Daniel Newman
[Slide 12: Underlying Trends] Thank you, William. Before I get in to the numbers, I would like to make some comments about our business model and how we add value. GDS develops and operates high performance data centers. We begin by securing long term tenure of entire buildings in shell state. We either build these shells ourselves, work with property partners who build for us, or convert existing buildings. We invest in structural works and all the critical systems needed to turn the building in to a data center. To give some idea of proportion, our investment is typically multiples of what the real estate alone would cost. We provide move-in ready data center capacity configured to suit the needs of individual customers. Essentially, we have two types of customers: Cloud and large internet who order in high volume and high-power density; and, large enterprises with small to medium order size, lower power density, but high redundancy. We operate the whole of every one of our self-developed data centers as a service for our customers. Our customer contracts specify numerous service level parameters which we commit to deliver. Our operating track record is as important to our customers as the quality of our facilities. Our customers install their IT equipment in our data centers, deal directly with the telecom carriers for connectivity, and manage their own IT operations or, if they wish, outsource to us. For the avoidance of doubt, our customers do not operate any part of our data centers; our customers are not tenants; and, our service contracts are not leases. Getting back to our 2Q18 results, starting on Slide 12, where we strip out the contribution from equipment sales and the effect of FX changes… On a quarter on quarter basis, our service revenue grew by 13.6%, our Underlying Adjusted NOI grew by 15.9%, and our Underlying Adjusted EBITDA grew by 21.3%. Our Underlying Adjusted EBITDA margin increased by 2.3 percentage points to 35.3%. [Slide 13: Revenue & Area Utilized] Turning to Slide 13… The main driver of revenue growth in 2Q18 was the 20,000 square meter increase in area utilized. We define “area utilized” as the net floor area of data centers in service which is revenue generating pursuant to customer agreements in effect. Area utilized is usually occupied by customers, but our ability to recognize revenue, bill and collect is not dependent on physical occupancy. There have been some reports recently about a short-term slowdown in server demand. I cannot comment on the validity of those reports. All I can say is that what we are seeing in terms of move-in by hyperscale customers is acceleration, across data centers in all our markets. Monthly Service Revenue or “MSR” per square meter in 2Q18 was within the range which we have seen over multiple quarters. On a per square meter basis, selling prices are stable and, as the backlog is delivered, we expect MSR per square meter to stay within the established range. [Slide 14: Underlying Adjusted NOI & EBITDA] As shown on Slide 14, profit margins are on an upward trend. The growth drag at the Adjusted NOI level which we saw in the past few quarters has now reversed and we continue to realize operating leverage on the corporate cost base. [Slide 15: Stabilized & Ramping-up Data Centers] As shown on Slide 15, our data center area in service is now 48% stabilized / 52% ramping up. The proportion which is stabilized is actually lower than in 1Q18, mainly due to 3 new projects or phases of projects coming in to service. However, the utilization rate of the ramping up portion is much higher at nearly 45% compared with 31% last quarter. The commitment rate is 98% for stabilized and 93% for ramping up – these are truly remarkable commitment rates by anyone’s standard. As an aside, please note that our GZ1 data center is now 100% committed and 100% utilized. Almost 80% is taken by one of our largest customers. [Slide 16: Cost Breakdown & Operating Leverage] At the corporate level, on Slide 16, SG&A is down to 11.2% of service revenue. If we factor in full delivery of the backlog, the current level of SG&A represents around 6-7% of service revenue. [Slide 17: Capex & On-Going Investment] Turning to our capex on Slide 17… In 2Q18, our capex paid increased to Rmb 1.2 billion, including payment of Rmb 231 million as front-end consideration for the equity of the GZ3 and SH11 acquired entities. The table of data centers under construction shows that we have around 36,000 square meters of capacity not yet committed. At our current run rate, its equivalent to around 2 quarters new customer commitments. [Slide 18: Financing Obligations & Liquidity] Turning to Slide 18… After completion of the convertible bond offering in June, our cash position has risen to nearly Rmb 4.5 billion ($673 million). When we raise capital out of GDS Holdings we receive proceeds in US dollars offshore. We use these proceeds mainly to capitalize our data center project companies onshore with equity. At the initiation of each new project, in our internal records, we allocate cash as the project equity. This ensures that we always have enough funds to capitalize our committed projects through to completion. Depending on the progress of projects and the requirement to fund capex, we inject cash in to the project companies onshore and spend it. Cross-border cash transfers require various China regulatory approvals and must have a specific qualifying purpose. We cannot transfer cash onshore for general purposes and leave it idle. We cannot convert it in to Renminbi without producing capex contracts and invoices. Furthermore, once cash has been injected in to project companies, it is not easy to move to our other onshore subsidiaries as a result of restrictions in our financing agreements and regulatory prohibitions on intra-group cash transfers. Our effective interest cost on borrowings for 2Q18 was 6.2% excluding the CB. Our interest coverage, based on Adjusted EBITDA over reported interest, was 1.7x excluding capitalized interest cost or 1.4x including it. During 2Q18, our net debt increased by Rmb 4.6 billion to Rmb 7.2 billion. Rmb 1.3 billion of the increase was due to capital leases associated with the GZ3 acquisition and additional resources we secured in Beijing, some of which are categorized as held for future development. Mainly as a result of these capital leases, our Net Debt / LQA Adjusted EBITDA ratio increased to 7.8x. Over time, we aim to bring our Net Debt / EBITDA ratio down to conventional levels. However, given our current rate of growth and scale of development, we feel that this metric does not give a good sense of our balance sheet strength. On the one hand, we have stabilized projects which are leveraged at less than 3x. On the other hand, we have incurred debt to finance projects which are ramping up or under construction with little or no EBITDA. Yet this debt is to a large degree covered by customer contracts, a substantial majority with investment grade counterparties. Such contracts tenors which are longer than most of our loans and total contract value far in excess of our outstanding loans. [Slide 19: Financial Resources] I am often asked how much capacity we can develop with our existing financial resources. On Slide 19, we show an illustrative indication. Assuming that we are able to leverage capital injected in to projects 1 to 1 with project debt, we currently have sufficient capital to finance around Rmb 9.0 billion ($1.4 billion) of new investments. On the commitment side, the cost to complete all data centers currently in service and under construction is Rmb 2.2 billion ($333 million). Adding to this the cost to complete projects initiated during the current quarter, the remaining amounts payable for acquisitions (including the maximum amount payable under contingent performance obligations), and the cost of the Hong Kong site, we would require a grand total of Rmb 5.5 billion ($831 million) to complete every data center in service or under construction as of today. The total capacity of such data center portfolio would be around 192,000 square meters. Based on this illustrative indication, we have, potentially, Rmb 3.5 billion available for further new commitments. Turning to Slide 20… We have an excellent track record in managing receivables. Days Sales Outstanding at 70 days is well up to the highest industry standards. We have not had a bad debt since 2015. In other words, we have collected 100% of what we recognized as revenue in the past 3+ years. Any company which recognizes revenue as service is rendered and bills in arrears – which is standard operating practice in ours and many sectors – will have unbilled receivables. Prime examples are companies such as AT&T and PG&E. Unbilled receivables is not a mandatory disclosure under US GAAP but we have chosen to disclose it because it provides transparency to the length of the billing cycle. In the past few years, the proportion of our revenue coming from Cloud and large internet customers has increased significantly. We bill these customers quarterly in arrears. Nonetheless, we have managed to maintain an acceptable DSO. Over 80% of our unbilled receivables at the end of 2Q18 relate to revenue recognized during the quarter. Unbilled receivables are easy to audit as they are tied to contracts in force and are part of a repeating quarterly billing cycle. For Days Payable Outstanding, we look at the opex cycle and the capex cycle. Both are reasonable. You can look at them on a combined basis if you prefer, but to exclude total capex purchases from the denominator, as some have done, yields a completely meaningless ratio [Slide 21: Project Acquisitions] During the past quarter, we had 2 acquisitions which closed – GZ3 and SH11. I would like to take this opportunity to provide some commentary on our acquisition strategy and the deals which we have done. Faced with strong demand from our customers, we started a few years ago to look for data centers which we could acquire to enhance our resource supply. As shown on Slide 21, between May 2016 and the present, we have completed 7 acquisitions. 2 of them (BJ4 and BJ5) are categorized as acquisitions, but were really just a mechanism for acquiring new project companies with agreements relating to property and power capacity. Net of those 2, it’s 5 acquisitions in the usual sense of the word. We have been highly selective in the data centers which we have acquired. They must fit our profile in terms of asset quality and suitability for serving our target customer segments. We have also been very careful in the way that we structured the acquisitions, deferring as much consideration as possible and linking it to the achievement of project milestones. This is referred to as “contingent performance obligations”. Out of the total equity consideration for the 5 real acquisitions of Rmb 1.3 billion, Rmb 752 mlllion has been paid and Rmb 506 million is contingent. We have involved legal advisers and financial advisers in due diligence. We have always bought 100%, paid in cash, and done so at single digit multiples of projected stabilized Cash EBITDA. The outcome has been highly successful. Every acquired data center has met or exceeded our expectations. For example, we were able to obtain additional power capacity and upsize the area of SZ5 by 53% post acquisition. Over 96% of the committed area of the acquired data centers is taken by our Top 10 customers. [Slide 22: Guangzhou Data Center Cluster] Our first acquisition was of the GZ1 data center. As you can see on Slide 22, it is one of three data centers which we now own in the same location. We acquired GZ1, GZ2, and GZ3 in three separate transactions, only moving forward when our acquisition criteria could be satisfied. It should not surprise you given the integrated nature of the Guangzhou development that, while each data center was owned by a different set of selling shareholders, the lead investor was the same. Based on our satisfactory experience of GZ1, we also acquired SZ5 from a different set of selling shareholders with the same lead investor from the Guangzhou cluster. If you examine the corporate records of the acquired companies for GZ1/2/3 and SZ5, you will find that prior to our acquisition they share some directors, contact details, etc., for the simple and obvious reason that they had the same lead investor. You will also find that at the time when we acquired these companies, we replaced the seller appointed directors and other personnel with our appointees at our earliest opportunity. [Slide 23: SZ5 Transaction Structure] I will illustrate this by going through the example of the SZ5 acquisition on Slide 23. Data center services is a restricted industry in China. It requires a Value-Added Telecom Services license. There is a foreign ownership restriction for licensees. In order to comply with this regulation, we have set up a group of PRC companies which hold the relevant licenses and over which we have total control by contractual relationships known as variable interest entity or “VIE”. The structure has been very common in China for the past 20 years and moreover, companies with several trillion dollars-worth of market cap listed in the US utilize this structure. Optimally, for each data center project, we put the provision of services and commercial contracts in to a LicenseCo that is part of our VIE Group, and the data center operating assets into an AssetCo that is under an intermediate holding company in Hong Kong. This ensures regulatory compliance and facilitates financing. It is not always possible. Sometimes the operating assets are in the LicenseCo, but wherever it is possible, this is the most efficient model and the one we use. When we acquire projects, we sometimes find that they are already structured in this way. As you can see on Slides 23 and 24, this was the case with SZ5, GZ2, and GZ3. In order to implement the acquisitions, we first took over the respective holding company in Hong Kong. In the case of SZ5, it is called RDTJ. RDTJ had a wholly owned PRC subsidiary, in essence the AssetCo for SZ5. I will refer to this subsidiary as WGYL. The acquisition of RDTJ was completed on 24, May 2017, as evidenced by the publicly available Annual Return filed with the Companies Registry, Hong Kong, which shows that we also replaced the seller appointed director on the same date. At that point in time, we had effective control over the entire target group by virtue of ownership and contractual relationships. The second step in the transaction is the acquisition of the LicenseCo. In the case of SZ5, it is called Shenzhen Yaode. The acquirer, Beijing Wanguo, is not a subsidiary of GDS. It is part of our VIE Group. Due to the longer time required to complete change of company ownership in China, the Shenzhen Yaode acquisition completed on 29, June, 2017. We replaced its seller appointed directors and other representatives on the same date. On the prior day, we replaced the seller appointed directors and other representatives of WGYL. In China, it is a requirement to file with the State Administration for Industry & Commerce or SAIC for change of ownership, change of directors, etc. Accordingly, we filed with SAIC for the change of ownership and directors of Shenzhen Yaode and for the change of directors of WGYL. However, the change of ownership of WGYL was indirect and did not require an SAIC filing at that time. In our 2016 20F, we refer to the pending acquisition of a data center in Shenzhen, i.e. SZ5, and specifically state that it involves agreements for the simultaneous acquisition of a Hong Kong company and a PRC company. In our 2017 20F, we refer to the acquisition of SZ5 as the acquisition of a “target group”, in other words “a group of companies” – plural. In describing the financing arrangements, we disclose the names of the three companies at the heart of the SZ5 transaction – RDTJ, WGYL, and Shenzhen Yaode. Allegations have been made stating that the SZ5, GZ2, and GZ3 acquisitions were undisclosed related party transactions. Given the facts that the acquired data centers had the same lead investor, the acquisitions involved a target group as disclosed in our SEC filings, WGYL was a central part of the SZ5 acquisition, SAIC filings for WGYL show its owner to be RDTJ, annual returns for RDTJ are publicly available in Hong Kong showing its change of ownership and directors, it is clear that the allegations are baseless. By failing to accurately describe these facts, the short seller made, in our opinion, an egregious error. To compound this failing by making defamatory allegations of fraud and self-dealing, is in our view, totally unacceptable. We encourage our investors and the legitimate analysts who cover us to base their conclusions on objective facts and not on what we view to be obvious and self-interested distortions or mis-statements of those facts. [Slide 24: GZ2 & GZ3 Transaction Structures] Because we have always been 100% committed to transparency, I also want to address the transaction structures for the GZ2 and GZ3 transactions, which are shown on Slide 24. Once again, each acquisition involved a target group, two transactions (one offshore / one onshore), and the filings show that the seller appointed directors, etc. were replaced by our representatives when the acquisitions completed. The composition of the GZ2 target group is disclosed in our 2017 20F. GZ3 was completed this year and we have not yet filed our 2018 20F. [Slide 21: Project Acquisitions] If you go back to Slide 21 where we show the equity consideration, it should be obvious by now that disclosure in an SAIC filing of consideration paid for the onshore acquisition does not include the consideration paid for the offshore acquisition. There is one more point which I need to address. One of the prior shareholders of GZ2 made a disclosure required by domestic stock market regulations which included references to the revenue and property and equipment of GZ2. The disclosed revenue under PRC GAAP did not include power income which we are required to book as revenue under US GAAP in our financial statements. The disclosed property and equipment did not include leases, which under US GAAP, we are required to capitalize in our financial statements. Once again, we believe that the mischaracterization of these facts reflects, at a minimum, obvious and elementary errors. [Slide 25: Contract Backlog Build-Up] Turning to Slide 25… At the end of 2Q18, our backlog had increased again slightly to nearly 57,000 square meters, which is equivalent to 65% of our area utilized at the same date. [Slide 26: Updated Business Outlook] Finally, on Page 26… Our 2Q18 results clearly show that we are tracking ahead of expectations in terms of revenue and Adjusted EBITDA growth. The accelerated move-in has brought forward new service revenue that we expected to commence later in the year. Given where we are today, we are raising our original guidance ranges for revenue and Adjusted EBITDA by approximately 3%. This means that the bottom end of the new range is nearly in-line with the high end of the original range. We have also had to accelerate capex and new project initiation in order to keep up with the sales demand. Accordingly, we are raising our capex guidance for the full year to around Rmb 4.0 billion ($604 million). This includes Rmb 662 million ($100 million) which we expect to pay for the Hong Kong site acquisition. In the future, we may sell and leaseback the site, but for now it is included in our capex. With that I will end the formal part of my presentation. We would now like to open the call to questions. Operator?
