that is needed for daily consumption especially by population centers whereas here uh that creates a competition or water for irrigation whereas in Shizang where it's very likely and in some parts of uh the autonomous region you would have uh no population uh in some parts of Sizan.

[00:05:25]So in this case the competition between water for cooling and water for other purposes is not as sharp at least or in some cases not existent. And the bottom line here is that it's water that's uh diverted for this use but it's not that not water that's going to be wasted.

[00:05:54]>> Okay. And and also traditionally data centers treat heat as a kind of waste.

[00:05:59]But as we know this Yaang project it is uh it has this waste heat recovery system where uh this heat is replacing thousands of tons of coal for heating nearby buildings and facilities. Uh but can this model be scaled up to provide lowcarbon heating for homes or industry?

[00:06:19]Certainly uh ch Chinese industries have been good at doing this. Uh I know of a situation a case in the Shandong Peninsula whereby uh heat that's you know generated that that is a byproduct from nuclear power generation and then that heat is treated and then it's it's transported some 100 kilometers away from the reactor that be nuclear and then to power the heating uh for at least 3 million homes. Now that is replaces the the coal that's needed I think by uh the magnitude of millions of tons a year. So this is just what I'm really trying to say is that for Chinese engineers and energy companies to use the uh uh the heat uh efficiently without having to turn to coal. This is just handy.

[00:07:23]>> Okay. And also we are hearing that this center uh could save 320 million kilowatt hours of electricity and cut 280,000 tons of carbon emissions each year. But any idea what baseline are these savings being measured against?

[00:07:40]Well, here in China, this baseline for such measurements in terms of carbon relative carbon emissions or the kilowatts hours of electricity saved, the baseline is often uh coal based uh power generation. And in that case the uh of course you have different uh types of coal and then different the amount of coal would be used differently but whether or not you have a coal fired power plant that's what's called below critical critical or super critical and in China we have more and more of these super critical generators but the bottom line is that it's measurable that you know solar power uh can be a effective and measure measurable replacement of coal fired power.

[00:08:37]>> Yes, but we know that AI is extremely energy intensive. So are these efficiency gains enough to offset the rapid growth in demand?

[00:08:47]>> Well, demanding electricity does not just come from AI. Although you know AI and the energy challenge is often uh highlighted and there are lot of uh extended issues involved and as I mentioned before it's location specific uh if you think about growth in demand for quality electricity it's not just AI it's precision engineering and [clears throat] both of you who are comfortable sitting in Beijing right now, but for our homes, we're not comfortable enough in the sense that um uh we still use LNG, which is or other forms of gas to uh satisfy our daily cooking needs. Imagine a situation when you know district by district here in Beijing we can replace the energy for uh household keeping uh cooking with electricity that demand would be very very high and very very steady as well.

[00:10:01]So let's not what I'm really trying to say that growth in demand however rapid in AI is a fact of life but number one AI itself including the chips that produced may become uh there's certainly a uh a uh business case for encouraging uh less energy intensive AI products the chips beginning with chips and data center that that form the bulk of the data centers. And then the another important point is not to let such concerns about AIdriven energy demand be demonized. After all, AI when it's properly managed and sufficiently regulated does serve a larger economic and social purpose.

[00:10:59]Okay. And I think there's another issue because we know high altitude ecosystems are often quite fragile. So do we fully understand the environmental impact of large data centers in these regions like we are building these projects, these factories, facilities there? Is there a risk that we are shifting environmental pressure from cities to these more remote and more vulnerable areas?

[00:11:26]any human activity has an environmental impact or footprint.

[00:11:35]Uh it it's we're not really choosing having to choose between a rock and a hot place. But comparatively, what other alternatives do you have? Do you uh situate AI centers closer to urban areas that would be you know in terms of water in terms of power a direct competition so I think it's relative and uh the the core point here is let's leave this to the regulators to the uh project designers and encourage them to minimize uh environmental impact also encourage them to invest in environmental protection wherever they see the most optimal or second optimal sites are for these AI centers. [snorts] >> Yeah. And also could large scale computing affect local microclimates even in these cold regions like could it cause some certain level of of warming in these regions?

[00:12:42]Certainly you know it's imaginable as I said before every human action has an impact but uh again it's a everything is relative I I do believe these such concerns should be put in perspective >> okay and we know these centers are primarily processing data for eastern China but to what extent does Xi Jang itself benefit like do they actually contribute to local dig ization and um for instance can AI also help support things like ecosystem monitoring in the region? Certainly certainly actually if anything [clears throat] uh traditionally has been one of the uh pioneering sites for testing uh what today we call AI but in the good old days we call telecommunications or uh ICT information communication technologies in part because of uh precisely because of the concerns about you know reaching to the difficult and fastline localities and in terms of weather in terms of uh the change in temperatures to have these gears uh meet the test of the extreme weather patterns change or the even you know the uh the uh frequent landslides in some part of the country. you know, yeah, those lands slides still you have power cables strong enough to minimize the damage. So, Shan benefits there no doubt about it and it has been a tradition here in China as I said to test out some of the uh equipments in that region.

[00:14:36]>> Yes. And actually this is part of China's eastern data western computing strategy. Can you tell us more about that? Like what is the thinking behind this national plan?

[00:14:47]This the data gathered in the east uh and then computed basically meaning analyzed and processed in the west is precisely a energy uh saving uh strategy.

[00:15:03]The point one point two is data centers are also businesses and by having these data centers registered in Shang or other parts of southwestern China. It's also a way to ensure that local governments have a steady stream of revenue and uh certainly you know this would also help uh increase employment in the local areas.

[00:15:34]>> Is is high altitude or cold climate computing a scalable global solution or is it just something limited to specific geographies? To what extent this can be scalable uh abroad?

[00:15:48]The decision is in the hands of industry and legal regulators abroad.

[00:15:58]uh different societies place different values and of course uh mind you here in China due to our industrial capacity and also our track record of investing in long distance transmission of both electricity and data we can do it but in other societies I would think as I said before number one it's mental regulation and number two it would uh require a associated capacity in building the transmission systems in those countries or localities. [snorts] >> Yeah. And also in climate science we often see that making a process more efficient can uh kind of make it cheaper and that in turn leads to greater overall use of it. But is there a risk that by making AI or these data centers in Shiang more efficient, we could actually accelerate AI demand to the point that it offsets the carbon savings?

[00:17:01]>> Green AI is efficient but then the degree of the green is associated with many issues. the hat source of the power generation and the layout and actual designs of how an AI center runs and frankly speaking the continuous monitoring and upgrade of energy consumption monitoring of the consumption and then improvement in operations.

[00:17:31]um AI can be accelerated in this way because it's cheaper otherwise businesses won't do it then nobody would pay for that but whether or not this can offset carbon savings and if that carbon savings relative to is measured against similar scale of AI deploy deployment that's one way of doing it but if you measure that against a carbon savings from other industries or the the broader socioeconomic context. I don't quite see the connection. But in terms of projectbased assessment, yes, green AI is a network.

[00:18:16]>> Well, I mean there's really a lot of debate right now about um the climate impact of AI and data centers. Just how green can AI really be? like do you think they can become net zero in the future or even climate positive? What would it take to get there?

[00:18:31]>> Net zero and climate positive these remain on the books of climate science in laboratory contexts. Anything can be achieved but uh uh I would think the impact of AI especially the data centers differs from one society to another.

[00:18:52]again uh as I have heard from associating with professionals who work on this the kind of energy intensity that's associated with today's AI data processing the there is room and there are efforts as well to reduce energy intensity I would think as a rule of the thumb the workable or actionable policy choice and industry choice is to move towards net zero and climate positive rather than making net zero or climate positive a precondition.

[00:19:27]>> Okay. But I mean what you believe we can do at the current stage to work towards this goal?

[00:19:34]>> At current stage we should try to adopt the most energy efficient sets of equipments along each and every stage of the way. Point one. Point two is in the laying the cables, construction of the sides and operations, make updated regulations and strictly enforce them and of course all these come with investment and then make timely adjustments to future deployment plans.

[00:20:04]>> Okay, thank you professor Jao Jun of international political economy at the school of international studies at Fing University. Thank you so much for joining us today. And for more on China's Eastern Data Western Computing Strategy, Vivian Nunes spoke with Limbo Chang, Dan of the China Institute for Studies in Energy Policy at Shaman University. As we heard, more and more data centers are being relocated to western China. What more can you tell us about that decision and the context given the environment there and the clean energy?

[00:20:36]>> Well, uh it's because uh one is uh cheap. You are going to build a data center next to a large wind and solar base and that wind and solar in the west can provide very cheap uh energy for the for a data center. The other one is the natural cooling system on average about 10 months uh the the temperature there can maintain about eight degrees Celsius and that is really going to save a lot of money for the data center. So this this is really economydriven uh uh movement uh from to locate the the data center from east to the west. I think most importantly this national strategy we have a strategy called is data computing west and that's been promoted for uh for a few years now and the the important is that if you are national strategy the government normally uh will simplify the approval procedure and can also uh provide other support such as network construction and also uh to connect the data center to the renewables. So all in all I think that this not only economydriven uh movement is also uh supported by the government >> and we hear so much about the global AI race being characterized as a chip war but is this contest really about securing cheap and abundant energy ideally from renewable sources? I think that both are important that chip is important obviously uh you need to have chip to run the the AI center on the other hand that that electricity also very important uh you have chip without electricity you cannot run uh on the other hand that you don't have a good chip you cannot uh run efficiently uh you need to consume more energy to achieve the end results so both are important but I think that what important is that the can you wait the timing is really important in this case for example that that doesn't require one to two year to build most but the electricity supply normal takes longer than that uh you have a really have a huge electricity most effective electricity like China then you can you can use electricity right away but on the other hand if US uh you have the chip you don't have electricity uh and also you cannot wait as I mentioned earlier the data center one to two years but to build a as termite in US but that's what they doing at this moment nuclear is really too far away uh hydro uh coal is pretty far away uh it takes a longer time to build even natural gas take at least two years and can run to seven years based on the some news on the network that say that you need to wait for the manufacturer of the turma equipments so that all in all you cannot wait but I do believe that uh in China the wind and solar Plus any storage can really support a data center in the US because it only take about six to 10 months to build a large scale wind solar plus energy storage but probably plus a small gas turbine they'll be good enough uh we don't need to have a very strong grip to support it and by the way that the the grip in in US is not really strong okay compared to China is pretty far away uh so I think that in this case Electricity chip is important.

[00:24:02]Electricity could be even more important for the US for China. I think next uh we have the electricity system. We have a cheap solar and and also the renewable is can be combined with the can support the data center can also combine with the the low carbon strategy. So everything the in the electricity side is advantage for China. No question about it. and but the China need to support the chip to move forward.

[00:24:29]>> That is Limbo Chang, dean of the China Institute for Studies and Energy Policy at Shaman University speaking with Vivian Nunes. And that wraps up this edition of Climate Watch. You can find more episodes on your favorite podcast platforms. Just search for Climate Watch and hit subscribe. Don't forget to leave us a quick review or share it with a friend. [music] You can also email us at [email protected] with your questions or feedback. [music] And stay updated on the latest climate stories with our Substack newsletter, Climate [music] Choices. I'm Jawing.

[00:25:02]Thank you so much for listening. See you next time.

[00:25:19]>> [music] [music]