Hello, ladies and gentlemen. Thank you for standing by for the GDS Holdings Limited's Fourth Quarter and Full Year 2024 Earnings Conference Call. At this time, all participants are in a listen-only mode. After management's prepared remarks, there will be a question-and-answer session. 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 fourth quarter and full year 2024 earnings conference call of GDS Holdings Limited. The company's results were issued via newswire services 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. 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 US 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 US 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 release and this 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, Laura. Hello, everyone. This is William. Thank you for joining us on today's call. The race is on for AI in China. We saw the beginnings of it last year when cloud and Internet companies increased their capital expenditures. This led to an initial wave of demand for AI training in remote locations. Now the race has gone to another level with demand for AI inferencing in Tier 1 markets. Based on our dialogue with our customers, this type of demand could run into multiples of gigawatts over the next few years. Looking at the opportunity from GDS' perspective, it is an exciting time to be a data center company again. The opportunity in Tier 1 markets plays to our strengths. We are, by far, the best positioned in terms of land and power to fulfill this kind of demand. The largest cloud and Internet companies in China are all our largest customers. A key fact affecting the timing of customer deployments is the availability of chips. For deployments over the next few quarters, we do not see any significant risk, and we are willing to commit to new business. However, for deployments further into the future, we think the right approach for us is to wait and see. The demand-supply situations in Tier 1 markets continue to improve, and we have the flexibility to decide when to move forward. We just executed our first asset monetization transaction. From a financial perspective, this enabled us to address immediate opportunities without deviating from our current path and the strict discipline. As our asset monetization program becomes fully established, we will have flexibility to do more while delivering on our commitments to shareholders. Several years ago, we laid out a strategy to get GDS back on track with steady growth and a stronger financial position. We remain firmly committed to this strategy. We focus on Tier 1 markets where we can add the most value. We prioritize delivering the backlog. We remain highly selective about new business, pursuing orders which match our inventory and which have fast move-in schedules. We incur capital expenditures when needed with short lead times ahead of our customer move-ins. We recycle capital through asset monetization, which is repeatable and scalable. And we create additional value through our equity stake in DayOne, which is now a standalone business. Let's review our progress in implementing this strategy. Our gross move-ins during 2024 were 79,000 square meters, all organic and all in Tier 1 markets. This is the highest in our history. The move-in rate picked up in the first quarter of 2024 and has stayed at a consistently high level into the current year. The pickup was due to a combination of backlog delivery and new orders with fast move-ins. As shown on Slide 7, we started 2025 with 110,000 square meters of backlog for area in service. We expect to deliver over half of this during the current year. We ended 2024 with a utilization rate of 74%. We expect utilization to increase to the high 70s percentage by the end of 2025. Our gross additional area committed during 2024 was 49,000 square meters, similar to the past two years, in-line with our strategy. We targeted new business to absorb our inventory. A good illustration is the three new orders we won in the fourth quarter of 2024, all related to capacity in service or under construction. During the first quarter of 2025, we won a massive new order with an existing hyperscale customer for around 40,000 square meters or 152 megawatts, split across two sites in Lanfang and Changshu. It is the largest single order in our history in China. This new order requires us to deliver data center within six months. The customer committed to move in fully within the following six months. The whole cycle from obtaining the new order to full utilization is about one year. This is high-quality AI-driven new business with no chip supply risk. It fully satisfies all of our criteria for capital expenditures with a short lead time, fast move-in and a long contract tenor. Furthermore, the sites are existing campuses where we already invested in past years. As a result, we only need to incur the cost to complete, and we are able to meet the deadline for rapid delivery. For AI inferencing in Tier 1 markets, hyperscale customers typically require sites with at least 15 megawatts of available capacity deliverable within a short period of time. Fortunately, we are very well positioned in this regard. We have multiple sites suitable for AI inferencing around Beijing, Shanghai, Shenzhen, and Guangzhou. After completing the 152 megawatts new order, we will still have around 900 megawatts of developable capacity. As demand continues to grow, there are still a few sites in Tier 1 markets with the necessary scale and the time to market. This should benefit us. Turning to Slide 13. I would like to share some operating updates for DayOne, which became our equity investee upon closing of its Series B equity raise. In 2024, DayOne accomplished a historical 340 megawatts of new commitments. DayOne ended 2024 with 467 megawatts of total IT power committed, most of which will be billable within the next two years. DayOne's sales pipeline is highly visible and strong. DayOne is confident of doing over 250 megawatts of new commitments during 2025 and remains on track to hit 1 gigawatt of total IT power committed in less than three years. I will now pass on to Dan for the financial and operating review.

Thank you, William. DayOne Data Centers, previously known as GDS International or GDSI, completed and closed its Series B equity raise on December 31, 2024. At closing, GDS' equity interest in DayOne was diluted from 52.7% to 35.6%. Accordingly, GDS deconsolidated DayOne as a subsidiary and recognized DayOne as an equity investee. In the consolidated financial statements for the quarter and year ended December 31, 2024, DayOne's operational results and cash flows have been excluded from the company's financial results from continuing operations and have been separately itemized under discontinued operations. Retrospective adjustments to the historical statements of operations and cash flows have also been made to provide a consistent basis of comparison for the financial results. Furthermore, retrospective adjustments were also made to categorize DayOne's assets and liabilities as 'assets and liabilities of discontinued operations' on balance sheets for the comparative periods. From the first quarter of 2025 onwards, DayOne will appear in our financials as a single line in our income statement and a single line in our balance sheet. However, in our earnings presentations going forward, we intend to continue disclosing key financial and operating information for DayOne similar to what we disclosed when DayOne was a segment of GDS, so that investors can keep track of DayOne's performance and the value of our equity investment. Although we will no longer present GDS and DayOne on a consolidated basis, we did provide guidance on a consolidated basis for 2024. I would highlight that our pro forma consolidated adjusted EBITDA for 2024 was above the top end of our guidance range. From now on, I'm talking about GDS continuing operations. Starting on Slide 17. In the fourth quarter of 2024, revenue increased by 9.1% and adjusted EBITDA increased by 13.9% year-on-year. In 2024, revenue increased by 5.5% and adjusted EBITDA increased by 3% year-on-year. If we normalize the numbers by excluding one-time items in 2023 and reversing the BOT projects transfer in 2024, our revenue and adjusted EBITDA would have grown by 7.9% and 7.7%, respectively. MSR per square meter declined by 2.3% in the fourth quarter of 2024 compared with the fourth quarter of 2023, in-line with our expectations. Looking forward, we expect MSR to decline slightly over the next year and we assume that power tariffs remain at current levels. Adjusted EBITDA margin for 2024 was 47.2% compared with 48.4% in 2023 or compared with 47.8% in 2023 excluding the one-time items. This implies that on a normalized basis, EBITDA margins were flat. For 2024, our capital expenditures totaled RMB3 billion, in-line with our revised guidance. Our base case capital expenditures for 2025 was RMB2.5 billion. However, we will incur an additional RMB2.3 billion as the cost to complete and deliver the 152 megawatt new order. Offsetting this increase, we expect to receive RMB500 million first installment of cash proceeds from the ABS transaction. In sum, we are giving guidance for around RMB4.3 billion of capital expenditures in 2025. Please note that this does not take account of the balance of proceeds from the ABS, further mega new orders or the proceeds of further asset monetization transactions in the current year. For the full year of 2024, our cash flow before financing is positive RMB379 million. Once again, this is in-line with our financial target. In 2025, with additional capital expenditures for the 152 megawatt new order, cash flow before financing will be negative. However, if we factor in debt deconsolidation and the deferred cash proceeds from the ABS transaction, we would still see no increase in our net debt. I'll come back to this point in a minute. As shown on Slide 24, at year-end 2024, the cash balance was RMB7.9 billion and the net debt for last quarter annualized adjusted EBITDA multiple was 6.8 times. Turning to Slide 26, we recently announced our first asset monetization transaction. This involves selling 100% of the equity of certain data center project companies to an SPV managed by a major Chinese securities company with back-to-back issuance of ABS. For the avoidance of doubt, the ABS represents the equity of these projects and it is not a liability of GDS. The ABS is 70% subscribed by top-tier institutional investors in China, led by China Life, while GDS subscribes for the remaining 30% and retains the rights for ongoing operation of the underlying data centers. The ABS will be listed on the Shanghai Stock Exchange as a standardized security product. The total enterprise value, or EV, for the transaction is up to approximately RMB2.9 billion, implying an EV to EBITDA of around 13 times. The total equity consideration is up to approximately RMB1.7 billion or RMB1.2 billion net of the 30% reinvestment by GDS in the ABS. The upfront cash proceeds are around RMB500 million and the deferred net cash proceeds are around RMB700 million. The reason why there are deferred proceeds is because the underlying data centers are still ramping up. Upon closing, we will deconsolidate existing debt of around RMB1.2 billion. We are making good progress with our public REIT or C-REIT application. It is moving forward faster than expected. C-REITs are not permitted to invest in the equity of unlisted companies; however, they can invest through ABS. As shown on Slide 28, with the ABS transaction expected to close in the next couple of months, we can cover our 2025 capital expenditures at RMB4.3 billion without increasing our net debt. We expect our net debt to last quarter annualized adjusted EBITDA multiple to come down to just over 6 times at the end of the current year. With the recovery in our share price, our 2030 convertible bond is now deeply in the money. If we treat this convertible bond as converted, our year-end net debt to last quarter annualized adjusted EBITDA multiple will be around 5.5 times. Turning to Slide 29. For the full year of 2025, we expect our total revenues to be between RMB11.29 billion to RMB11.59 billion, implying a year-on-year increase of between approximately 9.4% to 12.3%, and adjusted EBITDA to be between RMB5.19 billion and RMB5.39 billion, implying a year-on-year increase of between approximately 6.4% to 10.5%. In addition, as I already mentioned, we expect capital expenditures to be around RMB4.3 billion. On Slide 30, we look at our guidance a few different ways. Our official guidance takes into account the deconsolidation of the data center projects underlying the ABS. On a normalized basis, if we assume the ABS did not happen, our adjusted EBITDA growth for 2025 at the midpoint would have been around 10.7%. This is consistent with the objective we set of getting back to double-digit growth. Alternatively, if we take our official guidance and then add on the gain on the sale of the data center projects, the adjusted EBITDA growth for 2025 at the midpoint is around 16.7%. Lastly, the additional capital expenditures, which we will incur for the 152 megawatt new order, in 2025 will lead to higher growth in 2026. Our current and very preliminary view is that adjusted EBITDA growth could be in the low teens for 2026, before taking out a further mega new orders or asset monetization. Finishing on Slide 31, we're not providing guidance for DayOne. However, we note that DayOne ended 2024 with a run rate adjusted EBITDA of around US$60 million. Based on the expected ramp-up, we’ll see an increase by multiples over the next two years. We'd now like to open the call to questions.

Thank you. We will take our first question. Your first question comes from the line of Yang Liu from Morgan Stanley. Please go ahead. Your line is open.

Thank you for the opportunity to ask a question. I would like to have some visibility in terms of your plan to spin off DayOne and let it go public. Could management update us in terms of the current plan and schedule? Yes, that is my question. Thank you.

Thank you, Yang. I think, last quarter, some investors asked the same question, but we don't have a clear view, right? Now, I would like to say, we do have the plan. The IPO plan is more visible, and we plan to list as a company within 18 months. So, I think, this is achievable, and we are very confident based on the current international business, DayOne's business grows so fast, and we are very confident it will be a very successful IPO in the next 18 months and create more high value for our current shareholders.

Thank you. May I follow up in terms of the C-REIT's progress that Dan just mentioned? You see faster-than-expected growth progress here. What is the status now? Is this under NDRC or under CSRC or stock exchange? And should we expect it to come out in the next one, two, or three quarters? What's your expectation now? Thank you.

I think we've made significant progress, but we can't disclose that information just yet. Once we are able to share, we will provide an update immediately. Last quarter, we aimed for the end of this year regarding the C-REIT's progress, but it looks like it may be four to six months ahead of our expectations.

Thank you. We will take our next question. Your next question comes from the line of Sara Wang from UBS. Please go ahead. Your line is open.

Thank you for the opportunity to ask questions. I have two questions, mainly on the China business. So, first of all, may I ask whether the current capital expenditures are based on existing orders on hand as management just mentioned that includes the more than 150 megawatts order win in the first quarter? And how shall we think about new order wins throughout 2025? Second question is regarding the existing vacant capacities. William just mentioned the AI inferencing demand from hyperscalers now requires more than 50 megawatts project size, and whether the existing capacity utilization ramp-up is mainly driven by non-AI demand? Thank you.

The first question is about our recent deal win in the first quarter, which we announced. We have noticed a strong pipeline, but we are closely monitoring the chip supply situation, as it is crucial for AI deployment in data centers in China. There is some uncertainty around the chip supply, and we are being cautious in our approach. The demand from hyperscalers is robust, as indicated by their capital expenditure guidance, and we believe this demand will persist over the next three to five years. While we are confident about the demand this year and in the coming years, we are taking a patient approach due to potential supply uncertainties with chips. We are ready to act on orders as needed, but our strategy is currently to be selective with new orders. The second question addresses the shift in the AI landscape. Initially, investments were focused on AI training, but DeepSeek has accelerated the demand for AI inference in China sooner than expected. This shift aligns well with our capabilities. The requirements for inference differ significantly from those for training; inference needs to stay close to the traditional cloud to support enterprise applications and requires very low latency. This aligns well with our resources, positioning us positively for this next wave of demand for inference in our data centers.

Very clear. Thank you.

Thank you. We will take our next question. Your next question comes from the line of Frank Louthan from Raymond James & Associates. Please go ahead. Your line is open.

Thank you. Can you describe the types of customers and workloads you are seeing? What proportion of that is related to AI compared to more traditional cloud enterprise business in China right now? Additionally, what is the current book to bill rate? In other words, how long does it take from when you sign a contract to when you start fully billing at the agreed terms? Historically, this process took quite a while. What is that duration like today? Thanks.

Currently, the workload in the Tier 1 market is primarily driven by AI inferencing rather than training. The training surge we’ve seen occurred over the last two years, and it’s not part of our current strategy. We are concentrating our resources in the Tier 1 market, aligning with our business strategy. Demand in this market is mainly fueled by AI inferencing, which is also accelerating traditional cloud deployment more rapidly than before. Regarding our capacity, we are utilizing our current capabilities and the CapEx capacity of the data centers under construction to meet customer demand. The lead time from when we secure an order to when we are fully operational is now around 12 months, which is a significant improvement compared to previous years when it often took two years or longer.

Are those lead times contractually obligated? Or is that just how quickly the customers want to move?

Yeah, absolutely. And as I mentioned, the deals which we select, the contract length is much longer than before. Based on our current position, we are sitting in a very good position to negotiate new terms compared with the last couple of years where we weren't.

Thank you. We will take our next question. Your next question comes from the line of Timothy Zhao from Goldman Sachs. Please go ahead. Your line is open.

Great. Thank you, management, for taking my question. I think the first question really is regarding the supply and demand dynamics that you see in the Tier 1 markets. As you mentioned that I think by the end of this year, the utilization rate of GDS is, I think, approaching like the high 70s. Just wondering if you have any sense on the industry-wide utilization rate? And also, how do you think about the pricing environment in the Tier 1 cities? And secondly, is regarding DayOne. I think you mentioned that you foresee around 250 megawatts new commitments for DayOne this year. Just wondering if you can provide some color on the orders or the demand? And what types of customers are you seeing that are contributing to this new commitment? And what are the underlying demands like AI versus non-AI? And if there's any risk regarding chip availability in this region? Thank you.

The primary concern in China is that the Tier 1 market is just beginning. Other major AI companies have provided their official guidance starting this year. In recent years, growth has been largely driven by training, but this year's guidance indicates that demand will shift from training to inferencing over the next three years. This transition is starting to benefit us. However, there is still an imbalance in supply and demand in the Tier 1 market, which is just beginning to take shape. Personally, I believe that this situation will stabilize in six to 12 months, with a potential turnaround in demand and supply after a year. Currently, there are many individual data center players in the Tier 1 market that were previously well-resourced, but most are now quite fragmented and do not meet the current AI demands. On the other hand, a few players still possess substantial resources in this market. Given time, these companies will adapt to the AI demand. Our strategy is to be very selective in pursuing orders, aiming for deals that align with our criteria. Additionally, we are open to the possibility of improved pricing. If that happens, it would indicate a favorable and healthy market for us. In regards to the DayOne customer, the new orders are coming from a variety of cloud and video companies across different countries, applications, and workloads, showcasing significant diversification in last year's international orders. Additionally, in Southeast Asia, the primary deployments are focused on high-performance CPUs and cloud services, rather than AI. The main workloads from both Chinese and US customers revolve around cloud growth and video applications, with an emphasis on high-performance CPUs rather than GPUs. Overall, the percentage of the market represented by these products remains relatively small, and I believe the new chip policy will not significantly influence the demand profile in Southeast Asia.

Thank you. That's very clear.

Thank you. We will take our next question. Your next question comes from the line of Jonathan Atkin from RBC Capital Markets. Please go ahead. Your line is open.

Thanks for taking my question. One China and then one, I guess, DayOne. So, what's the use of the ABS proceeds? And can you give us a little bit of a flavor for the customer profile, margin profile, weighted average lease expiration? Just any color about those stabilized assets that you're issuing capital off of? And then, the DayOne question is maybe a little broader. You broke ground in Chonburi, I think, just a couple of days ago. What's the use case you see for Thailand? And then, any kind of update on Batam, what's going well, what are some of the challenges that you're seeing relative to your last conference call? Thanks.

Yeah, the ABS proceeds can be used either to pay down debt and deleverage or to reinvest if the right opportunity is there. We look at new investment opportunities as being one part of the equation and asset monetization as being the other part of the equation. So, this asset ABS issue has been achieved at a good time because we also presented with a very good new investment opportunity at around the same time. When you put it all together, we are able to increase our capital expenditures, but keep our debt at the same level or lower and be able to achieve, at the end of this year, lower net debt to EBITDA. We have a lot of assets that are suitable for asset monetization treatment. We selected assets for the first transactions that we thought would be highly acceptable to investors. The asset we chose for the ABS happens to be one that we acquired a few years ago, and it has mostly financial institution customers, which obviously financial investors have high recognition for those kinds of customers. But, of course, it doesn't have to be this way. For the C-REITs, we chose a different seed asset with quite a different profile. It's more of a cloud internet customer.

Let's discuss the groundbreaking in Thailand. We recently announced that we are starting to build a new data center, which indicates strong customer demand. This new campus in Thailand will be our largest to date, and we have received mixed demand from both US and Chinese customers. We are confident that demand will continue, positioning Thailand as a new hub in the Asia Pacific region. Regarding Batam, we successfully completed the first two phases as promised to our customers and are currently working on the remaining phases. The Batam project is progressing well, and we anticipate increased demand for it as well.

If I could ask about the domestic market in China, you mentioned strong demand in the Internet sector and also referred to DeepSeek. There seems to be a significant ecosystem of AI startups in China. How do you view the sales prospects from AI startups in comparison to established Internet companies that are increasing their capital investments?

I believe the demand is primarily driven by established companies. We are also observing significant enterprise-type demand emerging, as many small enterprises are beginning to experiment with AI. Internally, the sentiment among Chinese enterprises is quite positive, as they are all looking to harness AI to enhance efficiency and boost revenue. This trend is very popular at the moment. Over time, I expect that demand will be fostered by various industries, which is clearly evident. I believe this will unfold in the next few years, and it has already begun.

Thank you. We will take our next question. Your next question comes from the line of Daley Li from Bank of America Securities. Please go ahead. Your line is open.

Hi, management. Thanks for taking my question. I have two questions. The first is about our future series issuance. What do you anticipate the valuation range for this series to be? In the China market, we've observed that warehouse valuations are quite high, around 20 times EV to EBITDA. What is our expected valuation range and yield? My second question pertains to the move-in pace for the China market. Looking at the net quarter-by-quarter move-in pace from clients, we've noticed an increase in rush orders for AI chips in the first quarter. Can we expect a faster ramp-up in the second quarter or beyond? Thank you.

Yeah, Daley, thanks for your question. There are around 50 C-REITs listed in China and we categorize them by the nature of the underlying assets. There are around 25 where the underlying assets are commercial real estate, industrial, business park, logistics, and so on. We think that subset is the best benchmark for a potential data center C-REIT. Those 25 companies, there are two or three outliers, but if we exclude them, what remains is trading in a very well-defined range in terms of dividend yield. I believe that dividend yield is the driver of their valuation, and the multiple is derived from that. The dividend yield is quite concentrated around 5%. If we take that as a reference and assume conservatively that we would offer a data center C-REIT at a yield premium, we can derive what the implied multiple would be for us in terms of our asset monetization, and it's quite attractive. We've set a benchmark of 13 times with the ABS, and we've stated that the investors in the ABS had the explicit intention when the time is right, when all the qualification criteria can be met, to inject that ABS into a C-REIT. So clearly, they expect to be able to do that at some kind of valuation multiple pickup.

In terms of the move-in pace, the new order is on a six-month move-in schedule, which marks a significant change compared to the last couple of years.

Thank you. As there are no further questions, I'd like to now turn the call back over to the company for closing remarks.

Thank you all for joining us today, and we'll see you next time. Bye.

Thank you.

This concludes this conference call. You may now disconnect your line. Thank you.