Hello, ladies and gentlemen. Thank you for joining GDS Holdings Limited's Second Quarter 2025 Earnings Conference Call. Today's conference call is being recorded. I will now turn the call over to your host, Ms. Laura Chen, Head of Investor Relations for the company. Please go ahead, Laura.

Thank you. Hello, everyone. Welcome to the Second Quarter 2025 Earnings Conference Call of GDS Holdings Limited. As the company's results were issued via newswire services earlier today and are posted online, a summary presentation, which we'll refer to during this conference call, can be viewed and downloaded from our IR website at investors.gdsservices.com. Leading today's call is Mr. William Huang, GDS' Founder, Chairman and CEO, who will provide an overview of our business strategy and performance. Mr. Dan Newman, GDS' CFO, will then review the financial and operating results. Before we continue, please note that today's discussion will contain forward-looking statements made under the safe harbor provisions of the U.S. Private Securities Litigation Reform Act of 1995. Forward-looking statements involve inherent risks and uncertainties. As such, the company's results may be materially different from the views expressed today. Further information regarding these and other risks and uncertainties is included in the company's prospectus as filed with the U.S. SEC. The company does not assume any obligation to update any forward-looking statements, except as required under applicable law. Please also note that GDS' earnings press release and its conference call include discussions of unaudited GAAP financial information as well as unaudited non-GAAP financial measures. GDS' press release contains a reconciliation of the unaudited non-GAAP measures to the unaudited most directly comparable GAAP measures. I'll now turn the call over to GDS' Founder, Chairman and CEO, William Huang. Please go ahead, William.

Thank you. Hello, everyone. This is William. Thank you for joining us on today's call. We delivered a solid second quarter, growing revenue by 12.4% and adjusted EBITDA by 11.2% year-on-year. We raised net proceeds of USD 676 million through the issue of convertible bonds and equity in the international capital market, strengthening our holdco balance sheet. More recently, we achieved a significant milestone in our onshore asset monetization strategy with the successful completion of our C-REIT IPO, the units of our series are now trading on the Shanghai Stock Exchange at implied cap rates of below 5%. This is a major breakthrough, giving us access to the China equity capital market on highly advantageous terms. Our gross move-in during 2Q '25 was around 20,000 square meters, which is consistent with the level over the past 5 quarters. Our utilization rate has continued to climb, reaching 77.5%. Moving over the next few quarters will remain solid, driven by delivery of the 152 megawatts order, which we signed in 1Q '25. We expect to deliver 35% of our total current backlog in the second half of 2025. In Q2 '25, gross new bookings were 23,000 square meters, mainly from traditional Internet and the cloud business with a good mix of customers and locations. AI demand was relatively quiet due to the uncertainty of chip supply in China. Customers have a number of options across both imported and domestic sources for the chip. It's a complicated mix of performance, technology, availability, and other considerations. We think that it will take some time for customers to decide which way to go. We are very confident about AI-driven demand over the medium and long term. However, we are still in a period of wait and see. During this period, we think that the most important thing is for us to be prepared to respond, ready in terms of developable capacity and ready in terms of access to capital. On the capacity side, we have around 900 megawatts of powered land held for future development in and around Tier 1 markets. We believe the coming waves of AI demand will be mainly for inferencing. This kind of demand is latency-sensitive and will require relatively large sites distributed across Tier 1 markets. For operational reasons, customers will seek to deploy capacity for inferencing within established cloud environments. We have multiple sites suitable for AI inferencing around Beijing, Shanghai, and Shenzhen. We have undertaken preliminary site preparations so that we can develop with a short lead time. This is an important consideration for customers. We believe there is a good chance that we will develop all of this 900 megawatts and more over the next few years. The issue is only one of timing. On the financing side, we completed the first-ever data center ABS transaction in China in late March. We then followed this up with the first-ever data center REIT's IPO in China in August. By pioneering these transactions, we have proven our ability to recycle capital from stabilized data center assets. This comes at the perfect time as we can use the proceeds to fund new investment opportunities. Furthermore, the terms on which we have monetized assets established a benchmark for the value of our stabilized data centers in Tier 1 markets, creating potential to unlock more value for shareholders. Our powered land and our monetization vehicles are unique in China and give us a significant competitive advantage as we enter the AI era. Lastly, I would like to share some operation updates for our equity investment in DayOne. In 2Q '25, DayOne added a phenomenal 246 megawatts of new commitments, which brings its total power committed by customers to over 780 megawatts. The new order in 2Q '25 included an anchor customer commitment for its Thailand project. More recently, DayOne announced that it has secured a second campus site in Finland, building out its successful market entry. DayOne is well ahead of schedule to meet the target of 1 gigawatt of total power commitments within 3 years. I will now pass on to Dan for the financial and operating review.

Thank you, William. Starting on Slide 13. In 2Q '25, revenue increased by 12.4% year-on-year. This resulted from an increase in total area utilized of 14.1% and a decrease in MSR per square meter of 1.7% as compared with 2Q '24. In 2Q '25, adjusted EBITDA increased by 11.2% year-on-year. Adjusted EBITDA margin for 2Q '25 was 47.3% compared with 47.8% in 2Q '24. Following the completion of the ABS transaction in late March, we deconsolidated the underlying projects for the whole of 2Q '25. Following completion of the sale of stabilized data centers to a C-REIT in late July, we will deconsolidate these projects during 3Q '25. As we report earnings over the next 3 to 4 quarters, the reported revenue and EBITDA growth will be impacted because the comparison will not be apples to apples. We estimate that the apparent year-on-year growth rate, without making adjustments to normalize for the asset monetizations, will be about 6 percentage points lower. We will continue to point this out on future earnings calls to ensure the underlying trend is clear. Starting with 2Q '25, without the ABS transaction, the year-on-year adjusted EBITDA growth rate would have been 13.9% as compared with the reported 11.2%. As shown on Slide 17, the ABS transaction took place on an EV to EBITDA multiple of 13.3x based on the maximum potential sale proceeds and the projected stabilized EBITDA. This was a good start considering where GDS is trading as a listed company on NASDAQ and the Hong Kong Stock Exchange. However, for the C-REIT IPO, we achieved an even higher multiple of 16.9x at the IPO price of RMB 3 per unit. The units started trading on the Shanghai Stock Exchange on the eighth of August. The closing price yesterday was RMB 4.04 per unit, about 35% up from the IPO price. At this level, the C-REIT is trading at 22.8x the projected '26 EBITDA disclosed in the offering memorandum. This is close to double the current year trading multiple for the GDS China business after adjusting for the assumed value of our equity investment in DayOne on a sum-of-the-parts basis. Under the current C-REIT regulations, we must wait 12 months before undertaking the first post-IPO asset injection. We started preparing some candidate assets of various sizes to give us the flexibility to dimension the next monetization in accordance with our financial requirements. It's important that we continue to grow and diversify the C-REIT so that it remains a viable option for us to recycle capital when it's in our interest to do so. With the C-REIT platform in place, if we assume that we invest in new projects, ramp up, operate, and monetize after 5 years at a cap rate in, say, the 5% to 6% range, the return on investment is at a very acceptable level. Turning to Slide 18. When we gave CapEx guidance earlier this year, we spoke of RMB 4.8 billion of organic CapEx. This RMB 500 million net proceeds in the current year from the ABS transaction resulted in CapEx guidance of RMB 4.3 billion. We are now deducting a further RMB 1.6 billion net proceeds from the C-REIT transaction, which was not previously factored in. This brings our CapEx guidance down from RMB 4.3 billion to RMB 2.7 billion. On Slide 19, in 2024, we achieved positive cash flow before financing, with the benefit of some capital recycling from DayOne back to GDS. In 2025, despite the fact that our organic CapEx is much higher than in the past few years, we expect our cash flow before financing to be close to breakeven with the contribution from our asset monetization transactions. Turning to Slide 20. During the second quarter, we raised USD 535 million through the issue of a 7-year convertible bond with a 2.25% coupon and a 35% conversion premium. We also raised USD 142 million through a simultaneous follow-on equity offering. One of the main purposes of this capital raise was to enable us to repay short-term debt at the holdco level and to either repurchase if possible or potentially redeem a convertible bond issued in 2022, which is currently out of the money and due in March 2027. Now net debt to LQA adjusted EBITDA decreased from 6.6x at the end of 1Q '25 to 6.1x at the end of 2Q '25. The reduction in consecutive quarters was partly due to the cash proceeds of the ABS, which were received during 2Q '25 and to the cash proceeds of the follow-on equity offering. As shown on Slide 21, if we take account of the C-REIT transaction on a pro forma basis, the net debt to LQA adjusted EBITDA ratio will come down to 5.9x. If we further adjust for the value of our reinvestment in the ABF and C-REIT listed securities, the ratio will come down to 5.7x. On Slide 22, we have already used part of the proceeds of the offshore capital raise to repay working capital loans due in 2026. As you can see, we now have 3 convertible bonds outstanding. The 2022 bond is out of the money. Hence, we show the maturity based on the potential put in 2027. The liability is covered by cash, which we are holding in reserve. The 2023 bond and the recently issued 2025 bond are both in the money. Hence, the maturity is shown based on the final maturity dates in 2030 and 2032, respectively. Turning to Slide 23, where we gave guidance earlier this year, we already assumed that the ABS will be deconsolidated in 2Q '25. However, the C-REIT transaction, which we completed in late July, was not factored into our 2025 guidance at all. Nonetheless, we are maintaining FY '25 revenue and adjusted EBITDA guidance unchanged, notwithstanding the deconsolidation of the C-REIT assets, while we are making mathematical adjustments to CapEx guidance to deduct the C-REIT cash proceeds. Finishing on Slide 24. DayOne's power utilization jumped from 143 megawatts at the end of the first quarter to 213 megawatts at the end of 2Q '25. This contributed to revenue growth of 244% and adjusted EBITDA growth of 265% year-on-year during the second quarter. Considering its fast expansion, including the recently announced second campus in Finland, DayOne is currently working on a Series C equity raise.

And now we're going to take our first question. It comes from Yang Liu from Morgan Stanley.

Congrats on the very solid results. I would like to ask about the future strategy regarding asset monetization in China. After the successful series IPO, does management benchmark the previous set target of 5x net debt-to-EBITDA as a long-term operation target for GDS leverage? Or are you more keen to adopt a more aggressive asset-light model to achieve better investment returns via the 5-year development cycle? How do you view your future strategy here?

Thank you for the question. There are several different considerations in the asset monetization strategy. One, of course, is the value at which we can monetize assets. The benchmark, which has been established in the ABS transaction and then at a high level in the C-REIT transaction, remains far above the level at which GDS shares are trading in the international capital markets. The implication of that is that every asset monetization is highly accretive to our shareholders. And I think that alone provides a strong rationale to monetize assets. Secondly, as we described in our prepared remarks, we feel like we are on the threshold of the start of another multiyear growth phase in this industry, which should present very good investment opportunities. The return on investment potentially is enhanced now that we know we can monetize assets at cap rates that are certainly higher than what we used to assume in our internal underwriting. The implication is that if we can monetize assets and reinvest, then we can create more value for our shareholders. You mentioned the consolidated net debt to EBITDA ratio; I did check back, and I think in 2023, I mentioned that we would target 5x within 3 years, which would give me about another year, I think. I believe we're approaching that level already, but we are now, as I mentioned, at a stage where some attractive new investment opportunities could present themselves. I don't think it's necessary for us to be overly aggressive about deleveraging if those opportunities arise. If they don't arise, then if we monetize on accretive terms, the deleveraging will naturally happen.

I have another question, if I may, regarding the development of DayOne. Given the company's belief that the previous 1 gigawatt target will be achieved far ahead of schedule, what is the current new target, for example, by the end of this year or next year in terms of the total area committed or megawatt committed?

Yes. I think based on our current footprint, we are building each growth engine in different regions right now. Finland is a very good example in Europe, and in Asia Pacific, we have already built a very solid and sustainable growth resource, land bank, and power, right? So our growth will be very solid in the next few years. Generally, we target every year at least 500 megawatts, yes. This is some internal KPI, but we commit to the market at least 300, right? Internally, that's ours. But now we have a very solid base to talk about this kind of number because we are not just growing in one country or one region. We have two regions, and in the next couple of quarters, we may enter some new regions as well. This will allow us to discuss more significant numbers and higher growth. Thank you very much.

Now we will take our next question and the question comes from the line of Sara Wang from UBS.

Again, congratulations on the solid results. I have one question regarding the customer profile. Given that the second quarter's gross move-in or new orders signed are still quite solid despite all the uncertainties around U.S. GPU exports, may I ask who the key customers are, separately for the move-ins and also for the new orders? What kind of workload do we expect for these new orders to carry? Is it mostly CPU or GPU? And if it's CPU, is it because the oversupply in the industry has been digested? Or if it's GPU, does that mean domestic substitution has achieved quite meaningful progress so that the supply chain uncertainty going forward should be mitigated?

Yes. I think for the first question about customer profile, as I mentioned, there are traditional Internet companies plus cloud service providers. This is in relation to some new orders, which we received this year. In terms of workload, there are both GPU types and traditional CPU cloud growth as well. So I would say this is quite hybrid.

I see. Maybe a quick follow-up on the demand side. Do we see any signs of price increase or MSR increase in the industry? The reason I'm asking is that I saw in the second quarter the MSR decline continued to narrow year-on-year and even increased quarter-on-quarter. However, if we assume the contract length is maybe 5 years on average, that means the contracts renewed this year were mostly signed 5 years ago, when industry rental prices peaked in 2020 or '21. As we renew the contracts this year, we still maintain a stable MSR. So what's the key reason behind?

First of all, let's talk about the market price. It's been stable since, say, the middle of last year, which is quite satisfactory. I mentioned in my prepared remarks if we evaluate new investment using a 5-year cycle from inception to exit, we can generate a very acceptable return, even if we consider exit cap rates that are considerably lower than where our ABS and C-REIT transactions were done. I think that's important because many industries in China are suffering in a deflationary environment. So the economics of our business remain very solid. The MSRs, I know Sara, because you asked about this previously in earnings calls. The MSR reduction is partly a reflection of the market price reduction you're referring to, but it also comes from changes in the mix. You go back 5 years, and most of our new business was coming from edge of town sites around Beijing and Changshu near Shanghai. Those were relatively early years for that kind of large edge of town campus, and there was a significant price and development cost differential compared to our downtown co-location data centers. So the MSR is not purely an indication of market price reduction; it also reflects the change in location mix. Over the next couple of years, we will continue to see our MSR decline, mostly due to the price reset of contracts like you've mentioned. We know pretty well what the dilution should be from price reset over the next couple of years, and that will get reflected in our MSR. The MSR will decline by a few percent when compared with the same quarter the prior year, but it should continue to decline by low single digits for the next couple of years. Beyond that, I believe we will start seeing a reduction in drag, and growth rates will reflect the volume growth in our business.

Now we're going to take our next question. The question comes from the line of Frank Louthan from Raymond James & Associates.

With the Series C round that you're looking at, are you still considering a broader public offering for DayOne in 2026, as I think you've talked about in the past? Can you break out some of the growth in DayOne between Southeast Asia and then Finland or any other EU sites that you're considering?

Yes, Frank. It's the shareholder plan to have an IPO of DayOne from a GDS perspective, particularly because we would like to have the opportunity to distribute the shares to our shareholders. I think it remains the case that we are targeting to have an IPO within 18 months. The Series C that we're considering did not anticipate that there would be another equity round pre-IPO. The performance of DayOne has far exceeded our expectations. It's been phenomenal, and that is what is driving the Series C. I can't rule out there being other capital raises before the IPO, and DayOne has plans to do some mezzanine debt as well. Those pre-IPO rounds are a result of the success the business is having. Regarding your second question about Europe, our European presence is currently in Finland in the Helsinki area. We have a first campus for which we obtained an anchor customer commitment. I won't provide precise details, but it's well over a 100-megawatt commitment, and I expect that we will build on that quickly in terms of follow-on commitments. The strategy of DayOne is to be a pioneer in creating new markets. It's not easy to achieve, but DayOne has done it multiple times now, working closely with different customers to de-risk our market entry. This provides us the opportunity to build on that base. We believe Finland is a very attractive location for data center operations because of access to renewable energy, competitive power tariffs, and a supportive operating environment. What you're seeing now is just the foundation and de-risk market entry, secured resource expansion, and the opportunity to add significantly to that.

Now we're going to take our next question. The question comes from the line of Edison Lee from Jefferies.

My first question is on DayOne. You have roughly 780 megawatts committed power. Could you give us some insight as to the split between Chinese customers and U.S. customers of that 783 megawatts? I believe your longer-term objective previously mentioned was a 50-50 split, right? I just want to know the progress on that.

Frankly speaking, the current percentage-wise, not significantly improved, but it's early stage. We encountered this situation over the last couple of quarters because every time we write to our key customers about their demand, we're trying to build our business. So I believe this will change in the next 2 or 3 years. The whole profile should change. Currently, I would say it's about 30% international customers and 70% domestic customers.

I see. Okay. A quick follow-up here on your guidance. You haven't changed your revenue and EBITDA guidance this year, and the first half is very strong, right? So I'm wondering how we should think about the second half growth, given that guidance has not been adjusted.

In the second half, we have the impact of deconsolidation of the C-REIT, but that wasn't factored into our original guidance at all. We will be deconsolidating the revenue and EBITDA from late July, which will have a material impact in terms of the EBITDA for the 5 months post-deconsolidation. Yes, I'm aware that the implied growth rate for the second half is lower than the implied growth rate for the first half, but we didn't feel it was necessary to adjust our revenue and adjusted EBITDA guidance at this point.

Does this mean it’s going to be impacted more by the move-ins in the second half and also potentially higher depreciation as you deliver more projects into the second half?

Well, I think let's see what the growth rates actually are as there are many moving parts.

Now we're going to take our last question for today, and it comes from the line of Gokul Hariharan from JPMorgan.

First of all, it looks like you have a fairly back-end loaded delivery schedule this year. Dan, could you outline how that will influence growth probably into next year, given that most of this is likely not captured this year or even early next year? Should we expect there to be a reacceleration in revenue and EBITDA growth, maybe in Q2 or Q3 next year, given you're delivering a lot of capacity in the second half of this year? That's my first question.

We are delivering a lot of capacity in the third and fourth quarters this year. However, the incremental revenue per square meter for that capacity is below our MSR. This edge of town capacity consists of a couple of large sites that are being developed specifically for AI inferencing with a very high power density. The impact of that may not be as great as it would be if this were a more traditional cloud business. I think we stick by our high-level direction of targeting high-single-digit EBITDA growth year-on-year. We have a backlog that will drive some of that growth, but we also need new bookings to fuel that growth. Currently, I think the new bookings are at a healthy level, higher than in the last few years, but are not reflecting mega orders like we saw in Q1 of this year. Until all that happens, I believe our growth rate won't really accelerate.

Understood. How are the conversations going with customers regarding some of these AI orders considering the supply situation? It seems to be a little more optimistic in Q3 compared to what it was in Q2. Are you seeing a lot more interest from customers? Is the sticking point still the certainty of chip availability, or are there other factors at play, such as the inferencing demand coming a little later in this AI cycle compared to the remote site training demand that has already occurred over the last couple of years?

We saw in the first quarter that if there were no chip supply issues, we would see much stronger bookings. That gives us confidence about the future. The key issue really is chip supply, and that's not resolved quickly. There have been policy changes, and I think right now, customers are waiting for the new technology in terms of the next generation of NVIDIA chips. It may not just be all about current models; it may involve the anticipation of the next big thing. I think we will have a clearer view on that in the next couple of months, and then we can talk more precisely about the timing of when we'll start to see those large orders. In the meantime, all we can do is prepare, and I think we're very well prepared. We've secured the powered land and incurred some CapEx to ensure that we can shorten the lead time. From our previous experience in China, as well as what we've observed at DayOne, we know that when customers are deploying AI, they usually have urgent demands. So we believe we are very well prepared in terms of developable land and have adequate access to capital, both the capital we've raised and our confidence in being able to recycle more. I don't think any other data center companies in China are as well prepared in both respects.

Dear participants, thank you very much for your time. Due to the time limit of today's call, we will not be addressing any further questions. At this moment, I would like to turn the call back over to the company for any closing remarks.

Thank you very much once again for joining us today, and we look forward to seeing you next time.

Thank you.

This concludes today's conference call. Thank you for participating. You may now all disconnect. Have a nice day.