GDS Holdings Ltd (GDS) 2017 Q4 法說會逐字稿

Hello, ladies and gentlemen. Thank you for standing by for the GDS Holdings Limited 4Q '17 and FY '17 Earnings Conference Call. (Operator Instructions) Today's conference call is being recorded.

I would 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 4Q '17 and Full Year '17 Earnings Conference Call of GDS Holdings Limited. The company's results were issued via Newswire services earlier today and are posted online. A summary presentation, which we will refer to during this conference call, can be viewed and downloaded from our IR website at investors.gds-services.com.

Leading today's call is Mr. William Huang, GDS Founder, Chairman and CEO, who will provide an overview of our business strategy and performance; Mr. Dan Newman, GDS CFO, will then review the financial and operating results.

Before we continue, please note that today's discussion will contain forward-looking statements made under the safe harbor provisions of the U.S. Private Securities Litigation Reform Act of 1995. Forward-looking statements involve inherent risks and uncertainties. As such, the company's results may be materially different from the views expressed today. Further information regarding these and other risks and uncertainties is included in the company's prospectus as filed with the U.S. SEC. The company does not assume any obligation to update any forward-looking statements, except as required under applicable law.

Please also note that GDS earnings press release and 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 will now turn the call over to GDS Founder, Chairman and CEO, William Huang. Please go ahead, William.

Thank you, Laura. Hello, everyone. Thank you for joining today's call. We have now completed 1 full year since our IPO, and it was a year of great achievement. We made tremendous progress across our business, delivering all that we said and more.

Our market position today is even stronger than a year ago. We see a substantial high level of demand for data center capacity in China, and the growth estimate has been raised. Our business outlook is the best that we have ever known.

The first highlight of 2017 was our sales performance. Total area committed by customers increased by almost 70%. The net addition to our contract portfolio is worth over USD 200 million in terms of annual recurring revenue, with an average contract life of around 5 years. This represents over USD 1 billion of total revenue secured in a single year. We believe it's an industry record, and what is more, the sales momentum is continuing very strongly into 2018.

We always say that we serve the customers who matters. As a demonstration of this, over 80% of our new business in 2018 came from existing customers. A critical fact behind this sales achievement was our ability to continuously expand capacity. We started the year with 13 self-developed data centers in service or under construction and ended it with 20. As of today, we are up to 23.

Sales and capacity growth is flowing through to our financial results. We started the year with a backlog of 24,000 square meters. We delivered the backlog, driving total value growth of over 50% and adjusted EBITDA growth of nearly 90%. We ended the year with an even larger backlog of 41,000 square meters, providing high visibility for future growth in 2018 and beyond.

Every one of our projects is fully funded. We successfully completed nearly USD 1 billion of debt and equity financing during the year, including the follow-on offering in January 2018. We have a stronger capital base to support our business plan.

Last but not least, we strengthened our relationships with key customers, including Alibaba and Tencent, who formally recognized us as their preferred supplier. We also established new strategic partnerships with highly influential industry leaders, namely CyrusOne in the U.S. and the State Development & Investment Corporation in China.

Now let's look forward, and I will start by discussing our market opportunity. It's Slide 5. 2017 was a year when the cloud really took off in China. Our digital economy is booming. Enterprise IT is moving on to the cloud at a very rapid rate. The leading cloud service providers are growing at almost 100%. Nevertheless, our public cloud market is still in its early stage and only around 10% of the U.S. and less than 2% of the total IT spending in China.

Cloud is paving the way for new technology, which is more and more compute and data-intensive, such as machine learning and artificial intelligence, which are huge focus areas for research and investment in China. China is also gearing up to lead the world in larger-scale 5G deployment, which will make Internet of Things a reality.

To find out how all of this translates into data center demand, we only have to talk to our large customers. They are at the leading edge of cloud and the new technologies in China. From what they have shared with us, they see their requirements for data center capacity going to higher and higher levels. They are looking to us to grow our supply in sync with their requirements.

Let's turn into the Slide 6. In our view, the optimal business model for the data center industry is to serve high-volume cloud and Internet companies co-located with high-value enterprises. GDS is already there. Over the past few years, we targeted capturing demand from all the hyper-scale cloud service providers and from the larger Internet companies, which own the most valuable data. These customers give us continuous growth, high fill-in rate, scaled economies and the leading market position. They also give us the highly strategic connection points into their platforms, which are key to driving enterprise co-location.

As a result of this targeting, cloud plus Internet has grown from 47% of our business 3 years ago to 64% 1 year ago, to 79% at the end of 2017. We already serve most of the customers who matter in China. They keep coming back for more capacity. And as they roll out their platforms in all the Tier 1 markets, we keep winning additional cloud connection points. For example, just in the last quarter, we got 2 more POPs from the Baidu cloud and 2 more POPs from AWS.

The hyper-scale cloud and the large Internet companies typically deploy their platform in Tier 1 markets for latency-sensitive data and application and in remote low-cost locations for the rest. In Tier 1 markets, they require multiple data centers in geographically separate zones and a neutral place to connect with their customers and partners. This is where GDS comes into the picture. We create value for our customers by providing the right resource in the right place at the right time. Historically, our customers have outsourced 100% of their requirement in Tier 1 markets. And from what they have discussed with us, this remains their objective.

In remote locations, our customers tend to build by themselves. However, they have started to look at outsourcing this part of their requirement as well. While our business remains focused on Tier 1 markets, we are prepared to follow our most important customers where they need us to go, providing that we can structure the investment to make adequate returns. In this context, we recently agreed to build and operate 3 data centers for one of our largest customers at a remote site in Hebei province in the far northwest of China.

On the enterprise side of our business, the key metric which we focus on is customer account growth. In 2017, we added almost 30% to our enterprise customer base. We have had particular success in the e-payments vertical, hosting all the leading service providers such as Alipay, Tencent pay, UnionPay, as well as the Central Bank settlement and clearing system for e-payments.

Going forward, a major driver for the enterprise customer franchise will be the unique ability to connect through our hub to all the major cloud platforms in our data centers. We also see potential for creating connection hubs for the financial service ecosystem. We have already soft launched our Cloud Connect hub. And in the next 1 or 2 quarters, we plan to upgrade and fully launch this service, initially focusing on the Shanghai market.

Let's move to the Slide 7. One of the benefits of serving the larger customers in China is that it gives us insight into what is coming. We foresaw the accelerating demand curve and stepped up our efforts to secure more resource supply. There are significant barriers to entry in securing qualified real estate and sufficient power in Tier 1 markets in China. This is why supply is only just keeping up with demand. We have successfully dealt with these challenges through: our dedicated local team in each market, partnering with governments-related property developers for built-to-suit data center shells, delivering significant benefits for the local economy and achieving optimal levels of power efficiency to address environmental concerns.

At the beginning of 2017, we had just over 60,000 square meters in service and another 25,000 square meters under construction. During the year, we initiated 5 new projects and acquired 2 data centers, one of which was under construction and one which was operational. By the end of the year, our area in service had increased by 66% to over 100,000 square meters. And we had reloaded the development pipeline with 24,000 square meters under construction at year-end and another 15,000 square meters initiated in January 2018.

While we are adding all this supply, we are also sustaining exceptionally high commitment rates: 90% for area in service and 50% for area under construction as of today. In addition to what you see, we have a significant pipeline of resource, which we have not yet activated. At the end of 2017, we had an estimated 46,000 square meters held for future development, which is mainly additional phases of existing data centers. We have also entered into 4 MOUs for leasing buildings that we expect to provide us with additional 46,000 square meters.

Let's look to the Slide 8. Our business is in a great position. Over the past year, we have extended our market leadership. We are the only platform player in China, far ahead of any other in the carrier-neutral sector. We have close relationships with the customers who matter, and they are always looking to outsource. We operate in a market where pricing is stable and the returns are at acceptable levels.

Continuing our strategy to be the home of the cloud in China, how do we build for that here?

First of all, we must keep up with the accelerating demand from our existing hyper-scale customers by scaling up our resource supply.

Second, we will use our resource advantage to establish relationships with additional strategic customers who we are targeting.

Third, we will enhance our Cloud Connect product to grow our enterprise business. We believe that the cloud connection points in our data centers give us a unique opportunity, which will become more and more valuable as enterprises adopt hybrid cloud and multi-cloud solutions.

Fourth, we will lower our unit development costs through design and procurement initiatives. This will make us even more cost competitive and help protect our return. We are looking at various innovations around standardization, block redundancy, ultra high power densities and intelligent power management.

Fifth, we will leverage our partnership with CyrusOne for cross-border business opportunities into China and all of China and with SDIC for entry into the new markets. We will also seek other innovative approaches to investment outside of Tier 1 markets.

With that, I would now hand the call over to Dan for the financial and operating review.

Thank you, William. Total revenue and adjusted EBITDA both came out higher than the flash financials, which we disclosed in January, and higher than the top end of the guidance for last year.

Let's look more closely at this, starting on Slide 12, where we strip out the contribution from equipment sales and the effect of FX changes. In 4Q '17, our service revenue grew by 16.7%, and our underlying adjusted EBITDA grew by 17.9% quarter-on-quarter. For FY '17, our service revenue grew by 58.7%, and our underlying adjusted EBITDA grew by 111.7% year-on-year.

Turning to Slide 13. The main driver of revenue growth in 4Q '17 was the 11,000 square meter increase in area utilized, out of which 5,000 square meters came from the backlog and 6,000 square meters came from the acquisition of Guangzhou 2 data center completed in October. The monthly service revenue, or MSR, per square meter over the past 4 quarters has been within a defined range and remained so in 4Q '17. The average selling price in the backlog is similar to our current MSR. So as the backlog is delivered, we do not expect the MSR to change materially.

As shown on Slide 14, profit margins are on an upward trend, but it's not a straight line. At the data center level, which is illustrated by NOI margins, there are 2 things going on: data centers are filling up and reaching optimal profit levels, which typically means an NOI margin of around 55%; at the same time, we are in rapid expansion mode with new data centers acting as a temporary growth drag. In 4Q '17, we brought a lot of capacity into service, 3 data centers, Beijing 3, the first phase of Shenzhen 4 and Shanghai 4, which together account for around 17% of our total area in service at year-end. The fixed cost of these data centers is what impacted NOI margins in 4Q '17.

Slide 15 illustrates the mix of our portfolio by stage of development. What's important to note is that we have high commitment rates across the board. As data centers come into service and fill, NOI margins will trend up to benchmark levels. It's just a matter of time.

At the corporate level, as shown on Slide 16, we continue to realize operating leverage on our SG&A. In 4Q '17, this was sufficient to offset the growth drag that I was just talking about. Accordingly, underlying adjusted EBITDA margins were higher in 4Q '17 than in 3Q '17. SG&A, excluding depreciation and amortization and stock-based compensation, hit 13.4% of service revenue. If we factor in full delivery of the backlog, the current level of SG&A represents around 8% of service revenue.

Turning to our CapEx on Slide 17. In 2017, we accelerated our investment activities, adding 7 new projects, including 2 through acquisition. Our full year CapEx paid was just over RMB 2 billion compared with RMB 1.1 billion in 2016. For the self-developed area in service, our unit cost averages out at under USD 10,000 per square meter. This capacity has an average power density of nearly 2 kilowatts per square meter. So the unit cost per megawatt averages out at under USD 5 million.

For the area under construction, if you do the maths, unit cost per square meter appears to work out slightly higher. This is because the power density of this capacity is also higher at nearly 2.3 kilowatts per square meter. In addition, the cost includes front-end power infrastructure and building shell and core costs, which will support several later phases of development. After taking this into account, the unit cost per megawatt for this part of our portfolio is actually lower, about 5% lower.

Now I would like to update you on the progress of each project which was under construction at year-end. Starting with Shanghai 5 Phase 1, it will enter service soon and it will be close to fully committed by the end of 1Q '18. Shenzhen 5 Phase 2 is reserved for expansion by the anchored customer in Phase 1. We expect to formalize the commitment from this customer in the near future. Hebei 1, 2 and 3 are the data centers which we are building to suit for one of our largest customers. They are all 100% pre-committed.

Shanghai 6 is pre-committed 45% by China's leading online travel company. Shanghai 7, which is not actually shown here, is being built to suit for us on the same campus as Shanghai 6. We will include it as under construction when the shell and core are handed over and we begin to incur CapEx later this year. We are trying to accelerate Shanghai 7 as we have pending demand for it.

Chengdu 2 Phase 1 is an expansion project of our Chengdu campus. It is already 25% pre-committed by a major cloud customer. The pre-commitment rate will increase significantly in 1Q '18 with demand from another major cloud customer. Shanghai 8 is a new project which we initiated in January this year. It is located close to our existing Waigaoqiao campus, and anchor customer commitments are coming soon.

Turning to Slide 18. There has been a significant change to our financing profile since we last reported results at the end of 3Q '17. The $150 million CB, which was outstanding, has been 100% converted. We received $100 million proceeds from the equity issuance to CyrusOne in October 2017, and then a further $205 million net of underwriting commissions from the follow-on offering in January 2018. We refinanced about RMB 700 million of bank borrowings, which were due in 2018, with new facilities with longer maturity. In addition, we obtained another RMB 300 million of short-term working capital loans.

So what does all of this get us? Our approach to data center financing is to inject equity into each development company and leverage it with project debt. All of the projects, which we have announced to date, are fully financed with equity and debt. Most of the equity, which we raised in the last few months, is still available for allocation to new projects, which we expect to initiate through this year and next. As we move forward, a similar approach will be adopted to financing these projects.

The credit market in China has been quite tight for a while. Despite this, we successfully secured RMB 3.9 billion, equivalent to over USD 600 million, of new debt facilities during 2017. This gives us a strong basis of confidence for the debt financing ahead. We have an excellent track record in the onshore banking market. In addition, we are actively considering alternative sources of debt, either to lower cost or extend tenor. Around 65% of our debt is floating rate linked to the People's Bank of China 5-year rate. This is a semi-market rate which is quite stable. We believe that we can comfortably deal with the changes in the interest rate environment.

Slide 19 shows our backlog buildup over the past year. At year-end, our backlog stood at nearly 41,000 square meters worth over $190 million in terms of annual recurring revenue, which is equivalent to 63% of our last quarter annualized service revenue. Delivery of the backlog provides high visibility for revenue growth, which brings me onto the subject of guidance on Page 20.

When we look forward 1 year, we start with a very stable base of area utilized already generating space. Our churn rate is exceptionally low, only 0.6% in 4Q '17. And over the course of 2018, we have only 8,700 square meters, roughly 8.5% of our total area committed, which is coming up for renewal. On top of this installed base, revenue growth is mainly derived from delivery of the backlog in place at the start of the year. The timing of delivery and, hence, revenue recognition, depends on when new data centers come into service and when customers move in.

Based on our current view of these factors, we expect full year 2018 total revenue to be in the range of RMB 2.46 billion to RMB 2.56 billion, implying a growth rate for total revenue of over 55%, at the midpoint of the range. We expect adjusted EBITDA to be in the range of RMB 905 million to RMB 935 million, implying year-on-year growth of close to 80%, at the midpoint of the range. Our total revenue and adjusted EBITDA guidance implies similar percentage growth rates for 2018 versus '17, as for 2017 versus '16.

We expect full year CapEx for 2018 to be in the range of RMB 2.6 billion to RMB 3 billion. The increase year-over-year follows on from the higher level of new customer commitments in 2017. The range takes account of potential upside in our current year sales performance.

We are not providing official guidance for sales. However, based on what we have already done in the year-to-date and our high probability sales pipeline, we are confident of beating the 41,000 square meters of net additional customer commitments which we achieved in 2017. We expect to make significant progress towards this target in the first half of 2018.

To give you some feel for the quarterly cadence of our financial results, move-in during the first quarter is normally at a lower level due to the major Chinese New Year holiday. Thereafter, increase in area utilized and, hence, revenue growth, will accelerate with higher growth in the second half of the year. This cadence is consistent with prior years.

Finally, I should mention that our guidance includes the assumption of one further acquisition of a data center under construction, a deal which we are very close to signing. As with our prior acquisitions, we expect the terms to be highly accretive. Our criteria for these deals are very stringent, but we are seeking more opportunities, and project acquisition will remain an integral part of our capacity sourcing strategy.

With that, I will end the formal part of my presentation. And we would now like to open the call to questions. Operator?

(Operator Instructions) We have the first question from the line of Jonathan Atkin.

I wanted to follow up on the last point that Dan made about acquisitions of either shell capacity or maybe fully built-out data centers. It sounds like the guidance contemplates some -- what are those projects? And can you tell us a little bit about what does the supply look like for those sorts of M&A activities through the year? Is it too numerous to count? Or are they kind of situational in certain metros? Provide -- I'd be interested in your perspective on that. And then the other question I had relates to the Cloud Connect hub. It sounds like you're starting to deploy it. Are there additional product features that you are looking to introduce? And how would you characterize customer demand for that product? Or is it still at a very early stage?

Jon, on the subject of acquisitions, when we look at the industry landscape, and people often ask us about competition, and one feature of the last one -- couple of years has been quite a lot of new entrants, who are typically developing 1 data center or 2 data centers and looking simply to make a successful project and maybe then to capture the profit from it. So this is what is giving rise to the acquisition opportunities. Frankly, the pipeline of those opportunities is getting larger. But we have very stringent standards. The data center facility needs to fit with our customer profile, and that's not often the case. So we screen out a lot of these opportunities. But what we're left with, yes, is there is good promise, I would say, for further acquisitions. And we like to do them while the data center is still under construction, so that we look at it really as an alternative strand of sourcing capacity.

Yes. Jonathan, let me answer your second question. Last year, we deployed the Cloud Connect solution to our customers. It's very early stage product. After almost a half year of testing we been working to develop more a functionalities for our customers and improve our user experience. I think in the next 1 or 2 quarters, we will officially launch this product. This is part of our strategy to build our products and become the home of the cloud. We want customers in our data centers to be able to conveniently connect to the cloud and have a great user experience.

And then a last question about the speed with which customers are moving in, so kind of the backlog to revenue conversion. Well, you have already preserved a state at the sites, and there's a substantial amount of pre-commitment. What is the approximate time lag between delivery of a data center and then occupancy of that committed space?

Yes, Jon, this is a characteristic of the contracts with the hyper-scale cloud and large Internet customers. As you can imagine, a very large customer does not deploy 100,000 servers on 1 day. So they secure capacity, and something that they find very valuable is having flexibility about the timing of the move-in. Now that's something that we are prepared and organized to give them as part of our skill to manage the timing and phasing of our own CapEx to minimize the front costs before they move in. So the contracts, we base it on the delivery schedules, which are really a kind of fallback or minimum. The contracts can give the customers flexibility for 12 to 18 months, but we can't forecast on that basis because almost every single contract is actually at a move-in which is ahead of the contractual minimums. So what we have to do is form a view based on what the customers tell us about their intentions. I would say on the whole, they fully move in within, say, 4 to 5 quarters.

We have the next question from the line of Gokul Hariharan.

My first question was on the strategy to move towards Tier 2 cities as we let them off the more remote areas. Could you talk about the competitive landscape that you encounter? I think we remember that a lot of the telcos had actually invested in a lot of the smaller tier cities and a lot of the remote areas over the last 2 years. So how do you think about the competitive landscape? It seems to be pretty different from what you encountered in the Tier 1 cities.

Yes, Gokul, I'll just preface this by saying, overwhelmingly, our business today and our strategy is based around Tier 1 markets. But we are responsive to our customers, and we will go where they want us to go. I really think the major distinction is between Tier 1 markets and remote locations. So in remote locations, there's nothing, so there is no competitive landscape. It's going to be a question of build-to-suit. Tier 2 markets, we're not really sure what Tier 2 markets are. We formed this partnership late last year with State Development & Investment Corporation and the 2 fixed-line telecom incumbents, China Telecom and China Unicom. And I think one of the objectives of that was to pool our strengths. So we're not going up against the telcos. We are partnering with them. And we found that quite an appealing way of establishing a presence in some new markets. Initially, only Tianjin is confirmed, but there could be one other city in a few quarters' time. And going forward, we may expand in those markets ourselves, but it's a very low-risk way of establishing our market presence. I think partnering with SDIC and telcos, I'd go as far as to say that the access is pretty much assured.

Yes. To echo Dan's explanation, our strategy to the Tier 2 markets is partnering with the SOE because it's always easier to get municipal government business commitment. So it will reduce our risk to get into this market.

Okay. Just one other question. The CapEx guidance of RMB 2.6 billion to RMB 3 billion, does that budget for acquisition cost as well? Or is that primarily self-built data center CapEx?

Yes, it does. It includes one acquisition that I referred to because that acquisition is imminent. And when we acquire a data center which is under construction, essentially, we're paying the cost to date, net any liabilities, and then incur directly the cost to complete, plus a relatively small premium. So that is included in the CapEx guidance for one data center. Now going back to, I think, the first question from Jon Atkin, there are potentially more than one, but there's only one that is definite enough to include in the CapEx guidance as of today.

We have the next question from the line of Robert Gutman from Guggenheim.

First, I'm not sure if you broke out the revenue from the acquisition, the data center acquisition that you made and the expected revenue from the one pending. And secondly, I was wondering, it seems like there's been this 30% year-over-year increase in enterprise logos. Do you see that -- how long do you see that continuing at that pace?

Yes, the acquisition was in October, and it was unusual because actually, this was an acquisition of a data center that was already operational. And I'll just explain the circumstance. It was next door to a data center which -- one of our existing data centers, which we acquired nearly 2 years ago. And actually, we had intended to acquire the second data center, the one we call Guangzhou 2, at the same time. But we had to wait because there were certain criteria around power capacity and redundancy that were not satisfied. By the time we were able to complete the acquisition, the data center was fully operational. So yes, it did contribute. I think if you want to estimate the revenue contribution, I mean, it's a couple of months. The revenue per square meter for that data center is below our average, but revenue per square meter doesn't indicate return. I'd tell you, the cap rate for that acquisition was something like 14%. So yes, I'll just leave it to you to make that estimate. But in terms of our sustainability of enterprise customer growth, 30% per annum.

Robert, this is William. As I mentioned earlier, we understand the optimal model of data center business is to, in parallel, to increase your hyper-scale customer, plus to increase your retail customer, which we call valuable enterprise customer. So to keep growing our enterprise customer is one of the sales strategies. If you look back last 3 years, almost every year, we continued to grow our customer number at 30% annual rate level. So in our resource plan, we still leave a lot of the capacity to bring in the valuable enterprise customer. This is our target. And we expect we still can maintain 25% to 30% annual account number growth in the next few years. This is very strategic and protects our investment return.

That's great. So if I had one follow-on, could you just update us a little bit on some of the initiatives or provide some color on the interaction with CyrusOne?

Yes. After CyrusOne became our shareholder, I brought our team to visit their data center and discussed all various business stuff in terms of design and construction and supply chain management and cost per order, customer referral. This is all part of the objective we try to achieve. We put this cooperation as 2-phase. First is case-by-case. We already start to work together to try to let CyrusOne help us improve our design for more cheaper and build more fast. And this is already start to work with one very, very valuable customer already. And on the other hand, we also help our Chinese customer to go to the CyrusOne data center in the U.S. I believe these 2 low-hanging fruits will happen in a few months. So this is all case-by-case right now. But we do have the midterm cooperation to achieve; to reduce our costs and try to standardize our design partially and get skill to improve our procurement power. We will maybe integrate their procurement and our procurement skill to squeeze our vendor together. This is our midterm target. And we also start to design some new data center in China with CyrusOne's support. This is our pilot case to deploy a new design to the China market. So this is all already initiated. Given 6 months or 12 months, this effort will contribute to the company pretty good benefit.

We have the next question from the line of Suzy Xu from Citigroup.

This is actually Arthur Lai. And so first of all, congrats for a record level of customer commitment and also growing by a lot. So I have a question on your presentation, Page 8, leverage our partnership. So you mentioned that the SDIC, actually, you signed the agreement with them and entered into the Tier 2 market. So from a street angle, I wonder why they choose GDS? Is it because you guys have uniqueness or because something interesting? And the second is how we benefit from this alliance. Should we think of it's a service revenue and lower CapEx and higher margin? And also, the third question is probably to Dan. In fact, how do you think of the timing of start to model these things?

Okay. Dan, I will answer the question, and maybe you can cover some of what I haven't covered. So why choose GDS? I think in China market, it's obvious. If they're smart enough, and they are smart, right? First of all, SDIC did a deal. They are a traditional state-owned investment giant and invest in all different kind of infrastructure, including power. So I think this type of business is what they were seeking for some new business. So they also have their government reputation. They have evaluated their partner almost 1 year before we signed the MOU. And their criteria is very simple. They want a data center company who has full capability: from design and construction and site selection and operation. And they also like that GDS has the best track record in the market; well-recognized by their customer. So this is the key criteria what they're seeking in their partner. After one year evaluation, they made a decision to work with GDS and also invited China Unicom and China Telecom to form a JV type company together. Our target is to build a data center in Tianjin first. But we still discussed another couple of the locations, which we can build our data center and serve the state-owned company and to host the government system. This is what's our plan. Currently still in discussion where we start the project and how many projects we should put in this year or next year.

(Operator Instructions) We have the next question from the line of Colby Synesael from Cowen and Company.

This is Michael Elias on for Colby Synesael. Two questions, if I may. Earlier, you mentioned that cloud took off in China in 2017. Can you better -- can you help us understand where China stands in the cloud adoption cycle relative to the United States? And second, I believe you mentioned earlier that 50% of the area under construction is pre-committed. Can you help us understand the mix of customers that are pre-committing the space and if it's a specific vertical or it's more broad-based?

What's the first question?

The first question was about the -- where is China in the cloud adoption curve.

Yes. As I mentioned, last year is a very big year for the China cloud. It took off. If you look at what Alibaba, Tencent, they announced, the cloud revenue growth is all in a 3 digit growth. I think this is just a start. Based on my observation, in last year, the total Ali-cloud reached -- because their financial year is the end of the first quarter of this year -- I think the total revenue will be close to the USD 2 billion. And they are 50% of the market share. They are the market leader. Compared with U.S. market, as I mentioned, it's just 10% of the U.S. peers. If you look at Amazon, their cloud revenue last year reached $20 billion already. And Microsoft's already reached $20 billion already. So Alibaba, is just around $2 billion. So it's quite early stage, but it's moved very fast. This is the fact. And we also see a lot of the financial institutions start to use the cloud. So this is a signal that means that they are in the testing period. I think in this year or next year, I believe a lot of the enterprise will adopt the hybrid cloud model.

Yes, the question about the backlog and pre-commitment. In almost -- most of the backlog, almost all of it is the top 5 or 6 cloud service providers and Internet companies in China. The sales cycle and the delivery cycle for enterprise is relatively short. Enterprises may only commit to a data center 3 months before it comes into service or it's not pre-commitment at all. So naturally, the backlog which will reach the enterprises are never going to be that large because they take delivery within pretty short order. So it's really the hyper-scale and large Internet which has created this huge backlog. And almost all of the new business from that segment is pre-commitment. In fact, we said on some previous earnings calls is that we're doing joint resource planning with our largest customers. They're asking us to align our resource plan with their requirements. And so they often evaluate and give feedback on projects that we're considering taking on. By the time we do decide to move forward, we already have a soft commitment from them. So when we disclose pre-commitment, that's contractual or legally binding. But we actually have a commitment that we consider to be real and quarters before that.

As there are no further questions at this time, I would like to turn the call back over to the company for closing remarks.

Thank you all once again for joining us today. If you have further questions, please feel free to contact GDS Investor Relations through the contact information on our website or the Piacente Group Investor Relations. Thank you.

Thank you. Ladies and gentlemen, that does conclude our conference for today. Thank you for participating. You may all disconnect.