[4 -> 9] Hello, ladies and gentlemen. Thank you for standing by for GDS Holdings Limited's fourth
[9 -> 15] quarter and full year 2022 earnings conference call. At this time, all participants are in
[15 -> 19] listen-only mode. After management's prepared remarks, there will be a question-and-answer
[19 -> 24] session. Today's conference call is being recorded. I'll now turn the call over to
[24 -> 29] your host, Ms. Laura Chen, Head of Investor Relations for the company. Please go ahead,
[29 -> 33] Laura. Thank you. Hello, everyone. Welcome to the
[33 -> 38] fourth quarter and full year 2022 earnings conference call of GDS Holdings Limited. The
[38 -> 44] company's results were issued to their newswire services earlier today and are posted online.
[44 -> 49] A summary presentation, which we will refer to during this conference call, can be viewed
[49 -> 57] and downloaded from our IR website at investorsgdsservices.com. Leading today's call is Mr. William Huang,
[57 -> 63] GDS founder, chairman, and CEO, who will provide an overview of our business strategy and performance
[64 -> 70] Mr. Dan Newman, GDS CFO, will then review the financial and operating results. Ms. Jamie
[70 -> 75] Koo, our COO, is also available to answer questions. Before we continue, please note
[75 -> 81] that today's discussion will contain forward-looking statements made under the safe harbor provisions
[81 -> 88] of the U.S. Private Securities Litigation Reform Act of 1995. Forward-looking statements
[88 -> 94] involve inherent risks and uncertainties. As such, the company's results may be materially
[94 -> 100] different from the views expressed today. Further information regarding these and other risks
[100 -> 107] and uncertainties is included in the company's prospectus as filed with USSEC. The company
[107 -> 113] does not assume any obligation to update any forward-looking statements except as required
[113 -> 119] under applicable law. Please also note that GDS earnings press release and this conference
[119 -> 125] call include discussions of unaudited GAAP financial information as well as unaudited
[125 -> 132] non-GAAP financial measures. GDS press release contains a reconciliation of the unaudited
[132 -> 138] non-GAAP measures to the unaudited most directly comparable GAAP measures. I will now turn
[138 -> 144] the call over to GDS founder, chairman and CEO, William Huang. Please go ahead, William.
[144 -> 150] Thank you. Hello, everyone. This is William. Thank you for joining us on today's call.
[150 -> 158] Despite the challenging environment, our business continues to deliver solid results. In 2022,
[158 -> 166] we grew revenue by 19% and adjusted EBITDA by 15% year-on-year. We won significant new
[166 -> 174] business in China while at the same time adjusting our development program to the current pace of
[174 -> 184] growth. We accelerated our international expansion with notable success. And on the funding side,
[185 -> 190] we took steps to access new sources of capital and strengthen our financial position.
[192 -> 200] Entering 2023, we are looking forward to a recovery. Customers are more positive about
[200 -> 207] their business outlook. When their business pick up, it will flow through to us quite quickly.
[208 -> 216] We could see this happen over the next few quarters. Meanwhile, we are continuing to
[216 -> 225] strengthen our operations and finances. Looking further ahead, the fundamentals
[225 -> 233] for our industry remain very strong. On the one hand, AI and other new technologies are driving
[233 -> 242] waves of demand. On the other hand, data center supply in China's tier 1 markets is gradually
[242 -> 250] becoming more constrained due to the limitations on land and power. As market conditions improve,
[251 -> 256] we are well-positioned to outperform with our customer relationships,
[256 -> 268] contract backlogs, and established asset bases. In 2022, we won 74,000 square meters or 178
[268 -> 275] megawatts of net new bookings, all of which was for our organic data center developments
[275 -> 287] in tier 1 markets. A big highlight was our 64 megawatt win in Johor with significant
[287 -> 293] contributions from the international. Our four-year bookings were up to the level of the
[293 -> 303] past few years. This highlights the importance of our strategy to follow the customer. For 2023,
[303 -> 310] we are targeting a similar level of new bookings, with once again a significant contribution from
[310 -> 321] international. In 4Q22, we won two new large high-scale orders. LF15 is the first order for
[321 -> 330] our new campus at Xianghe in Langfang, Hebei province. The order is from a large internet
[330 -> 338] company which is migrating from our downtown site in Beijing. The customer is relocating to
[339 -> 351] LF15 and to one of our other campuses in Langfang. BJ14 Phase 2 is the second phase of development
[352 -> 361] at our Greenfield site in the Tongzhou district of Beijing. The order is from a first-time large
[361 -> 372] internet customer in the recruitment sector. This win typifies our strategy of matching
[372 -> 380] resource inventory with high potential customers. The new commitments come with confirmed moving
[380 -> 387] schedules. Over the past three years, we have made significant progress with new customer wins
[388 -> 397] and diversification. Large internet companies accounted for 63% of new bookings
[397 -> 405] in 2022, as compared with 23% in 2020. Enterprise and financial customers
[407 -> 415] averaged around 20% of new bookings. This shows how we have been able to evolve our strategy
[415 -> 424] to capturing growth from different customer segments and different locations. Customer
[424 -> 433] moving in 2022 was affected by the lockdowns and other macro factors. Going forward,
[433 -> 440] we are focusing on customers who can commit to fast-moving schedules, typically large
[440 -> 446] internet customers moving faster than cloud customers. Some of the large internet orders
[446 -> 456] we won in 2022 have relatively short moving periods. Hence, the moving rates could pick up
[456 -> 465] as these new commitments towards the end of this year. We have a large backlog totaling
[466 -> 474] 260,000 square meters, nearly half of it related to data centers which are moving ready.
[475 -> 486] This gives us high visibility for future growth with reduced capex. To adjust to the current
[486 -> 496] environment, we have slowed down our capacity expansion. In 2022, we brought 28,000 square
[497 -> 506] meters into service. In 2023, we plan to bring further 58,000 square meters in service.
[507 -> 518] All of which are 33,500 square meters in mainland China and 24,500 square meters in
[518 -> 526] international. Going forward, we will target to reduce the lead time from investment to customer
[526 -> 535] moving. Looking beyond our current construction program, we have over 300,000 square meters
[535 -> 543] of area held for future development in mainland China, which can support multiple years of future
[543 -> 551] demand. The mainland consists of land and powerful campus-type developments at good locations
[552 -> 559] in national hot markets. Our pipeline aligns with the government's east-data, west-computed
[559 -> 567] policies. This is a valuable resource which will give us significant competitive advantages
[567 -> 578] as supply becomes tighter in the year ahead. All of these priorities are directed at
[580 -> 587] further strengthening our strategic market position and improving our operating and
[587 -> 596] financial efficiency. The initial phase of our international expansion is focused on Hong Kong
[596 -> 599] as the regional hub for Great China
[600 -> 606] and Singapore Johor Batam as the regional hub for Southeast Asia.
[606 -> 614] These two hubs rank among the 10 largest data center markets globally.
[614 -> 618] By leveraging our whole market customer relationships,
[618 -> 622] cost advantages from our prefab products,
[622 -> 627] and the proven execution capability in high-scale development,
[627 -> 631] we can accelerate delivery to our customers
[631 -> 635] and rapidly establish market-leading positions.
[635 -> 640] Our first self-developed data center in Hong Kong
[640 -> 643] will enter service in the next couple of months.
[643 -> 647] We have secured anchor customer commitments
[647 -> 652] from leading China and global cloud service providers.
[652 -> 657] Hong Kong One is the first in a multi-year development pipeline
[657 -> 659] for purpose-built data centers,
[659 -> 663] classed in an ideal location
[663 -> 667] for serving both hyperscale and enterprise customers.
[667 -> 674] Hong Kong's position as the primary gateway for network connectivity
[674 -> 677] between China and the rest of the world
[677 -> 683] assures its long-term position as a data center hub.
[683 -> 687] We have put together a unique set of assets
[687 -> 694] ideally suited to meeting new waves of demand.
[694 -> 698] We are making great progress with both land acquisitions
[698 -> 703] and the customer commitments in Southeast Asia.
[703 -> 707] In Johor, we are constructing three data centers
[707 -> 712] for delivery later this year and early next year.
[712 -> 715] We are going from breaking ground
[715 -> 720] to delivering 64 megawatts over five quarters.
[720 -> 725] We have acquired or secured options
[725 -> 732] over the adjacent sites to enable us to scale up.
[732 -> 736] In the next few months, we expect to receive an additional order
[736 -> 740] which would take us to nearly 100 megawatts
[740 -> 743] of commitments in Johor alone.
[743 -> 746] For existing projects,
[746 -> 752] the priorities are to win further orders for Johor and Bagan,
[752 -> 757] build up our local management team,
[757 -> 760] and deliver the initial capacity.
[760 -> 764] At the same time, we aim to establish new projects
[764 -> 768] in Kuala Lumpur, Jakarta, and other new markets
[768 -> 770] in Southeast Asia and beyond.
[770 -> 773] While our position in mainland China
[773 -> 776] is well set for years to come,
[776 -> 779] we believe that GDS International can become
[779 -> 782] a significant second growth engine.
[782 -> 790] Now I will pass on to Dan for financial and operating review.
[790 -> 793] Thank you, William.
[793 -> 797] Starting on slide 22, where we strip out
[797 -> 802] the contribution sales and the effect of FX changes.
[802 -> 807] In 4Q22, our service revenue grew by 1.5%,
[807 -> 811] and underlying adjusted EBITDA was slightly down
[811 -> 814] by 0.2% quarter on quarter.
[814 -> 821] For FY22, our service revenue grew by 19.2%,
[821 -> 827] and underlying adjusted EBITDA grew by 14.5% year on year.
[827 -> 832] Turning to slides 23 and 24,
[832 -> 834] service revenue growth is driven mainly
[834 -> 836] by delivery of the committed backlog.
[836 -> 842] Net additional area utilized during 4Q22
[842 -> 846] was around 10,700 square meters.
[846 -> 851] Around 7,950 square meters was in Tier 1 markets,
[851 -> 855] and the remaining 2,700 square meters approximately
[855 -> 857] was from BOT projects.
[857 -> 861] Net additional area utilized for FY22
[861 -> 864] was around 51,000 square meters.
[864 -> 867] Around 29,000 square meters was in Tier 1 markets,
[867 -> 871] and the remaining 22,000 square meters approximately
[871 -> 873] was from BOT projects.
[873 -> 880] For FY23, we expect additional area utilized,
[880 -> 885] net of churn, to be similar to the level seen in FY22,
[885 -> 889] i.e. around 50,000 square meters of net add.
[889 -> 892] We disclosed on our last earnings call
[892 -> 895] We disclosed on our last earnings call
[895 -> 898] that one large Internet customer
[898 -> 901] will move out of our downtown data centers in Beijing.
[901 -> 907] As a result, we will record 17,000 square meters of churn
[907 -> 912] spread across the first three quarters of 2023.
[912 -> 916] We have already won back more than 17,000 square meters
[916 -> 919] of new commitments from this customer
[919 -> 922] for two sites in Langfang.
[922 -> 926] We will start to move into these sites during 4Q23.
[926 -> 930] As William mentioned, towards the end of the year,
[930 -> 933] we will also start to deliver some other new contracts
[933 -> 935] with faster move-in schedules.
[935 -> 940] Accordingly, the cadence of move-in in FY23
[940 -> 943] will be quite heavily weighted to the back end.
[943 -> 946] The good news is that a pickup later this year
[946 -> 949] will feed into FY24 growth.
[949 -> 954] Monthly service revenue per square meter
[954 -> 959] was RMB2194 in 4Q22
[959 -> 964] compared with RMB2237 for the previous quarter,
[964 -> 968] a decline of 1.9% quarter on quarter.
[968 -> 974] Over the course of FY23, we expect MSR per square meter
[974 -> 977] to decline by around 4%,
[977 -> 982] comparing 4Q23 with 4Q22.
[982 -> 987] This forecast decline is mainly due to change in location mix,
[987 -> 992] including further BOT move-in.
[992 -> 995] Turning to slides 25 and 26,
[995 -> 1000] for 4Q22, our underlying adjusted gross profit margin
[1000 -> 1004] was slightly up on the prior quarter at 50.9%,
[1004 -> 1007] while our underlying adjusted EBITDA margin
[1007 -> 1011] was down at 44.4%.
[1011 -> 1016] In FY22, our underlying adjusted gross profit margin
[1016 -> 1022] was 51.2% compared to 53.3% in FY21,
[1022 -> 1026] and our underlying adjusted EBITDA margin
[1026 -> 1032] was 45.6% compared to 47.5% in FY21.
[1032 -> 1037] The margin decrease was mainly due to elevated power tariffs
[1037 -> 1039] throughout last year,
[1039 -> 1044] which we estimate knocked 1.5 percentage points off our margin,
[1044 -> 1047] and growth drag from international expansion
[1047 -> 1050] of around a further one percentage point.
[1050 -> 1054] The midpoint of our guidance for total revenue
[1054 -> 1056] and adjusted EBITDA
[1056 -> 1061] implies an adjusted EBITDA margin for FY23,
[1061 -> 1067] which is similar to the level seen in 4Q22.
[1067 -> 1070] We have assumed no reduction in power tariffs
[1070 -> 1072] through the course of this year,
[1072 -> 1077] and continuing growth drag from international expansion.
[1077 -> 1079] Turning to slide 27,
[1079 -> 1083] 2022 was a transition year
[1083 -> 1086] in terms of bringing down our CapEx in mainland China,
[1086 -> 1088] on the one hand,
[1088 -> 1091] and accelerating international investment, on the other hand.
[1091 -> 1095] Our organic CapEx in mainland China
[1095 -> 1100] was around 5.8 billion RMB for FY22,
[1100 -> 1103] which is a few billion lower than in the past couple of years.
[1103 -> 1107] International CapEx was 2 billion RMB,
[1107 -> 1111] while acquisition CapEx was around 3.5 billion RMB.
[1111 -> 1115] We are guiding the total CapEx in FY23
[1115 -> 1118] of around 7.5 billion RMB,
[1118 -> 1125] comprising a further reduction to 3.5 billion RMB for mainland China,
[1125 -> 1129] and an increase to 4 billion RMB for international.
[1129 -> 1133] Replacement CapEx included in the mainland China number
[1133 -> 1138] is running at around 200 million RMB in 2023.
[1138 -> 1141] On slide 28,
[1141 -> 1144] we provide some further data points relating to CapEx,
[1144 -> 1149] starting with mainland China.
[1149 -> 1151] At the end of 2022,
[1151 -> 1157] we had around 152,000 square meters under construction.
[1157 -> 1162] The total cost to complete this capacity is RMB 7.4 billion,
[1162 -> 1166] which we expect to incur over the next three years.
[1166 -> 1168] With this additional expenditure,
[1168 -> 1171] our total capacity and service would increase
[1171 -> 1175] to around 667,000 square meters,
[1175 -> 1182] sufficient for us to grow our billable area by around 71%.
[1182 -> 1183] As you can see,
[1183 -> 1187] a relatively small amount of incremental investment
[1187 -> 1190] is required to support a large amount of growth,
[1191 -> 1194] because much of the investment has already been incurred.
[1194 -> 1197] Turning to international,
[1197 -> 1200] our total cost today...
[1200 -> 1207] for the five data centers under construction in Hong Kong and Johor totaling over 100 megawatts
[1207 -> 1217] is RMB 4.1 billion or U.S. dollars 590 million. The total cost to complete is RMB 3.4 billion
[1217 -> 1227] or U.S. dollars 490 million. Looking at our financing position on slide 29, at the end
[1227 -> 1236] of 2022, our net debt to last quarter annualized adjusted EBITDA ratio was 8.0 times on a consolidated
[1236 -> 1246] basis. If we exclude the debt of the international business and add back net assets to our cash,
[1246 -> 1254] our net debt to last quarter annualized adjusted EBITDA ratio was 7.1 times. If we further
[1255 -> 1261] exclude capital work in progress for our construction program in mainland China, our net debt to last
[1261 -> 1269] quarter annualized adjusted EBITDA ratio would have been 5.0 times. Our effective interest rate
[1269 -> 1279] for FY22 dropped to 4.7 percent. Over the course of 2022, we maintained our cash position and ended
[1279 -> 1288] the year with 8.6 billion RMB or 1.2 billion U.S. dollars of cash on our balance sheet. In
[1288 -> 1295] accordance with our treasury policy, we put cash on deposit with banks which are investment grade
[1295 -> 1303] rated and pre-approved by our board. In addition, we hold cash in equity and debt service reserve
[1303 -> 1311] accounts as required by our project financing facilities. We previously had a small amount
[1312 -> 1320] in accounts with SBB for short-term operational purposes. As of today, this amounts to 2 million
[1320 -> 1329] U.S. dollars which is in the process of withdrawal. Turning to slide 30, in January of this year,
[1329 -> 1339] we raised 580 million U.S. dollars gross proceeds from the SUAD new CB with 7-year maturity. We
[1339 -> 1348] subsequently repaid 150 million U.S. dollars of working capital loans. In June of this year,
[1348 -> 1356] we expect to repurchase 300 million U.S. dollars of an existing CB when it is put. As a result,
[1357 -> 1365] we will have almost no debt repayable at GDF's holdings level until 2027. The debt which is
[1365 -> 1374] repayable over the next few years is amortizing long-term project level loans which we refinance
[1374 -> 1382] on a regular basis. On slide 31, we provide some more color on current year funding requirements,
[1383 -> 1390] starting with mainland China. We expect our operating cash flow to be around 1.5 billion
[1390 -> 1398] RMB to FY23, and as I mentioned, organic capex to be around 3.5 billion RMB,
[1398 -> 1406] resulting in negative free cash flow before financing of around 2.2 billion RMB. We aim
[1406 -> 1415] to fund this gap using a combination of project debt and asset monetization. At the end of 2022,
[1415 -> 1426] we had RMB 8.5 billion of committed but undrawn project loan facilities in mainland China. We have
[1426 -> 1432] recently obtained regulatory clearance for our offshore China data center fund. We expect to
[1432 -> 1438] sign the limited partnership agreement and the sale and purchase agreement for the first data
[1438 -> 1447] center asset shortly. The net cash proceeds will be approximately 1.45 billion RMB after deducting
[1447 -> 1453] our 30% capital commitment to the fund. We will have the option to do more such transactions to
[1453 -> 1460] release equity if we choose to do so. We're also in the process of signing a formal framework
[1460 -> 1466] agreement with a leading Chinese insurance company for an onshore version of the China
[1466 -> 1473] data center fund. Over the next couple of years, we expect our operating cash flow for mainland
[1473 -> 1481] China to increase, while capex remains at similar or lower levels to the guidance for FY23.
[1482 -> 1488] Hence, we expect to become free cash flow positive for mainland China within three years.
[1490 -> 1496] Turning to international, as mentioned previously, we expect total capex for international
[1496 -> 1506] of 4 billion RMB for FY23. We expect to finance 50% of this, say 2 billion RMB or 290 million
[1506 -> 1513] US dollars with project debt. For the balance, we are evaluating a number of options for raising
[1513 -> 1519] equity at subsidiary level, including private equity at our international holdco level,
[1520 -> 1526] and or by bringing in local partners at country holdco or project holdco level.
[1529 -> 1539] Turning to slide 32, for the full year 2023, we expect total revenues to be between 9.94 billion
[1539 -> 1550] RMB to 10.32 billion RMB, implying a year-on-year increase of between approximately 6.6% to 10.7%.
[1552 -> 1560] We expect adjusted EBITDA to be between 4.43 billion RMB to 4.6 billion RMB,
[1560 -> 1570] implying a year-on-year increase of between approximately 4.2% to 8.2%. In addition,
[1570 -> 1576] as previously mentioned, we expect capex to be around 7.5 billion RMB for the full year.
[1578 -> 1581] We would now like to open the call to questions. Operator?
[1582 -> 1588] Thank you. If you would like to ask a question over the phone, please press star 1 and 1 on
[1588 -> 1592] your telephone and wait for your name to be announced. And to withdraw your question,
[1592 -> 1599] you can press star 1 and 1 again. For the benefit of all participants on today's call,
[1599 -> 1603] please do limit yourself to one question. And if you do have more questions,
[1603 -> 1608] then please reenter the queue. And we'll now take our first question.
[1609 -> 1617] Please stand by. The first question is from the line of Jonathan Atkin from RBC. Please go ahead.
[1618 -> 1625] Thank you. So, my question was mainly around Southeast Asia and wondered if you could maybe
[1625 -> 1630] talk a little bit about where you see relatively greater challenges to developing capacity
[1631 -> 1640] in Johor versus Bataan, as well commercially in those markets, setting aside kind of the initial
[1640 -> 1644] customer discussions that you've had and the anchor commitment that you have in Johor.
[1646 -> 1652] What are the prospects for those markets to serve sort of Singaporean requirements or act as a
[1652 -> 1658] greater kind of Singapore availability zone versus serving kind of domestic and kind of regional
[1658 -> 1665] demand on a standalone basis? In other words, how tightly coupled do you see Asian demand and maybe
[1665 -> 1674] prospectively Western demand in Johor and in Bataan with respect to kind of Singaporean?
[1674 -> 1680] How tightly coupled is that with Singapore versus acting on their own as hubs? Thank you.
[1684 -> 1691] Hi, John. Let me start with the second part of your question. And the way we view the
[1692 -> 1699] Singapore market and the surrounding areas is as a proxy for the Southeast Asian market.
[1700 -> 1706] I think the majority of the data center capacity in Southeast Asia is concentrated in and around
[1706 -> 1713] Singapore. And a large part of that is there to serve the region, not just to serve Singapore.
[1715 -> 1720] When we look at the situation in Singapore, it's well known because of the historic
[1721 -> 1728] moratorium since 2019 that the developable capacity in Singapore is largely being built out.
[1729 -> 1735] And what we've seen from third party market research is that utilization rate capacity
[1735 -> 1745] in Singapore has gone to low to mid 90%, which is more than full in industry terms.
[1746 -> 1750] The Singapore government is going through a process. They may allocate
[1751 -> 1756] some additional quota to allow some growth in Singapore. But we think that's only going to
[1756 -> 1764] address a very small part of the incremental demand. Therefore, as we've seen in other markets
[1764 -> 1772] where we have a presence like Beijing, Shanghai, Shenzhen and so on, the excess demand will spill
[1772 -> 1778] over. And we believe that it will spill over to Johor and Bataan. We think they will both be
[1780 -> 1787] vibrant and high growth, large scale data center markets. We think that when you look at it from
[1787 -> 1794] the perspective of how the hyperscale customers, particularly cloud service providers deploy,
[1794 -> 1799] they need the diversity of different locations. And I think having
[1800 -> 1811] Singapore is operationally very beneficial. In terms of the pace of development, I think
[1811 -> 1819] we went out on our own in terms of making commitments to both Johor and Bataan. Frankly,
[1819 -> 1828] I think we leapfrogged many other players in doing that. I think that having that dual
[1828 -> 1836] strategy will give us a marketing edge and a solution that others cannot offer. So far,
[1836 -> 1843] what we've seen is that Johor has taken off a little faster, primarily because of the
[1843 -> 1848] existing infrastructure there. I'm talking about power infrastructure and the network
[1848 -> 1857] connectivity. But we are very confident that the time is coming. It may be one year or
[1857 -> 1864] so behind. We've been talking to government in Singapore and government in Indonesia. There's
[1864 -> 1875] a very strong commitment to making Bataan a successful location for data centers. We're
[1875 -> 1884] happy to be a pioneer. In the next few quarters, we'll be able to make progress in solving some of
[1884 -> 1890] the basic infrastructure challenges, even if it necessitates investing in submarine cable
[1890 -> 1897] networks ourselves to kickstart that. Then I think the market will start to pick up momentum.
[1897 -> 1906] Thank you. If I could just ask a brief follow-up. Inside of China, can you talk a little bit about
[1906 -> 1913] customer availability for obtaining servers? Are there supply chain constraints that are
[1913 -> 1920] affecting their ability to procure equipment and ultimately move in? With respect to central
[1920 -> 1926] Beijing, what are your prospects for refilling that capacity with other demand?
[1926 -> 1941] I think in China, we just mentioned this year is a transition for China's whole business. But
[1942 -> 1949] in the last couple of weeks, I came back to China and visited a lot of governments and our customers
[1949 -> 1956] as well. I think I'm quite encouraged by our customers, especially the large internet and
[1956 -> 1962] cloud service providers. Based on our current pipeline, I think the demand, we can see the
[1962 -> 1969] leading indication is pipeline. I think it's very strong. But for the cloud, when I visited
[1969 -> 1976] the cloud service provider, they are all very ambitious. I think they are maybe slightly a
[1976 -> 1983] little bit later than all the internet companies and financial institutions. If you come to China,
[1983 -> 1995] I think I see in the first two months, the consumer data rebounded very fast. This is
[1995 -> 2002] a leading indication that all the supply chain industry will catch up with the recovery. But
[2002 -> 2013] it still needs time. In terms of the current share market, all the resources, I think we
[2013 -> 2019] are very confident. We have just a very limited inventory. So I think this is a very, very good
[2019 -> 2027] aspect. I believe and I'm very confident we will sell them easily in this year. So this is the
[2027 -> 2042] question. Thank you. Thank you. We'll now take our next question. Please stand by. This is from
[2042 -> 2049] the line of Yang Liu from Morgan Stanley. Please go ahead. Thanks for the opportunity to ask
[2049 -> 2058] questions. I have two questions here. The first one is regarding the demand. I think one of the
[2058 -> 2064] reasons to see very divergent demand from cloud and internet companies is that a lot of big
[2064 -> 2071] internet companies are moving away from public cloud. So if we combine the cloud and internet
[2071 -> 2083] together, when do we expect to see the overall demand recovery? Or put another way, in terms of
[2083 -> 2092] GDS full year guidance, what is the assumption of demand recovery behind it? And my second
[2092 -> 2109] question is, in terms of the competition, are you seeing more competitive bidding or price from Chinese
[2109 -> 2116] telcos in front of your big customers? Thank you. Okay, Yang Liu, this is William. I'll answer your
[2116 -> 2128] first question. In general, we just mentioned that our structure of the new booking has changed. If
[2128 -> 2138] you look at the last year, last two years, we used to be leading by almost 85% or 90% incremental
[2138 -> 2146] demand from the cloud service provider. This year, last year, almost zero from the cloud service
[2146 -> 2155] provider. So we still reached 70,000 square meters, right? This is all from the structurally changed to
[2155 -> 2164] the internet plus enterprise. So we still maintain this assumption for our this year's new booking. So
[2164 -> 2175] we just wait. If the cloud is recovered, that's all upside. So that's the profile assumption that we
[2175 -> 2188] made. You asked about what was the assumption in our guidance, in our business plan. So I think the key
[2188 -> 2197] driver of revenue and ultimately of EBITDA is the move in. I mentioned that we're expecting over the
[2197 -> 2204] full year, the move in to be about 50,000 square meters. So we have some exceptional churn. I'll
[2204 -> 2210] keep talking about it. But if you add that back, you'd see that the move in this year will be over
[2210 -> 2221] 60,000 square meters, which is around 15,000 square meters plus per quarter, which is a little higher
[2221 -> 2226] than the last few quarters, but similar to what it's been over the last six or eight quarters. So
[2226 -> 2234] that doesn't really imply much of a recovery. But one of the factors which is behind that number
[2235 -> 2241] is that, as we mentioned, a few contracts which start delivery in around September, October,
[2241 -> 2249] November, December, have faster than typical move-in schedules. Some of it could be quite
[2249 -> 2255] front-ended, which means that we may see quite a bit of move-in towards the end of the year.
[2255 -> 2261] It's not this year, January of next year. So that boosts the full year number without really
[2261 -> 2269] boosting the revenue much this year. But it does set up, I think, for potentially a good rate of
[2269 -> 2279] year-on-year growth in 2024. Yeah, and I want to add a more point. I think the focus for the
[2279 -> 2287] new booking is 75,000 square meters in our new guidance. This is all organic. Historically,
[2289 -> 2295] when we reached some 90,000 square meters, including a lot of acquisitions. So this is
[2295 -> 2303] still a very high level compared with all the global top players, still bigger than
[2305 -> 2317] most of the data center players in the world. Okay. I think in terms of the competition from
[2317 -> 2326] the telco, I think the telco, our strategy is Tier 1 market. We build out all our assets
[2326 -> 2333] in the Tier 1 market or edge of the top of the Tier 1 market. This is, I think, in the last,
[2333 -> 2340] nothing changed in terms of supply. Telco still left investment in this region, right? So I think
[2340 -> 2345] we are not worried about it. And on the other hand, we don't compete directly. They're also
[2345 -> 2350] our customers. I think we have very good relationship with the, work with the three
[2350 -> 2359] telco in the last 10 years. So recently, a few, I mean, a few telcos came to us and tried to
[2360 -> 2366] discuss some cooperation between us, because they also deployed their cloud service,
[2367 -> 2373] want to deploy their service in Tier 1 market. And if someone want to deploy cloud in service
[2373 -> 2381] in Tier 1 market, I think definitely GDS will be the first choice for them to partner with.
[2381 -> 2390] Otherwise, no possibility to provide their cloud service in this city, right?
[2394 -> 2399] Thanks a lot for the answer. Thank you. We'll now take the next.
[2400 -> 2419] Hi, thanks for taking my question. First, could you talk a little bit about what are
[2419 -> 2423] you hearing in the last couple of months? William, I think you mentioned there is some
[2423 -> 2430] excitement from the customers, but typically on the cloud side, which is still a significant
[2430 -> 2438] part of your backlog if you think about the next couple of years. Are they starting to
[2438 -> 2445] see utilization pick up in their existing infrastructure that they need to now start
[2445 -> 2451] accelerating the movement? Or do you think that the cloud acceleration is still going
[2451 -> 2460] to take a little bit more time? And secondly, also on the EBITDA front, we have seen some
[2460 -> 2468] erosion in EBITDA because of power costs and some, I think, mixed related challenges. Do we
[2468 -> 2477] see the guidance that we gave for FY23 as kind of like the sustainable level of EBITDA margin,
[2477 -> 2484] or do we think there could be further pressure as more international business comes online in 2024 and beyond?
[2484 -> 2496] Hello, Gogo, this is William. In terms of the cloud service provider, I think number one, they're all
[2496 -> 2505] making new business plans right now based on the current environment. Previously, a lot of
[2505 -> 2514] overhead is in China, like lockdown policy, now it's gone, it's moved up. And another is the new
[2514 -> 2526] government set up all the structure and all the platform. They made a couple of times statements
[2526 -> 2532] to support the platform and the social media, private company, support their growth,
[2533 -> 2540] continue to grow in China. So I think our customers are also very encouraged by the current
[2540 -> 2549] environment. So they are making a new business plan right now. In general, I think based on
[2549 -> 2556] my composition with our largest customers, top three customers, they're all very encouraged.
[2556 -> 2561] I think in terms of business plan, it definitely will boost in the next few quarters.
[2561 -> 2569] But in terms of moving, there's a lot of effects to impact the moving in terms of the shipment,
[2569 -> 2577] in terms of the previous, they still have some inventory, right? So I think in general,
[2577 -> 2587] I think that's why we still very conservatively assume cloud service provider moving pace still
[2587 -> 2596] maintain a solid level, neutral level, not aggressive. So it's normal. So I think,
[2597 -> 2603] but given the whole environment, looks like it will improve in the next few quarters.
[2610 -> 2617] Just to leave it on margin, I feel like this year's margin is probably the trough.
[2617 -> 2626] But it does depend on several different drivers. One, of course, is pricing.
[2627 -> 2631] It's another topic for discussion, but we feel like pricing is important.
[2632 -> 2640] And there's a setup which could see pricing recover in future years. Secondly, is power tariffs.
[2642 -> 2647] The input fuel costs have started to come down, but we haven't seen that reflected in power
[2648 -> 2654] tariffs yet. And even if it is reflected, it wouldn't benefit us this year. It could benefit
[2654 -> 2661] us in future years. And Chinese government policy may influence that if the government
[2661 -> 2668] wishes to see power tariffs come down, as was their inclination historically,
[2668 -> 2674] the tendency historically. And then the third part is international, where we are creating
[2675 -> 2680] significant value, or we will create significant value in future. But at the moment, it's EBITDA
[2680 -> 2688] negative. I think it will be EBITDA break-even in the first half of next year. But that's still
[2689 -> 2698] depressing on the margin. So it will take until 2025 before the EBITDA margin of international
[2698 -> 2706] is in the 30%. So that growth track will persist for some time. I think for forecasting purposes,
[2706 -> 2712] maybe you use this year as a trough and then see a slight step up over the next couple of years.
[2717 -> 2725] Got it. Thank you. Thank you. We'll now take our next question. Please stand by.
[2728 -> 2732] This is from the line of Frank Lelson from Raymond James. Please go ahead.
[2733 -> 2738] Great. Thank you. Can you comment on the state of the market for capital recycling and the
[2738 -> 2744] appetite for those assets? Is it the same or better or worse than when you originally put
[2744 -> 2749] this plan together? And I apologize if I missed this, but can you quantify the amount of capital
[2749 -> 2755] recycling you expect to do in each of the next two years? And is this sort of the low for CapEx
[2755 -> 2762] this year? Thank you. Let me start with the numbers. What I showed
[2763 -> 2779] was that in 2023, our pre-cash flow before financing is negative by around 2.5 billion.
[2780 -> 2788] And some of that can be financed by drawing down on project finance facilities. And then the remainder
[2789 -> 2798] needs to be financed with our capital. Having done that convertible bond in January,
[2798 -> 2808] assuming we go ahead and repurchase the CB as possible, on a net basis, we increased our
[2808 -> 2815] financial resources. And so I think pro forma, we have around 9 billion RMB of cash. So that is
[2815 -> 2821] available to be allocated to China, to be allocated to international if we choose to do so.
[2821 -> 2829] But we try to prioritize raising additional capital, recycling capital at the country level
[2830 -> 2840] before we allocate our cash. In China, if we assume that the target for this year is to recycle
[2840 -> 2849] around 1 to 2 billion RMB, we are already 10 months down the road of working on our offshore
[2851 -> 2857] China data center fund. We had the regulatory approval. When we signed the limited partnership
[2857 -> 2861] agreement, we also signed the sale and purchase agreement for the first asset to be transferred
[2861 -> 2868] to that fund. And that will realize cash proceeds to us, net of what we put into the fund,
[2868 -> 2876] the 1.5 billion, 4.5 billion, 1.45 billion. So that goes a substantial way to fulfilling
[2876 -> 2882] whatever requirement we have this year. I think the chances are that next year,
[2882 -> 2887] the requirement will be similar, if not less, because operating cash flow will hopefully increase.
[2888 -> 2895] In CapEx, I don't expect to be higher. It might be lower. So that negative free cash
[2896 -> 2900] for financing might be a smaller number. I don't think it's very challenging for us to
[2901 -> 2910] do monetizations at this scale. We have a substantial amount of capacity in this
[2910 -> 2916] offshore China data center fund to do more deals. We've been working for some time on
[2916 -> 2925] onshore equivalents. And now in the process of signing a formal term sheet, a framework agreement,
[2926 -> 2936] an asset has been identified and initial diligence done by the investor, which would also recycle at
[2936 -> 2944] least a few hundred million RMB of additional capital. And then just looking more broadly,
[2944 -> 2950] the REIT market in China is beginning to really take shape. There's a lot of interest. We're
[2950 -> 2959] being approached continuously by investors and banks looking to structure something involving
[2959 -> 2967] data center assets. I think it's an asset class that attracts a lot of interest in China.
[2968 -> 2975] For the China REIT market regulations, it's really designed for stabilized assets. With these
[2975 -> 2983] funds, we made the decision that suited our purposes to focus on pre-core projects that
[2983 -> 2988] were still under development, de-risked to a degree, but under development.
[2991 -> 2996] But that could be a stepping stone towards eventually being able to access the REIT
[2996 -> 2999] market. So I think we're very comfortable.
[3000 -> 3024] Okay, great. Thank you. Thank you. We'll now take our next question. Please stand by.
[3024 -> 3030] This is from the line of General Ying from Nomura. Please go ahead.
[3033 -> 3038] Hi, thank you for taking my questions. I have just one quick one question about China market.
[3038 -> 3045] So we are seeing the latest news about AI technology. So can we actually see any
[3045 -> 3050] potential demand rising from the market about the AI data centers? And do we expect the AI data
[3050 -> 3056] centers will be, you know, could be the outsourcing by like companies like GDS third parties or they
[3056 -> 3061] will, or some of the cloud companies will do the self-building things that will be outsourced
[3061 -> 3073] that will be self-built? Thank you. Yeah, I think you're right. I think now everybody
[3073 -> 3079] focused on the CHAPS GDP, right? But then they forget that in China, this is already a lot of the
[3079 -> 3088] big player already start to use the AI. AI is also very hot topic in China. And we also believe
[3088 -> 3094] this will be the new, will drive the new wave of the demand. And definitely GDS, our product,
[3094 -> 3103] our location all fits AI type demand in future, in terms of power density, in terms of total
[3103 -> 3109] power capacity and the location where we are. So I think we are very, very encouraged by the
[3109 -> 3115] new wave of the demand. This will affect our business in next few years. And just what
[3115 -> 3122] happened in the last eight years ago, when the cloud stopped, right? So I think we are very,
[3122 -> 3134] very happy. I think definitely, I think because the AI type demands ask for more high power
[3134 -> 3139] density. It's very difficult to build by themselves. So also it will be the trend.
[3140 -> 3151] That's what we believe. Thank you very much. Thank you. We'll now take our next question.
[3153 -> 3161] Please stand by. This is from the line of Edison Li from Jefferies. Please go ahead.
[3163 -> 3170] Hi, thank you very much for taking my questions. My first question is about your build-out
[3170 -> 3177] schedule or your delivery schedule. And I found that in 2023, 85% of your delivery is actually
[3177 -> 3182] taking place in the second half, right? Only 15% in the first half. So I wonder if that is driven
[3182 -> 3188] by your demand assessment or driven by your construction schedule? Okay, that's number one.
[3188 -> 3195] Number two is, on your guidance for 2023, obviously the EBITDA growth lags behind the
[3195 -> 3200] revenue growth. If you take out international business, what do you think the EBITDA growth
[3200 -> 3208] can be in 2023? And then lastly, one of your peers recently said that a major cloud service
[3208 -> 3214] providers in China in fact stopped self-building. And I don't know how true that is, but what is
[3214 -> 3220] your view on the cloud service providers self-building program, given the current state
[3220 -> 3237] of the market? Thank you. The first part about the delivery schedule. So we slowed down pretty
[3237 -> 3242] much our entire construction program. I think there's more than 20 projects in that program.
[3243 -> 3249] But as you know, all those projects have customer commitments, and those customer commitments have
[3249 -> 3256] a delivery date. So the ability to slow down depends on the customer agreeing to take delivery
[3256 -> 3264] at a later date. Not all of them. But we wish that to happen, and some of them would like us
[3264 -> 3270] to keep to the original delivery schedule, which is of course perfectly fine by us. We're just
[3270 -> 3279] trying to manage the occurrence of our CapEx to shorten the lead time between when we spend money
[3279 -> 3287] and when customers move in and start to generate income. So that's really what's behind the
[3287 -> 3295] numbers. We do have a substantial amount of capacity in service, which is not yet revenue
[3295 -> 3304] generating. I mean, our utilization rate is 71%. When you have a portfolio in service as large as
[3304 -> 3312] ours, and the commitment rate is 95%, utilization rate is 71%. So that means there's 24% which is
[3312 -> 3321] committed but not utilized. That's a big number. That's our backlog for air in service. It's 120,000
[3321 -> 3332] square meters. So that could move in practically with no additional CapEx. So you've got to take
[3332 -> 3340] that into consideration, right? The completion of the projects under construction is not the only
[3340 -> 3347] growth driver when we already have so much move-in-ready capacity. William, would you like
[3347 -> 3358] to answer the question about self-builds? The question is that one of your peers recently said
[3358 -> 3367] that a very big cloud service provider, in fact, will stop self-build. I think this is the topic.
[3367 -> 3381] I think it's not only one cloud service provider. I remember in the last couple of earnings calls,
[3381 -> 3388] I think we are the largest builder in China. We have a scale advantage. I think finally, some of
[3389 -> 3397] customers start to realize that. So I think in terms of economic perspective,
[3398 -> 3403] they realize it's not very efficient. So I think this is the topic
[3404 -> 3418] that we are discussing internally. So we are trying to encourage this for their future.
[3419 -> 3427] Some of them can also talk about selling some assets to us. We are also considering this option.
[3429 -> 3439] Let's ask about the EBITDA growth rate excluding international. We allocate SG&A to international
[3440 -> 3450] using sound accounting principles. With that allocation, the EBITDA of international is
[3450 -> 3460] negative about 100 million RMB in 2023. So if you like, you can add that back and that's what
[3460 -> 3468] GDS looks like without international. Okay, so on the revenue side,
[3469 -> 3478] do you have anything projected for 2023? Yeah, I think about 1.7% of our revenue
[3478 -> 3485] comes from international. Okay, got it. Yeah, thank you very much, Dan.
[3487 -> 3491] Thank you. I'd now like to turn the call back over to the company for closing remarks.
[3493 -> 3497] Thank you all once again for joining us today. If you have further questions,
[3497 -> 3501] please feel free to contact GDS Investor Relations through the contact information
[3501 -> 3505] on our website or the Pears-Saunders Youth Investor Relations. Next time, see you. Bye-bye.
[3505 -> 3512] Thank you. This concludes this conference call. You may now disconnect your line. Thank you.