Description

Company Overview

GDS, founded by William Huang (黄伟) in 2001, is the largest carrier-neutral data center (DC) provider in China. It operates 88 DCs with over 90% of the capacity in and around major Tier-1 cities across China. The geographic proximity to Tier-1 cities provides GDS with a distinct advantage, moat, and growth opportunities vs. peers within the industry (we will address this in detail below). Our investment thesis is as follows: 1) The DC industry in China is nascent and has ample opportunity to grow multiple folds over the coming years, 2) GDS has a moat today due to its Tier-1 cities focus and will further increase its moat through value-added services as the DC industry matures, 3) GDS’s management team has shown themselves to be strong leaders in terms of both execution and capital allocation capabilities, and 4) We have an opportunity to become long-term shareholders at a ~50% valuation discount vs. global peers, providing us with a margin of safety. As with any investment, there are investment risks to consider: 1) In asset heavy industries we always need to keep an eye on competition and supply vs. demand dynamics, 2) There is meaningful customer concentration, therefore we need to monitor the health of GDS’s customers, 3) Potential adverse changes in its operating cost structure as GDS has long-term contracts, which may lead to margin compression, and 4) Balance sheet risk as GDS is highly levered and spends a significant amount of capital to grow its business through the aid of debt and equity financing. We have diligenced each of the core risks and believe that management’s track record alongside long-term industry tailwinds will help to minimize the overall risk profile. Ultimately, our investment in GDS satisfies our three main criteria of: 1) A company with a tremendously long runway for growth, since we can see the DC industry grow to multiple folds its current size, while continuing to gain share, 2) Investing with a meaningful margin of safety due to GDS being valued at a ~50% valuation discount vs. global peers, while exhibiting more than triple their growth rate, and 3) Backing a strong management team that has weathered and thrived through multiple cycles within the industry.

Thesis #1: Growth Opportunity

The DC industry in China is nascent and has ample opportunity to grow multiple folds over the coming years. To arrive at this conclusion, we need to benchmark China vs. the US (refer to analysis on the right).[1] To start off, we know the current market size (TAM) and online population for both regions.[2] In addition, we know the total data consumed for each country on a GB per capita per month basis. Most notably, we can see that US data consumption is ~6x that of China on a per capita basis. This gap has been shrinking rapidly; just two years ago the delta in consumption would have been ~10x. The reason for this gap is due to the nascent cloud and “as-a-Service” (aaS includes, IaaS, PaaS, and SaaS) industries within China. As of today, the aaS industry in China is only ~10% the size of the US, but growing rapidly.[3] We believe that over time, China’s aaS industry will be equal in size (if not larger) than the US because it is a proven form of efficiency gains. From US data, it has been proven that the aaS industry increases companies’ overall IT efficiency by 30-50%.[4] With China’s aging demographics, improving its overall efficiency is paramount and will lead to a longer duration of strong GDP growth, an area that the government cares very much about.[5] Using these available datapoints, we can calculate a per unit TAM for China’s DC industry (RMB0.21) vs. the US (RMB0.17). As we can see, the unit TAM is higher in China vs. the US, which is not a common sight in other sectors or industries. The reason for the China premium is due to China’s unique internet infrastructure (discussed under Thesis #2: Moat). Building up to a mature China DC industry, we first use the US unit TAM (for conservatism) and apply a combination of mature US data usage metrics alongside China’s online population. The resulting TAM is ~5x larger than it is today, indicating ample growth opportunity for the foreseeable future. Note that this exercise only assumes that China’s data usage catches up to the US. The reality is that for both countries, the total data usage continues to grow exponentially due to the advent of 5G use cases such as consumer and industrial Internet of Things (IoT), smart cities, AI, big data, digitization of everyday life, autonomous vehicles, etc.

In addition to the market opportunity, we must also consider GDS’s positioning within the industry. As of today, GDS has high-single digit (HSD) market share in China. Its closest competitors in the carrier-neutral space all have less than half the number of cabinets that GDS has and exhibiting slower growth in terms of the absolute number of new cabinet additions because they lack the inventory (land bank plus power quotas) to support future demand. This results in GDS taking share in the carrier-neutral market as just a few years ago its share was in the mid-single digit (MSD) range. In addition, the three telco providers (carriers) in China hold roughly half of the DC market share. The China telcos, which are all SOEs, have been consistently losing share to the carrier-neutral peers due to their slower pace of operations and lower quality products. As a reminder, SOEs are not typically run for profit and efficiencies, which is why we tend to avoid them altogether. Lastly, to triangulate on GDS’s mature-state market share, we look to our neighbors across the Pacific; Equinix (EQIX), which is the largest carrier-neutral DC operator in the US, has high-teens market share. With GDS taking share from both carrier-neutral DC operators and the China telcos, over time we should see GDS arrive at a similar market share vs. EQIX (we have assumed steady-state market share of mid-teens to be conservative). As outlined in the analysis above, with a future market TAM of ~5x vs. today and the ability to double its market share, GDS has an opportunity to grow its revenue ~10x vs. today.

Thesis #2: Moat

GDS has a moat today due to its Tier-1 cities focus and will further increase its moat through value-added services as the DC industry matures. As we have alluded to above, the fact that GDS has capacity and inventory in all Tier-1 cities across China is a moat for its business. To understand this phenomenon, we need to dive into the internet infrastructure in China vs. the US. The best way to describe the US internet infrastructure is to use a “web” analogy (refer to first chart below). To transfer data from A to B, it is a direct path. Therefore, to cater to all the nodes, DCs can be positioned anywhere. As shown, DCs positioned in any location, 1, 2, or 3, would work for the entire network. In contrast, the best way to describe China’s internet infrastructure is through a “decision tree branch” analogy (refer to second chart below). China’s internet infrastructure is built with multiple layers and all data must ultimately filter through one of the ~20 nodes at the national level.[6] These ~20 nodes house the Great Firewall of China, which limits access to the rest of the world. In addition, each node at every layer house China’s Golden Shield Project, which filters domestic communications (i.e. when certain things are not allowed to be said on WeChat).[7] Therefore, even though locations A and B are seemingly right next to each other, the route taken is long and tedious with filtering occurring at each stage of the journey – all of which ultimately increases latency. Under such a scenario, to provide the lowest latency, DCs should only be positioned in location 1, whereas locations 2 and 3 are suboptimal (even though some nodes are very close to the DC, the average distance travelled from all nodes is longer). Uncoincidentally, the ~20 nodes that make up the national level are effectively the Tier-1 (and Tier-1.5) cities and surrounding areas in China. To make matters worse, Tier-1 cities house the densest population within China leading to the energy infrastructure in these regions being stretched to their limits. Therefore, power in Tier-1 cities is rationed and any project that draws a significant amount of power (see: DCs) must receive power quota approvals before commencement. The low latency advantages coupled with the limited ability to obtain land and even more so power, make Tier-1 cities DC locations more expensive and a scarce resource, one that GDS has a clear lead on.

GDS has an advantage over its peers in a few areas. For one, it is the only carrier-neutral DC provider with significant capacity across all Tier-1 cities including Beijing, Shanghai, Guangzhou, and Shenzhen. Any DC customer who has a presence across China (such as Alibaba’s cloud platform, Alicloud) would prefer to deal with one provider such as GDS rather than multiple regional providers for different cities. None of GDS’s peers have a similar broad coverage across Tier-1 cities. The second advantage is that GDS has made the strategic decision to acquire land and power quotas in Tier-1 cities far ahead of near-term demand due to their scarce nature. The result is that GDS has enough inventory today to fuel more than seven years of future growth. Potential customers want to partner with DC providers that can match their own multi-year growth plans and no one other than GDS can deliver on this. Case in point, from GDS’s two closest peers, one of them was not able to secure enough land and power to support 2022 growth plans and had to lower guidance in early 2022, while the other one is close to running out of inventory in Tier-1 cities and has focused more on remote location DCs.

Lastly, we believe GDS will increase its moat as the industry matures. To understand this evolution, once again we will need to look towards the US and EQIX. Close to 20% of EQIX’s total revenue (over 30% of EQIX’s US revenue) is from interconnection. Interconnection is the direct transmission of data from two B2B software companies housed within the EQIX ecosystem using its own direct fibers connecting multiple DCs and within single DCs.[8] Having direct lines of communication decreases latency for B2B software, increases efficiency and customer satisfaction, and ultimately lowers customer churn. As more software migrates to the cloud, being plugged into the EQIX ecosystem is a must-have to remain competitive vs. peers. GDS and China are still in the preliminary stages of interconnect. Whereas EQIX has over 10k customers, GDS has just under 1k. China’s aaS industry is still ~10% the size of the US, but as it matures, we will start to see more value-add (see: moat) accrue to GDS as it builds an interconnection ecosystem. There is one slight wrinkle in China where only the three China telcos are allowed to own fiber, but there are ways for GDS to still build an interconnection network either through leasing the fiber from the telcos, building its own direct DC to DC fiber, or through cross-connect where GDS’s customers can communicate with each other within a single GDS DC or availability zone.

Thesis #3: Strong Management

GDS’s management team has shown themselves to be strong leaders in terms of both execution and capital allocation capabilities. William and GDS’s management team can be seen as visionaries in the industry. From humble beginnings in 2001, they were pioneers in the industry and first started to work with financial institutions to fulfil their needs. As the cloud providers started to scale in the 2010s, GDS saw an opportunity to partner with them and accelerate its growth through hyperscale DCs. Over the last couple of years, we have seen the emergence of internet giants, where GDS was again early to partner with them directly and cater to their needs. As of today, GDS can count all large internet and cloud companies in China as its direct customers. As the B2C industry in China matures, these companies have looked to expand internationally (predominantly into Southeast Asia), an area where GDS is again early to the game. Lastly, within China, SOEs demand for DCs continue to grow and GDS has already partnered with some of the leading customers in various state-owned sectors. We like management teams who take long-term views on their businesses and invest appropriately for the future; GDS’s management team does just that. This foresight has translated into a ~50% CAGR in terms of utilized cabinet growth and revenue growth and ~100% CAGR in terms of EBITDA growth from 2014 to 2021.

One of the most important traits we look for in management teams is capital allocation capabilities and GDS exemplifies this. A good example is how GDS solved a tough request from its largest customer, Alicloud. As a reminder, Tier-1 cities hold scarcity value (limited supply of land and power) and therefore garner higher ROIs vs. lower-tier cities and more remote locations. This creates an issue when Alicloud requested GDS to build a few specific built-to-suit (BOT) DCs in lower-tier cities, which would result in a lower ROI for GDS. As the DC business is asset-heavy, every RMB spent by GDS on lower-tier cities is an opportunity cost vs. spending that same RMB in Tier-1 cities to garner a higher ROI. GDS overcame this issue by partnering with infrastructure funds, in this case GIC (Singapore sovereign wealth fund). Initially the format was for GIC to put up 90% of the capital required to build these BOT DCs and earns a healthy ~10% return on its capital (guaranteed by Alicloud). GDS will invest the remaining 10% of capital required and incrementally charge a management fee, which would have resulted in GDS earning a ROI higher than its Tier-1 cities DCs. Since then, the investment split between GDS and GIC have been reconfigured to be closer to 50% each, but this example showcases GDS’s management team’s strong understanding of capital allocation.

Thesis #4: Valuation with Margin of Safety

As a technology infrastructure company, GDS is in the real estate business. For the real estate industry and REITs, businesses are valued on a free cash flow (FCF) yield basis. Aside from using industry valuation norms, we see ourselves as part owners of our portfolio companies and will always prefer to value them on their FCF as that is the true and final value accrued to shareholders. There is an added layer of complexity when valuing GDS since it is squarely in its investment phase and will be for years to come; China’s DC industry is still far from mature. In order for us to see the true earnings power of GDS, we need to make a few assumptions: 1) GDS stops investing in new DCs and only completes the remaining capacity in its backlog (note that GDS has ample cash reserves to do so without raising additional equity), 2) Run-rate utilization should be 95% as all DCs in-service have contractual utilization rates of 95% after the initial two year move-in period, 3) Normalized FCF margins of roughly 30-35% triangulated through our conversations with management teams, industry expert, and US peers. Ultimately, we arrive at an 8.1% FCF yield. When looking at yields, we should always benchmark it relative to the risk-free rate (RFR) and in this case, SHIBOR was trading at 2.5%. Therefore, we can say that GDS’s valuation is SHIBOR+5.6%, while growing normalized FCF at ~30% per year. Now we look to compare GDS’s valuation to global peers such as EQIX, DLR, CONE, CORE, and SWCH, which on average has a yield of 4.1%. With LIBOR at 0.2%, we can say that US peers trade at LIBOR+3.9%, while growing normalized FCF at ~8% per year. In comparison, we can see that GDS is trading at an ~45% wider spread on top of the RFR (5.6% vs. 3.9%), double the spread on a total yield basis (8.1% vs. 4.1%), while growing FCF at around ~4x vs. mature US DC peers (~30% vs. ~8%); we see this as a significant margin of safety.

Fast forward 12 months - What Happened?

Fast forward to the end of September 2022, GDS’s stock price is down 60% YTD. With such a meaningful decline in the stock price, we need to ask the pointed question: were we wrong on GDS? Because real life is not black and white, there can be a range of outcomes bookended by 1) the market is right, we mis-diligenced business fundamentals, and did not buy with enough margin of safety or 2) the market is wrong and is looking at GDS through a myopic lens. The hard part, and our job, is to have the intellectual honesty to differentiate between the two. We will outline and review the major events (news and noise) that have transpired since our investment and assess whether this has negatively impacted the fundamentals of GDS’s business. In addition, for any fundamental impacts, we need to better understand if it is temporary or longer-term in nature.

·         ADR delisting noise negatively impacted GDS’s stock price YTD and most notably over the three-day period starting March 10, 2022, where its stock price plummeted 51% (rebounded subsequently). Even though there has been some positive news around China and the US reconciling their differences by having the PCAOB visit HK to perform audits in September 2022, we do not underwrite to either a successful or unsuccessful outcome. GDS is dual-listed in HK and the US with the shares fully fungible between the markets and will not have its business be fundamentally impacted from any trading restrictions; nor would it impact our ability to trade the stock.

·         Covid resurgence throughout the year across China had a negative impact on sentiment across all Chinese stocks. Due to the difficulty in shorting A-shares (China on-shore listed companies), H-shares (HK listed) and ADRs (US listed) are even more negatively impacted due to funds looking to decrease their net exposure through shorts and hedges. Generally speaking, regional lockdowns have minimal impact on GDS’s business as demand and usage for existing capacity remains strong (arguably even stronger under lockdown). The one exception was the prolonged lockdown of Shanghai and Beijing from April through May 2022. During this lockdown, it was impossible to have new cabinets installed by GDS’s customers into GDS DCs, which led to a meaningful slowdown in growth. Although the Shanghai and Beijing lockdowns had a tangible negative effect on GDS’s growth in 2022, we see this as a temporary blip in the greater trend towards digitization and growth in DC demand.

·         Strained China and US relationships over Taiwan were exacerbated when Nancy Pelosi (US House Speaker) made a visit to Taiwan in early August 2022. This negatively weighed on all Chinese stocks, but we do not see any specific fundamental impacts on GDS.

·         US restrictions on advanced technologies (i.e. limiting AI chips from NVDA and AMD) led to fears that GDS’s customers will be unable to secure servers to install in GDS DCs. This is a fallacy as the restrictions were strictly for miliary usage and only placed on the latest versions of these chips.[1] In addition, GDS’s customers have access to multiple other chip producers that can fulfil their demand requirements.

·         The prolonged power crisis in China has led to power tariffs hitting their ceiling of 20% for the year. When we look at GDS’s financials, power costs as a percentage of revenue historically have been around 25%; YTD it has skyrocketed closer to 30%. The power crisis began towards the end of 2021 from a lack of coal reserves, which led to China rationing coal usage, decreasing power output, and increasing power prices. Although the coal crisis has been averted, 2022 has not been kind in terms of climate; China was hit by its most severe heatwave (hottest and driest) in six decades. Lower power generation (lack of water for hydroelectricity plants) and higher demand (air conditioner usage) has led to severe power shortages across large swaths of China, to the point that air conditioner temperature limits were imposed in multiple regions. Although, GDS can recoup half of its additional power usage costs from its customers on a lagged basis, there remains a negative impact on GDS’s financials. In due time, we believe power costs will normalize and power tariffs will recede, but that will not happen in the near-term as China grapples with its stressed power infrastructure. This issue may have longer-term negative impacts on GDS’s earnings power and we have factored in these incremental expenses in subsequent valuation analyses for conservatism.

·         Cost of debt funding continues to decrease for GDS. The weighted average interest rate for GDS’s long-term borrowings declined from 7.4% in 2019 down to 5.5% in 2021 and continues the trend in 2022. The decline in borrowing rate is structural in nature and we expect GDS’s cost of debt to continue this downward trajectory albeit at a slower pace. As the DC industry matures, banks are increasingly more comfortable with the cash flow generation profiles of these assets. GDS being an asset heavy business with significant amounts of long-term debt will benefit significantly from declining borrowing costs.

·         End-customer demand has slowed. We have highlighted that the Chinese government went into regulatory overdrive starting from mid-2021 resulting in a significant slowdown in growth and confidence in China’s internet sector. This weakness is exemplified by Alicloud’s growth rate slowing from ~30% in 1H21 down to ~10% in 1H22. Alicloud is GDS’s largest customers, which has resulted in GDS’s overall revenue growth slowing from ~30% down to ~20%. That being said, GDS has been working directly with large internet companies and government enterprises, therefore its reliance on Alicloud continues to reduce meaningfully and is showing better resilience. We believe that the weakness in Alicloud (as a proxy for public cloud in China) is more temporary than permanent as the public cloud industry is in the process of digesting China’s regulatory changes and the mean reversion slowdown following the strong demand arising from Covid in 2020 and 2021. Going forward, a renewed acceleration in growth (after sector normalizes) from the public cloud sector alongside GDS’s expansion with government customers should lead to renewed strength in GDS’s growth rate.

Update on Risks

·         Risk #1: Supply vs. demand dynamics (improving)

o    Through our channel checks, we know that industry dynamics are getting better. After a flurry of new money (private equity, infrastructure funds, and conglomerates) entered the DC industry in 2020 and 2021, we have seen this pool of capital dry up with not much to show for from a results perspective (i.e. new operational DCs). Potential DC customers are skeptical with new industry entrants and would much rather partner with GDS who has the experience and network of available existing capacity and future inventory in all Tier-1 cities and surrounding regions.

·         Risk #2: Customer concentration (temporary weakness)

o    The slowdown from Alicloud and other public cloud providers had a negative impact on GDS’s growth rate. We believe this to be more temporary in nature and through GDS’s diversification strategy (SOEs, large enterprises, and overseas), we have seen GDS to be more resilient relative to the public cloud providers.

·         Risk #3: Operating cost structure (negative)

o    The increased power tariffs have negatively impacted GDS’s margins and we believe this may extend for a longer period of time. This has negatively impacted our view on the terminal value of GDS’s business and we have lowered our normalized margin expectations accordingly.

·         Risk #4: Balance sheet risk (perceived and misunderstood negative)

o    GDS is a highly levered company. It can take on this leverage because its customer contracts are on average ten years in length, which provides GDS with tremendous visibility into future revenue and cash flows. Within these ten-year contracts, the first two years are considered a grace period where the customer only pays for capacity it is utilizing. At the end of the two-year mark, the customer will have to pay the equivalent of full utilization (95%) no matter what the actual utilization rate is. Therefore, the only uncertainty to GDS is the customer’s move-in pace in those first two years. Generally speaking, GDS will only start to construct a DC after it has received meaningful commitments from its customers, resulting in a ~95% commitment rate for its DCs in-service and ~65% pre-commitment rate for its DCs under construction.

o    The issue that GDS faced this year is that customers who previously placed orders have slower move-in vs. prior expectations (which led to slower revenue growth for GDS and less upfront cash recovery) while new customers are still placing orders, which require GDS to break ground and start construction. Historically GDS was able to use its stock price and raise capital when it had a higher market valuation to support its expansion plans. This year, GDS’s valuation has seen significant pressure (see updated valuations below) and therefore it has lost one avenue to raise capital. Investors have started to doubt GDS’s ability to fund its growth. The reality is that GDS has multiple avenues to raise capital without diluting its shareholders. It has a RMB10bn project-based credit facility that is undrawn, it can partner with other infrastructure funds (such as GIC mentioned above) on a single project or multi-project basis, and it can raise capital directly at much better valuations for its new Southeast Asia DC growth opportunity as required.

o    In short, GDS’s strong visibility into its future cash flows and its ability to create unique capital raising avenues provide us with the comfort that it does not have any liquidity issues over its horizon.

Changes to Our Thesis

·         Thesis #1: Growth Opportunity (intact)

o    Our view on China’s demand for digitization, DC TAM expansion opportunities, and GDS market share gain potential has not changed.

·         Thesis #2: Moat (strengthened)

o    As mentioned above, the competitive environment has become more benign, which benefits GDS. In addition, all Tier-1 cities have limited their land and power quotas over the last few years, making GDS’s inventory far more valuable today vs. a few years ago.

·         Thesis #3: Strong Management (intact)

o    Our views around GDS’s management team’s vision, execution, and capital allocation capabilities have not changed.

·         Thesis #4: Valuation with Margin of Safety (improved)

o   Using the same valuation framework as above, we rolled forward the financial metrics from 3Q21 to 2Q22. In addition, we have lowered the run-rate EBITDA margins to account for the higher power tariff expenses. The end result is a valuation yield that is close to three times as wide vs. our entry point (21.5% vs. 8.1%). What we found interesting was that in a rising rates environment (rising in the US as rates in China are declining) the spread for US DCs has narrowed from LIBOR+3.9% in December 2021 to LIBOR+1.5% in 3Q22 (refer to chart below). Which means that although the US RFR has increased from 0.2% to 3.8%, the overall FCF yield for US DCs saw only a slight uptick from 4.1% to 5.3% – limiting a dramatic drop in the stock price. On the other hand, China’s RFR decreased from 2.5% to 1.7%, but GDS’s spread has increased from SHIBOR+5.6% to

SHIBOR+19.8% – more than three and a half times! Keeping in mind that even in this weak macro environment, GDS’s growth rate (arguably at trough) is more than double its US peers. When comparing GDS to its peers, we believe that GDS is wrongfully penalized for non-fundamental related reasons (outlined above) and that over time the market will see the normalized long-term value that we ascribe to GDS.

Catalyst

LT earnings compounding