Sara Wang: Got it. Thank you. And then my second question is about William. So during the last earnings call, I think William has shared his intention to increase the shareholding in GDS. I’m just wondering if there’s any update on this. Thank you.
William Huang: Yes. I already bought the 50%, which I committed, right? So I still will continue to execute that.
Sara Wang: Got it. Thank you.
Operator: We are now going to proceed with our next question. And the questions come from the line of Timothy Zhao from Goldman Sachs. Please ask your question.
Timothy Zhao: Sure. Thank you, management, for taking my question. So my question is regarding the AI demand. I think since the last time that we spoke, I think I remember last quarter, you mentioned that AI driven demand was still in the early stage. Just wondering, given I think we have passed a quarter, three months or maybe longer. Just wondering if management have seen any updates regarding the move-in pace from the generative AI demand from the [indiscernible] in China. And when we think about the CapEx going forward, could management share any color on like what percentage of CapEx will be likely spent on the like high-power density cabinets or transforming the existing cabinets into the power – high power, density wise.
William Huang: Okay. Yes. In terms of AI demand, I think we are very clear in the international market, now the currently globally, I mean, data center demand mainly driven by the AI type of application, right? So this is a very, very clear chance which already happening in the U.S., in Asia right now. So we are – this is one part. But if we talk about AI demand in China, I think the China, in terms of the model – AI model, it’s a little bit behind the U.S. So I think it’s still in a very early stage. But of course, in the media, everybody talk of the AI. That’s just an indication that everybody try to step in. So I think in terms of how impacted to the data center industry, I think they maybe up to 12 months or 18 months, it will start to impact China data center demand. So in a significant way in my view, yes.
Dan Newman: For the second part of your question on CapEx, right, I think if you talk about AI-related CapEx or very high power density capacity, then I think that goes with liquid cooling. So really, it comes down to what percentage of our CapEx or what percentage of the capacity that we’re developing will we deploy liquid cooling. As a matter of fact, we have deployed liquid cooling going back more than two years in China. We’ve done projects both with what’s called cold plate cooling and also full immersion cooling. It’s been a relatively small part of what we’ve done so far in China. And for – in the international expansion, we did mentioned that, I think in the international market, maybe we’re seeing the flow-through from AI demand come quicker.
So, a significant part of what we developed in Johor is using cold plate, liquid cooling, and we developed a prefabricated cold plate liquid cooling module, which we manufactured in China and shipped to Johor. We’ve talked maybe a bit more generally about the economics and deploying liquid cooling. The overall unit development cost is slightly higher than if we use traditional – more traditional air cooling. But because liquid cooling delivers a lower POE and a higher power density, it also means that there’s more IT power capacity available to sell to the customer and to generate revenue. So you have to take into account these different components. And overall, in terms of total cost of ownership or economic returns to us as the data center operator, we think that liquid cooling will create some cost efficiency, where it can be deployed.
