21Vianet Group: Lynchpin Of China's Internet Industry

| About: 21Vianet Group (VNET)

A more readable PDF version of this report is available here.

The investing setups that generate the largest absolute returns generally involve making two calls that prove to be correct in the long term: a) early detection of a paradigm shift in a very large industry and b) identification of the company best-positioned to dominate that industry-wide transition.

Examples include Amazon (AMZN, now $305/share) at $10 after the Dot-bomb fallout, when its near monopolization of e-commerce was not so obvious … or Priceline (PCLN, now $1,013/share) at an unbelievable $11 after the 9/11-induced travel recession momentarily paused the destruction of the offline agency by the online self-service model … or Google (GOOG, now $868/share) at $100 post-IPO as online performance-based advertising was set to overtake display and other brand advertising … or Equinix (EQIX, now $166/share) at a mere $3 during a period of industry oversupply, before carrier neutrality became the gold standard in data center colocation.

As is the case with these precedents, we believe 21 Vianet Group ("VNET", or "the Company"), the leader in Chinese carrier-neutral data center services, is an archetypal franchise with durable and self-reinforcing moats that is positioned to dominate an industry at its inflection point in secular growth. The paradigm shift that China's Internet sector must navigate is a shift away from today's highly congested legacy network architecture, and we posit that VNET is the lynchpin to sustainable scale and efficiency. VNET's role as a key enabler of what we believe will be a renaissance period for China's Internet economy positions the Company as one of China's most important companies of any sector.

The market has not shared our bullish view, and only in the past quarter has sentiment begun to turn, largely due to the imminent (late Q3) monetization of VNET's cloud-based software partnership with Microsoft (NASDAQ:MSFT). Even after doubling from its all-time lows earlier this year, VNET still trades at a discount to its closing 4/21/2011 IPO price of $18.80 and is, as we will show, significantly undervalued relative to comparable but much slower-growth companies around the world in much more saturated end markets.

We believe VNET is too cheap largely because an appreciation of VNET's long-term outlook requires an understanding of several esoteric topics: a) how the Internet works at a physical level, b) how China's infrastructure differs from that of the US (and similarly developed markets), and c) how China's infrastructure is likely to evolve. We dedicated about half of this long report to providing a primer on these topics. The 12 sections can be read independently, so those familiar with any of the topics can skip them. At a minimum, we suggest reading sections I-II, "An Important Note on Utilization" of V, and VI-IX.

Any attempt to predict the exact outcome of a multi-year investment in a company as dynamic as VNET is an exercise in false precision. Therefore, we assign a short-term 12-month price target of $32 (~77% upside as of 10/10/2013). However, we believe that over the long term, VNET will compound equity at rates best described in multiples (as opposed to percentages) through a combination of earnings growth, multiple expansion, equity appreciation through cash flows (deleveraging and dividends), and the realization of various company-specific and industry-wide positive catalysts that abound well into the future.

I. Summary of Investment Thesis

  1. Opportunity to invest early, as industry demand accelerates, in archetypal wide-moat company that enjoys network effects, economies of scale, high customer switching costs, and pricing power
  2. "Arms dealer" to the Chinese Internet sector levered to every major secular growth trend in the world's largest (by usage) digital economy: growth of online data transmission and consumption, data center outsourcing, Internet usage, mobile and smartphone usage, e-commerce, online video, social networking, online gaming, and the pending monetization of cloud-based software and services
  3. Well-capitalized share gainer in a highly regulated industry with high entry barriers, uniquely positioned to consolidate the industry
  4. Attractive contract-based recurring revenue business model with high revenue and earnings visibility
  5. Attractive fundamentals:
  • Trading at sell-side consensus FY 2014 12.5x EV/EBITDA, despite an outlook for sustainable 30%+ growth with expanding margins
  • Our model calls for more robust sales growth and additional margin expansion, which values VNET at a forward 11.0x EBITDA multiple, at an absolute discount to comparable peers despite a three times faster relative growth rate
  • Healthiest balance sheet of comp set, with a positive net cash balance despite nearly doubling capacity in 2013 (direct public comps are 0.4-1.5x levered and expanding capacity much slower, if at all)

6. Opportunity to invest ahead of numerous catalysts, both imminent and longer-term

  • Imminent: continued capacity expansion to address pent-up demand, sustainable annual pricing increases, and ongoing rollout of higher-margin emerging businesses that include a content delivery network (CDN) growing at 60% per annum and a partnership with Microsoft to roll out cloud technology expected to generate $1 billion of bookings by 2016
  • Mid- to long-term: industry consolidation, effects of announced government stimulus, and modernization of Internet architecture that will allow for incremental, higher-margin services (in particular, commercialization of peering)

7. Uniquely experienced and well-connected management team that enjoys a strong relationship with the government

II. What Really Matters

Although we have prepared a comprehensive case for the substantial appreciation of VNET's shares, we believe Pareto's Principle (80% of the effects come from 20% of the causes) particularly applies to investing. Investment success usually boils down to making a correct call on just a few critical factors that drive the vast majority of the returns. This is important to keep in mind as we analyze the rapidly evolving Chinese Internet industry because of the confluence of multiple momentous structural forces at play. It is easy to lose sight of the forest for the trees.

Because we believe that VNET's leadership is firmly entrenched and growth is being driven by inexorable secular forces, the two factors that stand out as most important are:

  1. Continued success in expanding capacity profitably. Data center businesses are operationally intensive, require very high service levels, and demand ongoing capital investment. Since the Company's IPO in 2011, the management team has done an admirable job of rapidly and profitably building VNET's business and its key assets, and that level of execution must continue as the Company begins to reach a scale that rivals that of the largest data center companies in the world while navigating triple the annual growth of comparable global peers
  2. Continued success in extending offerings to keep pace with the unknowable technological innovations of the future in the data center industry. VNET has already extended its offerings beyond basic hosting and managed services to include content caching/delivery via the acquisition of the FastWeb CDN and (ii) China's first public cloud software offering via partnership with Microsoft. Staying ahead of the technology curve for future offerings will be important to long-term growth and customer lock-in

III. China's Internet Paradox

To describe the scale of China's Internet ecosystem, one must bandy about numbers that boggle the mind. China has only half the penetration of the US but is already 2.3x larger by users… poised to be 5x larger than the US at a similar level of industry maturity

Source: Kleiner Perkins Caufield Byers

China has only half the smartphone penetration of the US but is already 1.6x larger by users, representing a quarter of global smartphone subs (and growing very quickly).

Source: Kleiner Perkins Caufield Byers

The massive Chinese populace has a voracious appetite for Internet-related activities, outpacing the media time spent online of even US consumers…

Source: Kleiner Perkins Caufield Byers

… with a substantial amount of Internet media time upside as all but the most basic Internet activities are highly underpenetrated even among current Internet users (not to speak of the future users that make up the 58% of China where Internet usage has yet to penetrate).

Source: CNNIC, June 2013

Even China's most mature Internet industry, online gaming, after more than a decade of monetization, is still growing faster than the overall Internet sector in the US, and the three newest Internet industries to gain critical mass are experiencing rates of growth not seen in the US in over a decade:

Source: IResearch

The sheer enormity of China's Internet industry belies its full potential and how much growth still remains.

Source: CNNIC, World Bank, ITIJ, China Internet Watch

However, another set of numbers that is rarely discussed among investors describes the Chinese Internet ecosystem from a starkly different perspective:

Source: Akamai, CCINDEX

China, home of the world's largest and fastest-growing major Internet market, suffers connection speeds between 60-85% slower than the 10 Internet markets with leading connection speeds and more than half as slow as connection speed in the US despite an Internet population that is nearly three times larger and is consuming data at a growth rate four to five times faster.

What is more surprising is that this substantial lag in download speed in China comes after a nearly 26-fold improvement since 2011, when China's connection speed relative to the developed world was a mere 100.9 kbps!

Despite substantial ongoing investment from the government and the telecom industry, the speed of Internet access in China has slowed to a crawl, with only a 23% improvement to a national average of 2.6 mbps in the first half of 2013 from 2.1mbps in 2012.

As the inadequacy of China's Internet infrastructure chokes bandwidth supply, demand is exploding. Bandwidth use (as measured by petabytes per month of IP traffic use) is projected to grow at a 44% CAGR for the foreseeable future (as shown below), or double the rate of last year's quickly stalling improvement in connectivity speed.

Source: Cisco

Therein lies the most fundamental paradox of China's Internet industry: The world's largest Internet ecosystem is underpinned by one of the world's feeblest underlying infrastructures, just as demand for bandwidth, the lifeblood of the Internet, is inflecting towards historically unprecedented scale.

To state the obvious, China has some major structural problems to solve over the next few years.

IV. Background: History in the Remaking

Although at a virtual level, China's Internet industry has largely caught up and kept pace with the latest innovations of the Web in the US (e.g. search, social networking, mobile, local services, and to a lesser degree, cloud services), at a physical level, Chinese Internet is literally decades behind in its development. Despite the two countries' substantial political differences, we believe that the history of US Internet serves as a good guide to the future evolution of China's Internet infrastructure, since technology limitations and the best practices required to solve them have no boundaries.

The Internet is a collection of thousands of disparate and independently managed physical networks that interconnect to form a network of networks. This mega-network serves as the plumbing for the World Wide Web (the "Web"), which is itself a network of millions of websites that link to each other. Think of the Web as the browser- and, increasingly, smartphone-based user interface to the Internet and all its content. Websites and all digital content delivered via the Internet are stored on physical servers hosted in data centers that connect to the underground fiber-optic networks that comprise the structural backbone of the Internet. The rate at which these fiber-optic networks transmit data over the Internet Protocol (NYSE:IP) is called bandwidth, and bandwidth capacity is a function of both the underlying network technology as well as how that technology is configured (its network topology).

The much-ballyhooed term "cloud" is a virtual concept that refers to Web-based services. But in physical terms, the "cloud" is nothing new; it is just jargon for distributed computing, or the use of a network of physical servers that share the computing load required to offer Web services such as storage on Dropbox or software from Salesforce.com.

At a virtual level, the Internet's physical interconnections go unnoticed as people browse the Web or use a Web service, although occasional bouts of latency or downtime lend clues to the complexity of the underlying physical networks.

At a physical level, these independent networks of networks connect to each other at mission-critical telecommunication facilities called network access points, or NAPs, to form a network backbone that underpins all Internet communication. It is no exaggeration to state that the health of the entire Internet economy relies on the efficiency, scalability, security, and democracy of this backbone-so much so that every major government in the world regulates its use.

In the early days of the public Internet in the US, backbone network providers (at the time, just the major telcos, AKA "carriers", such as AT&T) exchanged their network traffic at government-controlled NAPs. As demand for Internet bandwidth exploded with the popularity of the Web, and the complexities of managing the Internet backbone overwhelmed the government's capabilities, the government privatized the operations of the Internet, transferring the NAPs to commercial providers such as AT&T and Verizon, thereby entrusting the continued development of the Internet's infrastructure to the forces of a free market for bandwidth. The term "NAP" was replaced with the term "Internet exchange point", or IXP, as various commercial services beyond simple network access were introduced by IXP operators. A watchful eye towards antitrust violations and national security matters is all that remains of the government's previously tighter governance of the underlying Internet infrastructure.

The commercialization of the US Internet led to two key innovations that allowed the modern US backbone to scale commensurately with Internet traffic growth: commercial peering and carrier-neutral colocation.

Peering and Interconnection

In today's US Internet ecosystem, there are hundreds of Internet service providers (ISPs) of various types that interconnect at various IXPs, ranging from the carriers/telcos that used to monopolize Internet access in the past (e.g. AT&T) to companies primarily focused on Internet service (e.g. Level 3 Communications) to specialists that focus on specific elements of Internet service such as hosting websites (e.g. GoDaddy). The economics of these interconnections are governed by contracts called peering arrangements.

A peering arrangement is an agreement among similarly-sized networks (peers) to share their networks by interconnecting them directly to speed the delivery of each other's traffic across the broader network. Commercial peering came about as Internet usage eventually got to the point that there was no single ISP that could possibly serve its demand through its network alone. The largest peers, usually backbone network ISPs, are labeled "Tier 1" and exchange their traffic for free, while smaller ISPs and other participants, labeled Tier 2 or 3, usually pay for "transit." The scalability achieved through the network effect of peering has allowed US Internet traffic to grow by literally four orders of magnitude from the mid-90s to today. Peering, by making data transmission across networks much more efficient, also substantially lowers bandwidth consumption and therefore the cost of Internet access.

The diagram below illustrates the enormous efficiency gains of moving a fully connected legacy "mesh" network (left) to a peering configuration (right):

Source: ix.kinx.net

Peering, which used to be the exclusive domain of the telco oligopoly and occurred at the government-run NAPs of the past, now occurs across various IXPs throughout the country (and, increasingly, the world), the largest of which are owned and commercially operated by companies such as EQIX.

EQIX and select peers who run IXPs as service offerings incremental to their colocation services (outsourced data center infrastructure and operations) have added further efficiency to the Internet value chain by introducing peering at the individual enterprise level for customers hosted in the same data center facility. Individual companies-e.g. a securities exchange and a sell-side firm, or Netflix and a network with streaming movie viewers-can now directly connect their servers to further speed data transmission, bypassing the public Internet altogether. These customer-to-customer connection services, called interconnections or cross-connects, speed connectivity to such a degree that even companies large enough to economically justify building and running their own data centers (such as Google) maintain presence in EQIX's data centers for interconnection purposes.

Carrier-Neutral (AKA Network-Neutral) Data Centers

A carrier-neutral data center is operated by a third party that is unaffiliated with the ISPs that provide Internet access within that data center. Examples of such US companies include EQIX, Interxion (NYSE:INXN), and Cyrus One (NASDAQ:CONE).

Colocation (the outsourcing of data center infrastructure and networking, and increasingly, operations and network management) at a carrier-neutral data center has several benefits: ISPs have to compete with each other for business, which results in optimal pricing to the end customer; (ii) ISP redundancy mitigates the risk of a single ISP failure; and a carrier-neutral network allows for the flexibility of switching ISPs easily, without costly and time-consuming physical data center moves, in the event that a particular ISP no longer satisfies a customer's needs. The advent of the carrier-neutral data center lowered the cost of Internet access while improving its quality (notably, speed and reliability).

Colocation at a carrier-neutral data center that provides interconnection services has become a best practice of data center management in the US. As the trend of outsourcing data centers continues globally, the largest carrier-neutral data centers that also run IXPs are uniquely positioned to dominate due to the inherent network effects of their offerings-the more nodes in their data center network, the more potential nodes in their exchanges, which improves network efficiencies, which begets more business from the largest customers, which incentivizes smaller customers in the larger companies' ecosystem to colocate and join the exchanges… and so goes the virtuous cycle.

China's Internet Architecture

The highly inefficient state-run architecture of the Chinese Internet somewhat resembles that of the US circa 1995, before its full commercialization.

Two state-owned enterprises (SOEs), carriers China Unicom and China Telecom, monopolize commercial Internet service in the north and south of China, respectively, as if an arbitrary geographical line that bisects China into hemispheres was drawn to divide each carrier's share of the market. Each carrier competitively dominates its respective market and has no meaningful presence outside it, where its service level degrades enormously and it is highly dependent on the other for access. Tier 2 and 3 ISPs exist, but their cumulative market share is small, and many of the ISPs are only local operators without a regional or national footprint.

Although in theory, having an oligopoly of just two Tier 1 ISPs administering bandwidth nationwide should simplify the ISP industry, in practice, the opposite is true. China Unicom and China Telecom act like holding companies, each with independently operated regional divisions in each of the 31 mainland Chinese provinces. There is not much centralized management at either telco. Every local office has its own management team, its own P&L, and administers access within each province autonomously, all of which makes nationwide change difficult and significantly complicates the picture for customers (more on this later as we discuss VNET's competition).

The provincial subsidiaries of China Unicom and China Telecom achieve national reach by interconnecting at just three state-operated NAPs located in three Tier 1 cities (Beijing, Shanghai, Guangzhou). As was once the case in the US, these NAPs severely bottleneck Internet traffic nationwide and are constantly and increasingly overloaded. It gets worse: Even within each carriers' network, limited inter-provincial interconnectivity bogs down traffic (the vast majority of traffic is inter-provincial, since most Chinese Internet companies are located in Beijing, Shanghai, or the Shenzhen/Guangzhou region), unnecessarily overloading the three NAPs tasked with brokering even inter-provincial traffic among neighboring provinces for the world's most populous nation.

Source: 21 Vianet corporate presentation

The only alternative to the overloaded state-run NAPs is VNET's private NAP, which is a major source of competitive differentiation for the Company. Over a period of more than 10 years, with a total investment of over $250 million, VNET built its own private backbone network (the 21 Vianet Inter-Cloud Express Network) that interconnects VNET's physical operations across 43 cities to the hundreds of networks of the state-run telcos and every major ISP in China, effectively making each VNET data center a private NAP.

Although no official industry figures for network efficiency are available in China, during our due diligence, we found that VNET customers such as Google report a three-fold or better improvement in connection speed during peak traffic hours when using VNET's superior network architecture.

Commercial peering, that is, peering beyond the interconnectivity between the state-owned telcos that monopolize the provision of bandwidth in China and select Tier 2/3 players, is not allowed by the government at this time. VNET has the technical ability to offer interconnection services for a fee but is currently legally unable to cross-connect customers directly and generates no revenue from peering. Although our investment thesis does not depend on the commercialization of peering technology in China, we believe it is highly likely to be approved by the government in the future to help address China's extreme network congestion. That would provide a major catalyst to accelerate VNET's top-line growth and margin expansion even further due to the very high-margin nature of the offering. ~19% of EQIX's 2013 revenue in the US will come from offering interconnection services (source: Jefferies: "Data Centers - A Global Perspective", p. 10).


Unsurprisingly, China has a highly protectionist and restrictive set of regulations for the ISP and data center industries. The regulatory hurdles to establishing a network in China severely restrict domestic competition and completely restrict foreign entrants.

Foreign companies are not allowed to obtain either of two major licenses required to operate a competitor to VNET: Internet data center (Pending:IDC) licenses and ISP licenses. Investment by foreigners is restricted to non-controlling stakes of less than 50%. These restrictions effectively eliminate global data center leaders from the picture in China, and we do not see this changing anytime.

The only foreign company that has ever managed a data center in China is EQIX, which via its July 9, 2012 acquisition of Asia Tone, a Hong Kong-based Asian data center company, currently has one small data center with a few hundred cabinets of capacity in Shanghai (VNET will end the year with ~20,000 cabinets nationwide). Since Hong Kong is a special administrative region that is technically governed by China but is not actually in China, where local laws apply, EQIX's acquisition was deemed legal. However, it is unlikely that the now EQIX-owned Asia Tone will be able to expand its presence in mainland China. Moreover, it is noteworthy that this Shanghai-based data center is actually a partnered data center, i.e. not self-built, but rather, leased from one of the local telcos' wholesale data centers. The Chinese industry insiders with whom we have consulted believe it is more likely that Asia Tone's licenses in China will not be renewed after expiry than it is for the company to be allowed to expand its Chinese operations.

If anything, after the Snowden fiasco, foreign participation in the Chinese market is likely to be scrutinized even further. This is significant not only when considering China's domestic market, but also has implications for inbound international traffic as China becomes an increasingly powerful force in the global Internet economy. Domestic leaders such as VNET stand to benefit exclusively from those trends.

Within China, the process of gaining and then maintaining licensure is onerous. In order to compete with VNET, a domestic company must not only gain both an IDC license and an ISP license, but the ISP license must be the broader-scoped inter-provincial license, which is much harder to attain, as opposed to the regional version. Licensed companies must pass the Ministry of Industry and Information Technology's (MIIT) annual inspection to stay licensed. Licenses expire after five years, after which they must be renewed.

In 2000, as the industry was still developing, more than 100 inter-provincial ISP licenses were issued. Today, only seven or eight companies that can operate inter-provincial networks remain. As for IDC licensure, no license has been issued since 2007. There is a rumor that Alibaba (an e-commerce company) and Huawei (the "Cisco of China") obtained licenses this year, but if true, those licenses were likely issued for the operation of IDCs for their own use. Both Alibaba and Huawei are major VNET customers, and a large portion of VNET's 2013 capacity expansion was pre-sold to Alibaba as it prepares for its highly anticipated IPO.

Regulation is not merely a barrier to entry into the ISP or IDC industry; it is also a substantial barrier to replication of VNET's Inter-Cloud Express Network (VNET's private NAP). In China, a private enterprise or individual cannot legally own or hold title to real estate assets, which include the exclusively state-owned underlying fiber-optic Internet backbone networks. Therefore, VNET negotiated long-term (20-year) leases from the state (e.g. China Unicom, China Telecom, and various smaller state-owned networks such as the national Education Network and the People's Liberation Army Network), and then built over 450 physical points of presence (PoPs) located in over 100 cities to interconnect their leased networks. The sheer time and government-level relationship capital required to replicate such an effort, let alone the access to project financing in the hundreds of millions of USDs required, makes the introduction of a competitive offering in the foreseeable future highly unlikely, if not altogether impossible.

V. US vs. China: At Different Points of Industry Transition

Foreign investors who are familiar with VNET often refer to it as "the EQIX of China" or, more recently, after its tie-up with Microsoft to launch cloud computing services this year, as "the Rackspace (NYSE:RAX) of China." Such comparisons are only valid to the extent that investors can navigate the alphabet soup of industry terms used to describe various data center business models and understand the key differences in China, where data center market dynamics are substantially different.

The On-Demand Model and the Progression of Data Center Services in the US

The following diagram illustrates the modern data center technology stack from which companies can choose to outsource various parts to third party service providers such as EQIX, RAX, et al.

Source: Gartner

At the most basic level, there is differentiation between the two hosting models. In the colocation business model, customers bring their own servers and networking equipment into an outsourced data center, while the fully managed hosting business model provides a turnkey solution in which the procurement, setup, operation, and maintenance of hardware and other technology capital (from servers and networking equipment to licenses for software to manage networks and systems) is also outsourced. Colocation customer mix skews to larger enterprises that require bespoke data center implementations and tend to purchase data center space "cabinets" or "cages" at a time (a cabinet encloses a rack in which servers and other hardware are "stacked", and cages are secured rooms that house multiple cabinets). Managed hosting customer mix skews to small and medium enterprises (SMEs) with less demanding implementations deployed on pooled infrastructure that can be shared by multiple customers. One key difference between the two models is the higher level of capital intensity of the managed hosting model. In today's world of higher specialization, managed hosting companies, which are focused on the value-added service element of data center hosting and the software elements of the technology stack, sometimes use third-party data center facilities from wholesalers, sometimes referred to as data center REITs, instead of their own.

In both hosting models, the data center company provides all the connectivity, networking, power, cooling, security, and other facility management services to operate a data center. Customers can choose from a menu of services and products offered atop the basic hosting services with varying degrees of direct involvement (from self-managed to fully outsourced). Many of these services are billed like a utility, on an on-demand basis, as the services are used (as opposed to billing for everything up front).

The on-demand, or "as-a-service", model allows customers to turn otherwise fixed infrastructure investments into variable costs with the flexibility to scale infrastructure up or down as necessary without the up-front capex and the long lead times required by the in-sourced/in-house approach. This economic benefit alone, even without the various practical and operational benefits, is one of the key drivers of the global trend towards outsourcing data center operations.

Over the past decade, technology business models evolved to favor on-demand/as-a-service models not only for physical infrastructure and network services but also, increasingly, for software. The largest modern service providers now offer some combination of infrastructure with value-added services related to the software that runs on top of it, which ranges from virtualization software that allows for the more efficient utilization of the hardware layer to the underlying operating system and middleware layers that allow end-user applications to interface with the hardware that powers them. Terms such as infrastructure as a service (IaaS), platform as a service (NASDAQ:PAAS), and software as a service (NASDAQ:SAAS) refer to the business models by which different parts of the technology stack are outsourced on an on-demand basis to a company that operates its services on "the cloud", a catch-all term that refers to distributed infrastructure that delivers computing over the Internet (as opposed to local networks).

In that sense, EQIX, RAX, and their data center peers who were once simply data center infrastructure companies are now rightfully considered on-demand cloud companies that are levered to the infrastructure outsourcing and cloud computing trends. EQIX is the preeminent colocation company while RAX is the largest managed hosting company.

In the US and other developed markets, there has been a trend within each of the various segments of the data center industry (wholesalers/REITs, colocation players, managed hosting players, etc.) to move further downstream as the core data center market becomes saturated. Increasingly, wholesalers are starting to bid for retail projects, lowering their contract size requirements in an effort to move downstream; colocation providers are moving up the technology stack to offer various on-demand products, with leaders such as EQIX differentiating through value-added services based on their unique network architecture (such as interconnection services); and managed hosting service providers are realizing a larger share of their total revenue from cloud offerings.

The charts below illustrate the stark contrast in end market demand growth between basic infrastructure offerings and on-demand services delivered over the cloud in the US:

The on-demand model has clear benefits for the service provider as well. For one, it is inherently a contract-based recurring revenue business model with high revenue and earnings visibility and predictability. Also, it has similarities to the movie theater or baseball park model of cross- and up-selling or packaging incremental value-added products with the core data center hosting offering to realize incremental revenue growth, margin expansion, and operating leverage from data center assets.

VNET is more like EQIX by core hosting model, with the majority of its revenue coming from larger colocated customers, although it does realize a minority of its revenue from a turnkey solution it calls its "container data center" service that has elements of RAX's core business as well. VNET expects to ramp up this turnkey product in the future to capture burgeoning demand from SMEs. As is the case with EQIX, VNET's investments in its own data center and network infrastructure are highly differentiating, much more so for VNET in China than for EQIX in the US. VNET's proprietary network is the only alternative to the state-run networks that everyone else must use, while there are multiple alternatives to EQIX.

Comparisons to EQIX and RAX fall short beyond these basic points, however, because China's industry is at an early stage of its long-term development while the US data center market is crowded and well-penetrated, with a much slower growth rate being only partially offset by higher-growth cloud services that don't even exist in China yet.

An Important Note on Utilization

To start a discussion of the differences between China and the US, it is instructive to analyze the growth path of the two local market leaders EQIX and VNET, which serve as good proxies for their respective market's supply and demand dynamics.

One metric that US investors tend to compare for both companies is the utilization rate of data center inventory as measured by billable cabinets. Without question, utilization is an important operating metric to track because it correlates to demand over the long term. However, there should be an enormous difference between the quarterly utilization rate of a data center company that is nearly doubling capacity year over year to meet pent-up demand and another that is growing inventory at single-digit rates. Comparing utilization between such companies is absurd, but that is how the market has been analyzing and valuing VNET since its IPO.

Furthermore, we have yet to see such comparisons being made with the same measure of utilization for each company, so the comparisons are not even based on the same metric.

There are two principal ways in which data center companies measure their overall utilization. The first method, which we will call the "cabinet method", simply divides the number of sold (billable) cabinets by the total number of cabinets in operation (total cabinet availability). VNET has always used the straightforward "cabinet method" when reporting utilization, so investors get an exact sense of how much of its physical real estate assets is generating revenue.

The second method, which we will call the "power method", eschews calculations based on absolute cabinet inventory in favor of power-adjusted availability. A data center has a limit on how much power it can generate to power the cabinets it hosts, so increases in power density requirements by customers will result in a decrease in the total number of cabinets that can be powered and, therefore, be sold. The "power method" of calculating utilization adjusts cabinet by making power, and not physical data center space, the limiting factor in the calculation. Excluding power limitations from the calculation and reporting utilization based on the "cabinet method" has historically been the industry standard, because a power capacity can be more readily increased (by installing a more advanced power distribution units) than physical data center space, which is also much costlier to add.

EQIX is the most prominent data center company that uses the power method when reporting utilization, so investors no longer get an exact sense of how much of their physical real estate assets are generating revenue. Prior to 2007, EQIX too reported utilization via the "cabinet method", but a close look at their 10-Ks shows that EQIX began using the "power method" in 2008. EQIX's 2007 10-K reported 2007 utilization as 58%, but its 2008 10-K reported it as 73%.

Our intention is not to condemn one method over the other; rather, we simply note that if the misguided comparison of utilization rates between a high-growth emerging market company and a low-growth developed market company is to be made, it should at least be made using the same metric.

Nevertheless, even by the misguided use of two different standards, VNET's cabinet-based utilization of 70.2% in the 2Q 2013 is not far from EQIX's power-based utilization of 75.2% as VNET ramps inventory at 70% per annum while EQIX's YoY capacity growth has been in the teens since 2009 and will drop to single digits in 2013.

Despite substantial data center acquisitions in 2007 (IXEurope and Virtu) and 2010 (Switch & Data) and various 2008-2010 tuck-ins which skew the figures favorably (especially in 2008) when compared to VNET's organic cabinet growth since inception, the inverse correlation between aggressive organic cabinet growth and utilization rate is not hard to see. Another relevant data point is the 78.4% utilization of European leader Telecity Group (OTC:TLCTF), a local data center leader that will grow capacity by mid-single digits this year.

In the context of VNET's rapid inventory expansion, a far more important metric is the ROI of incremental investment in data center or cabinet expansion, which for VNET is north of 30%, an outstanding figure. Another important metric that must be considered alongside utilization is how quickly new cabinets are sold to steady-state ~80% utilization. VNET has been filling deployed cabinets at an excellent pace despite substantially ramping growth over the past two years.

State of the Chinese Data Center Industry

Why is VNET able to grow profitably at such a rapid pace without significantly compromising utilization, something even EQIX, considered the paragon of the data center industry, was unable to do? We attribute the phenomenon to several key differences between China and the US:

  1. China's overall industry growth is unlike anything the developed world has seen in over a decade. Comparisons between VNET and EQIX circa the mid-2000s are more appropriate than the comparisons between the two companies today, where most investors seem to be stuck

2. The unique infrastructure challenges that are flaring up in China as bandwidth consumption explodes have created heretofore globally unprecedented demand

3. Despite record demand, the Chinese data center industry is in a constant state of supply constraint. China's capital markets are as underdeveloped as its underlying Internet infrastructure, and the resultant limited access to expansion capital has acted as a governor to data center expansion by private companies. During the same high-growth period in the US, the Internet bubble-driven glut of dumb money led to massive data center oversupply throughout the early to mid-2000s

4. The stringent regulatory hurdles that act as significant barriers to entry in China were not present in the US. To the contrary, well past its days of reckoning, the US market is still crowded and competitive, while China's market is dominated by the state-owned telcos and literally a handful of private companies led by VNET, which stands to benefit disproportionately with its growing network effects as it continues to gain scale and share

Can the Chinese market's fast-paced growth and VNET's profitable expansion continue?

We believe the Chinese data center industry is poised to grow at double digit rates for a decade and likely longer, increasingly outpacing global growth in the next several years as sales accelerate in China while they slow in the developed world.

Consider the following:

  • The US Internet industry is powered in part by approximately 1.5 million outsourced data center cabinets serving 254 million domestic users, but China has a mere ~150 thousand outsourced data center cabinets serving an Internet population that is almost 2.5x times as large (591 million)
  • This huge gap is even more pronounced when China's broadband speed-between three to seven times slower than that of developed countries-is considered. Small increases in bandwidth throughput as China's infrastructure constraints are resolved will yield large, incremental increases in usage and data center demand
  • China has a very young Internet demographic with higher bandwidth consumption tendencies than US users
  1. The average Chinese Internet user is only 25 years old, while the average US Internet user is 42 years old, and spends 21 hours online (source: China Internet Network Information Centre, Jan 2013)
  2. 77% of Chinese users stream online music, 66% stream online video, and 66% use social media, all top five activities behind text messaging and online search (source: China Internet Network Information Centre, Jan 2013)
  • The charts below illustrate China's sustainable runway for accelerated growth from a demographic perspective

Data center capacity in China must grow by a factor of 10x to catch up to today's level of penetration in the US

When combined, the US, the UK, Germany, and Japan (four of the most developed data center markets in the world) only have 43% of China's population but 25x China's data center capacity

  • China has yet to begin large-scale monetization of the cloud, which will serve as a major growth catalyst for the industry at large and VNET in particular

The only data center market in the world that will even come close to approximating the scale of the Chinese market is India's, but that market is still years away from inflection, hampered by infrastructure problems far more excessive than China's which will take many more years to resolve. The scale of the commercial opportunity in front of VNET is truly once-in-a-lifetime.

VI. VNET's Dominant Competitive Position

Important Players in China's Data Center Value Chain

The industry value chains of the US and China are comprised of the same roles, with data center wholesalers and retail colocation providers that operate their own infrastructure on one end, retail customers on the other, and various ecosystem companies that serve both ends in the middle. However, China's competitive environment is much simpler to analyze because its industry is generally less fragmented and, for now, devoid of the added complexity that cloud-based services introduce. Two key factors that limit competition and thereby also simplify the industry's major constituents are limited access to expansion capital and the government's parsimonious licensure policy.

In China, there are no pure-play wholesalers of note. With some small regional players excepted, generally, the telcos are the only companies in the data center wholesale business, although they act as both wholesalers and retailers at once. The telcos have data center customers they service directly, but they lease their directly owned data center facilities to a large majority of carrier-neutral data center providers who then resell to end customers.

On the retail side, there are only a handful of companies that operate their own data centers. VNET is the only large-scale provider that has self-built a majority of its cabinet inventory, with over 60% of its total inventory self-built and the rest leased from the telcos as of 2Q13. By the end of this year, 80% of its cabinets will be self-built. Other notable data center companies with a retail model include:

  1. China Netcenter ("CNC"), a public company listed in the A-share market, that derives less than 30% of its total revenue from a few hundred cabinets of inventory leased from the telcos and whose main business is its CDN service (CNC is more like American CDN leader Akamai (NASDAQ:AKAM) than like EQIX), which generates ~60% of total revenue
  2. Chengdu Doctor Peng Telecom and Media Group ("Dr. Peng"), also an A-share company, that is primarily a data center contractor that builds data centers for the telcos and derives a minority of its revenue from an inventory of approximately 4,500 cabinets, some of which are self-built

Currently, only VNET is meaningfully expanding carrier-neutral data center inventory because demand for data centers is so high in China that the telcos have no more wholesale inventory to lease.

After data center leasing and management, the next largest segment of revenue in China's data center industry is bandwidth reselling. The scarcity of bandwidth in China coupled with various service-related challenges in managing a direct billing relationship with state-run telcos make bandwidth reselling a lucrative, though lower-margin business.

The next largest value-added service segment is content delivery services. While pure-play data center companies try to optimize content delivery at the physical network level, using interconnected architectures and routing technology, CDN operators optimize delivery at the virtual (content and application) level. Like AKAM in the US, CDN operators build a network of servers (physical "nodes" that replicate the content nationwide for easier local access) running proprietary content optimization technology across third-party data centers throughout the country. Technologies such as local content caching (replication of content across a network to provide multiple points of quick access) and request routing (similar to VNET's smart routing, but in reverse, optimally routing requests from end users of content through the path of least resistance to the closest content nodes) can meaningfully reduce delivery speed and network load.

The largest CDN is owned by ChinaCache (CCIH), which competes with CNC and Dr. Peng, both diversified data center companies, as well as with pure-play CDN players Fastweb (acquired by VNET in 2012) and private company Dilian.

Defining VNET's Competitive Set

Nominally, VNET competes with telcos China Telecom, China Unicom, and to a lesser degree China Mobile, as well as various smaller data center companies, some of which also offer carrier-neutral facilities. Beyond its core data center offering, VNET also competes with value-added specialists such as CDN operators and bandwidth resellers. Ultimately, it can be argued that VNET also competes with its own customers, some of which run their own data centers for their own use and therefore have the option to bring their outsourced operations in-house.

From that perspective, which amalgamates the entire data center value chain into a single competitive set, market share as of year-end 2011 is divided as follows (more recent industry data is not available):

Source: IDC

Carrier-neutral data centers are gaining share as a class, and within that class, VNET is gaining share most rapidly. Research firm IDC reported VNET's market share as 5% in 2009 and 7.5% in 2010. Early estimates by VNET management peg the Company's share in 2012 at approximately 13%.

A narrower way to define VNET's competitive set is to restrict the analysis to carrier-neutral data center companies. After all, VNET does not directly compete with the telcos for the increasingly large universe of customers specifically looking for carrier-neutral facilities, and value-added specialists only come into the competitive picture once a data center provider has already been selected. Within its segment, VNET is the dominant leader in a fragmented market that consists entirely of smaller regional or local competitors without VNET's national footprint. Again, the chart below is as of the end of 2011.

Source: IDC

All of these market share numbers are deceiving, though, because the majority of Chinese cabinet inventory is leased from the telcos themselves, and a significant portion of the revenues from the smaller competitors such as CNC and Dr. Peng are generated from businesses other than hosting and managed services. The salient point is that irrespective of how its competitive set is defined, VNET is the share-gaining leader in the fastest-growing segment of the overall market, carrier-neutral data center services.

The most important way to define VNET's competitive set is to review how customers buy data center services and then consider who can offer an alternative. VNET's customers are generally Internet or technology-enabled companies running mission-critical Internet systems whose decision criteria prioritize quality of service, network connectivity, and solution scalability over price. By the criteria its customers consider most important, VNET has no competition.

Competitive Differentiation

Although VNET is a fairly new public company unfamiliar to most foreign investors, having listed on NASDAQ on April 20, 2011, the Company was established 16 years ago and is one of the most recognized and respected companies in the Chinese telecommunications sector, widely acknowledged as the leader in the data center service industry. In our interviews with current and potential customers or partners, both Chinese companies and American multinationals, it was common to hear VNET described as "head and shoulders above anything else available" and "the only company that offers service at an international standard in China."

VNET is unique among its competitors in three categories:

1. Network architecture

  • VNET's unique backbone network is the only marked improvement to the state-run NAPs

a) As mentioned previously, the 21 Vianet Inter-Cloud Express Network offers the only alternative to the three overloaded state-run NAPs

b) Threefold or better increases in network speed can be realized by customers who access VNET's backbone by colocating in a VNET data center, any of which serves as a private NAP to China's Internet backbone.

  • VNET uses BroadEx, proprietary smart-routing technology to further optimize its network

a) BroadEx acts like an intelligent switchboard that constantly monitors the VNET network for the fastest, most reliable, and most secure routes through which to direct traffic

b) BroadEx optimizes across both VNET's own and others' interconnected networks

  • VNET has the only carrier-neutral footprint with nation-wide coverage

a) Other carrier-neutral competitors mainly operate intra-provincial or limited regional networks

2. Data center quality

  • International ratings for data center operations range from Tier 1 (lowest) to Tier 4 (highest) and are based primarily on connectivity and power uptime (we are simplifying for brevity).
  • Tier 4 demands what is referred to as "four nines" of uptime, i.e. 99.995% uptime. Tier 3 demands "three nines" (99.982%). Tiers 1 and 2 both require "two nines" (99.671% and 99.741%, respectively). Tier 4 data centers are so expensive and operationally difficult to build and maintain that even the world's largest data center companies only offer them selectively, and some industry insiders consider the standard to be aspirational in practice despite the claims of the most elite service providers.
  • VNET, whose service levels guarantee three nines of network uptime and four nines of power uptime, has the only Tier 3 facilities in China, since the state-run NAPs themselves may not even qualify as Tier 1.
  • Of note, VNET is the first Chinese company to achieve ISO 9002 quality system certification by the American Registrar Accreditation Board and a certification by the United Kingdom Accreditation Service.
  • At the cabinet level, an important measure of capacity is power density, which is most often measured in kVA per rack. Best-of-breed international data centers are now building at a 6+ kVA per rack standard. Most Chinese data centers, which are either offered directly by the telcos or by wholesalers of telco inventory, build cabinets with power density of 2.2-2.3kVA per rack. VNET's data centers are built to support 5-6kVA per rack.

3. Service

VNET's best-in-class service warrants special mention, because the Company's competitive differentiation is generally believed to be solely technical in nature and is therefore underestimated by investors.

The trend towards outsourcing by Chinese companies is driven as much by economic and technology-based efficiencies as by the operational complexities of managing data center and network operations directly. Comparisons of VNET to foreign market leaders are inadequate because they fail to describe VNET's crucial role in addressing these local market complexities that arise from myriad infrastructure-related, regulatory, and other operational issues. One of VNET's most important roles is its ability to eliminate or dramatically reduce such complexities for its customers.

Specifically, VNET uniquely addresses important local market complexities in the following ways:

1. Bandwidth procurement

  • Like EQIX, VNET procures bandwidth from carriers and resells it to customers as part of its holistic service offering. However, unlike EQIX's customers, who overwhelmingly opt to deal with ISPs directly, the vast majority of VNET's customers buy bandwidth through VNET. The alternative would require that a separate contract, service level agreement, and billing arrangement be negotiated and maintained with each local branch of each ISP across multiple cities in each of China's 31 provinces.
  • Furthermore, the availability of bandwidth in China, given its scarcity, is never a guarantee, especially during peak hours.
  • Through the collective bargaining power of its large customer base, VNET abstracts all this complexity by purchasing bandwidth at favorable rates with enough headroom to operate with minimal disruption even during peak hours.

2. Centralized network management

  • Dealing with the highly fragmented corporate structure of the state-run telcos is not just a problem when setting up a data center and procuring its bandwidth. At least those issues are relatively non-recurring.
  • Much worse is the problem of managing a national network from support and service standpoints. In more developed countries, an ISP typically has a national network operating center (NYSE:NOC) to which severe problems can be escalated for investigation and resolution. At a minimum, regional NOCs have access to an overarching national NOC to help with emergencies even if customers may not directly be able to communicate at the national level.
  • In China, doing business with telcos directly means being serviced by government employees in a hierarchically fragmented organization with no centralized management. It is not uncommon to find that local branch employees are unavailable after 5 or 6 PM.
  • VNET offers the only 24x7x365 nation-wide NOC in the industry and abstracts the complexities of dealing directly with the ISPs.

3. Unique data center expertise

  • Recognition for VNET's expertise is evident in the special customer and partner projects in which the Company occasionally engages to maintain its most important ecosystem relationships
  • For example, VNET currently manages six data centers for China Telecom and China Unicom
  • VNET has also been asked by its largest customers, notably Tencent and BIDU, for assistance in building their own data centers
  • An important precedent occurred in late 2011 and early 2012, when Tencent, China's largest Internet company, tried to bring in-house its most critical systems, which had been historically outsourced to VNET, only to quickly return as a customer after failing

Demand for VNET's best-in-class service is so high that customers have historically asked the Company to lease lower-quality cabinet space from the telcos while it builds out its own network of data centers nationwide. Because VNET's expansion strategy is focused on dominating Tier 1 and larger Tier 2 cities first, given the explosion in demand in those regions (especially in Beijing, the "Silicon Valley of China"), larger customers who want to be colocated in smaller Tier 2 and Tier 3 cities will actually pay a premium to host their servers in VNET-managed wholesale space rather than lease the space and manage it themselves.

In summary, VNET is able to offer network speed, reliability, and redundancy that is unmatched in China with best-in-class service levels driven as much by the Company's unique technology assets as by its business process management and service responsiveness.

The Government: Friend or Foe?

It is worth spending a few moments examining the government's role given its heavy influence on the industry. We believe foreign investors generally misunderstand the government's overall agenda.

The government is involved at multiple levels:

  • All telcos are state-owned enterprises (SOEs) that exclusively provide bandwidth and, together, have largest share of the overall data center cabinets deployed in China.
  • The government is the only entity that actually owns and leases the underlying fiber-optic infrastructure of China's Internet ecosystem.
  • Licensure is strictly controlled by the MIIT, the government arm that governs technology policy.
  • The MIIT is ultimately responsible for resolving China's network congestion problems before they cause a national crisis.
  • The MIIT, in support of the overarching China State Council, also has the important job of administering stimulus to the Chinese technology sector, which is a key sector of focus in the new Chinese administration's domestic consumption-focused Five Year Plan.
  • Expansion capital, though limited, has generally been supplied to data center companies by state-owned banks.
  • The Chinese telecom sector is one of the most heavily regulated, and the government will likely continue to strictly control participation to a limited number of private companies.

On the one hand, the government that exclusively controls bandwidth and real estate assets also competes with private sector companies, stringently controlling entry into the market. More generally, the government appears to be reining in bank lending to private companies in an effort to control liquidity.

On the other hand, the government is keenly aware of the unsustainability of its current Internet backbone and clearly understands that technology infrastructure as a category must receive financial and policy stimulus to support the development of the Internet industry, a critical driver of domestic consumption.

On August 14th, the China State Council issued a public statement about its planned promotion of domestic IT consumption and singled out the improvement of information infrastructure consumption as a key area of stimulus. The government intends to increase fiscal and tax support for IT companies, improve infrastructure construction, and stimulate the sector to grow at over 20% annually to over 3.2 trillion RMB by 2015. And more recently, as reported by Marbridge Daily on September 22nd, Shang Bing, Vice Minister of the MIIT, stated that part of the government's "Broadband China" stimulus program involves investment of two trillion RMB through 2020 to increase the capacity of China's fixed broadband and mobile broadband infrastructure.

We believe that for VNET, the government is much more of a friend than a foe, and the historically strong relationship between VNET's management team and key policy-makers will only strengthen over time.

First, VNET has always received preferential treatment. VNET is one of the few companies to have successfully leased fiber-optic networks that form China's Internet backbone and is one of only two companies we are aware of with enough licensure to operate a private interconnected network with nationwide scale (CCIH is the other, although CCIH's network has only a fraction of VNET's access points and does not include VNET's nationwide data center footprint). VNET is classified as a "very important" technology company and has historically received favorable tax treatment. And VNET has quickly gotten approval from local real estate administrators for building in highly desirable locations such as Beijing as the Company transitions its data center network away from the cabinets leased from the government in favor of self-built cabinets.

Second, VNET has had preferential access to capital. From the sanctioning of the Company's IPO on a US exchange to raise equity to the credit that local banks have provided during the Company's early days, VNET is one of the few companies in the industry to have enjoyed a fully-funded growth plan. On March 15, 2013, the Company successfully raised at attractive terms one billion RMB through the issuance of dim sum bonds (RMB-denominated bonds that are issued in Hong Kong, allowing foreign credit investors to get exposure to domestic Chinese corporate credit). And more recently, on September 4, VNET received what is effectively an interest-free 900 million RMB loan from the Dongguan province as part of a joint venture between VNET and the government to further develop Internet infrastructure in the region.

Finally, the government is not merely a friend; it is both a customer and a partner. On the list of "who's who in corporate China" that comprises VNET's customer base are various government entities, several of which outsource extremely important networks to VNET (e.g. state-run banks).

VNET's partnership with the telcos has only strengthened over time. VNET is the largest bandwidth reseller for the telcos and operates six of their most important data centers for them.

Practically, it is not in the telcos' best interest to compete with VNET. Their data center business accounts for a mere 2% of telco revenue, and the telcos clearly do not have the technology or human resources required to compete with a highly specialized industry leader with a 16-year history of R&D such as VNET. Also, Chinese telcos have higher priorities to tackle, such as improving last-mile broadband speed, particularly for wireless. This explains why, despite burgeoning demand, telcos have not expanded their own data center capacity on pace with market growth and have even begun to lean on VNET for help in managing their data center assets. Rather than intensify direct competition with VNET, we believe it is much more likely that the telcos eventually decide to either outsource their data center business to VNET or get out of it altogether.

The government understands well that VNET is too important to fail and should continue to support the Company's development as the lynchpin to scaling China's Internet ecosystem.

VNET's Full Complement of Structural Moats

We believe there are only five sustainable structural moats in business, and VNET enjoys all of them:

1. Network effects

  • The 21 Vianet Inter-Cloud Express Network is inherently based on network effects which strengthen as more data centers and access points are added to the network's fabric

2. Economies of scale

  • VNET's business is not human resource-intensive, so the company will enjoy operating leverage as it builds out its infrastructure
  • VNET's scale already positions the company differentially in negotiations with key suppliers, e.g. bandwidth, data center leases, etc.
  • Economies of scale act as formidable barriers to entry, since new entrants would need not only access to capital but also the ability to operate with large cash losses for several years

3. Intellectual Property

  • VNET's proprietary network architecture and BroadEx smart routing technology were built after 16 years of operations and R&D
  • As the only company to have managed a carrier-neutral network of its scale over a uniquely interconnected and efficient network backbone architecture, VNET has accumulated over a decade of knowledge about optimizing, managing, and troubleshooting its own and customers' networks that even its largest customers have not been able to replicate

4. High customer switching costs

  • Switching from VNET's predominantly colocation-based hosting services requires costly and lengthy physical moves of mission-critical hardware out of a nationally distributed network of data center facilities followed by an even lengthier and costlier data center and network reconfiguration
  • VNET provides a service level that integrates its staff with the core business processes of its customers, and these relationships and operating protocols would need to be replicated
  • Customers who require optimal network speed within China have no viable alternative

5. Regulation

  • We believe that China will remain highly protectionist of its telecommunications sector and favor the support of a few large players over the completely free market that the US government chose when it privatized the industry

Beyond these structural moats, VNET also enjoys the following notable competitive advantages at this point in its industry's development:

1. Access to capital

  • Significant financial resources are necessary in order to fund data center construction, which requires technology capital, utilities, real estate, and physical security. Businesses with the cheapest access to capital are often able to gain meaningful advantages.
  • Even for industry leader VNET, access to expansion capital was a key rate limiter until its successful 2011 IPO raised enough equity capital to fund data center expansion at a pace more in line with industry demand. We believe VNET is the only carrier-neutral colocation provider with a fully funded growth plan in China and therefore stands to take share at an accelerated pace.
  • VNET's cost of capital has trended lower since its IPO, with the most recent RMB 900 million line of credit from the government obtained effectively interest-free.

2. Ability to consolidate the industry during its early stage of development

  • M&A is a natural business expansion strategy in the data center industry, since acquired customers and technologies can be readily integrated and then easily cross- and up-sold.
  • In the US, the data center companies that survived the post-2000 washout were able to make acquisitions to more quickly gain share of a troubled industry as the supply/demand imbalances corrected.
  • In China, despite the opposite supply/demand dynamic playing out, VNET is able to similarly acquire in accretive fashion the most promising smaller companies as they struggle to gain scale independently, thereby accelerating the Company's share gains during an ideal period in the industry's growth curve, as demand and technology advancement are booming.
  • Notable acquisitions, which ranged from 14.8 million RMB to 116 million RMB, include:
  1. Regional data center: 21ViaNet@Xian Holding Ltd on July 2012, to gain additional cabinets and data center space, primarily to better accommodate the needs for Tencent.
  2. Managed network services: Beijing Chengyishidai Network Technology Co., Ltd and Zhiboxintong (Beijing) Network Technology Co., Ltd., from September 2010 to December 2011; Guangzhou Gehua Network Technology and Development Company Ltd. in October 2011; Beijing Tianwang Online Communication Technology Co., Ltd. ("Beijing Tianwang") and Beijing Yilong Xinda Technology Co., Ltd. ("Beijing Yilong Xinda") in February 2013.
  3. Content delivery services: Fastweb International Holdings and iJoy Holding Ltd in September 2012.
  4. Cloud: Shanghai Cloud 21Vianet Network, Co., Ltd Shenzhen Cloud Information Technology Co., Ltd in October 2011.
  5. Also, in February 2012, VNET won a bid for radio spectrum in the 2.3 GHz band to provide broadband wireless services in Hong Kong for HK$150 million. The Company intends to use this spectrum to collaborate with China Mobile in 4G/LTE technology development that will position the partnership well for the forthcoming rollout of 4G in China and enable VNET to enhance its core data center business's mobile-related services.

3. Pricing power

  • VNET charges an approximate 30% premium to telco pricing for its hosting service.
  • VNET has consistently increased prices approximately 10% per year and expects to be able to continue doing so for the foreseeable future.
  • VNET has nearly tripled market share in the span of three years while exerting its pricing power.

VII. Customers

VNET is by far the largest carrier-neutral Internet data center service provider in China, with operation in 43 cities and a geographically diversified and loyal base of more than 2,000 domestic and multi-national customers that span many industries ranging from Internet companies to government entities and from blue-chip to small- and medium-sized enterprises. Among the Company's greater than 2,000 customers is a "who's who" list for each target industry, with nearly every major company represented. A small sampling of notable customers is shown below:

Source: 21 Vianet corporate presentation

As with any B2B business, customer satisfaction and customer concentration are two key factors to consider when evaluating the sustainability of VNET's leadership and the risk of losing future share.

As discussed previously, it is telling that VNET charges an approximate 30% premium to telco pricing for colocation, is able to raise prices approximately 10% each year, and was still able to triple market share over the last three years. VNET customers churn at a very low rate of 1.75% per month (as of Q1 2013), as measured by monthly recurring revenue, and none of VNET's top 20 customers have churned (as of Q1 2013), with the notable temporary exception of Tencent, who, as mentioned previously, tried to in-source their nationwide data center operations over a year ago, only to return to VNET within a couple of quarters to outsource to the Company even more of its data center operations.

Moreover, demand is so high for VNET's service that the company has increased cabinet capacity at a CAGR of 48% since 2008, almost twice as fast as the data center industry's overall sales CAGR of 25.2%. Notably, this rate of cabinet addition outpaces even the 44% CAGR by which overall broadband consumption has grown in China, which is particularly impressive considering the industry's constant state of physical data center supply constraint.

Another important indicator of customer satisfaction is the fact that, despite a corporate strategy to shift its cabinet inventory mix from partnered (i.e. leased from telco wholesale inventory) to self-built, VNET still plans to lease about 15% of its 2014 cabinet expansion at the request of customers that want coverage in regions where VNET does not currently have any self-built inventory. These customers are happy to pay the same price for partnered cabinets in regions where VNET has yet to expand to gain access to VNET's proprietary high-speed network and industry-leading service.

One of the most interesting outcomes of our due diligence was gaining an appreciation for how customers use VNET. Although smaller customers may use VNET exclusively, most large customers allocate only a minority of their data center budget to VNET (figures ranged from 15% to 50%). This is partly due to the fact that VNET simply cannot expand cabinet capacity fast enough to meet demand in the most active Internet regions such as Beijing and partly due to the fact that customers tend to host their most important mission-critical servers with VNET. Based on our interviews, customers generally choose VNET over even their own self-managed (in-sourced) data centers for their most important servers. The obvious implication is that customers will leave VNET last if ever forced to reduce data center space for any reason. And clearly, VNET has plenty of room to gain share of wallet with its large customers.

For a business whose largest customer is the Internet sector, which is highly concentrated in usage among the largest 10% that serve 90% of consumption, VNET's customer concentration is surprisingly favorable.

Source: 21 Vianet corporate presentation

We note that all of the metrics discussed in this section were achieved without the benefit of legally being able to offer interconnection services, which we consider the most profitable and "sticky" offering in the data center services industry. As we have discussed in Section IV, we strongly believe these services will be commercially standardized by the government within the next year or two. To wit, the announcement of VNET's joint venture with the Dongguan government made specific mention of interconnection technology as a key focus of the partnership.

VIII. Uniquely Positioned to Lead the Chinese Cloud Industry

State of the Chinese Cloud Computing Market

Although there has been much talk about various large Internet companies offering cloud services this year, the number of cloud deployments with meaningful monetization is negligible at this point. The following companies have announced cloud initiatives but all have made minimal commercial progress:

  • Alibaba: In an attempt to emulate Amazon's highly successful (in the US) AWS offering, Alibaba announced the formation of AliCloud, a subsidiary devoted to the development of cloud computing services that has largely floundered since inception.
  • Baidu (NASDAQ:BIDU): Announced in 2012 its intent to offer various Google-esque Web applications over the cloud. The most successful of these to date is Baidu Wangpan, a Web-based consumer storage solution much like Dropbox that has steadily gained users, with the last reported number around 70 million as of June 26, 2013.
  • Tencent (OTCPK:TCTZF): Not to be one-upped by Baidu, Tencent too launched a competitive storage offering
  • Kingsoft (OTCPK:KSFTF): This software company announced on August 12, 2013 that it would offer 100GB of cloud-based storage for free (double what Dropbox offers for free before charging).
  • Qihoo 360 (NYSE:QIHU): Two days later QIHU announced 360 GB of free storage, which led Baidu to announce that it would offer 1 TB for 1 RMB per user.

Aside from a handful of nearly free cloud-based consumer storage services from local companies in a race to the pricing bottom, the Chinese cloud industry is effectively non-existent. Foreign companies have fared no better and have yet to make inroads in the Chinese market.

So far, the sole participants in the cloud value chain who have benefitted financially, albeit indirectly, are infrastructure companies such as VNET, since cloud solutions increase Internet and therefore infrastructure utilization. Everyone else is currently in investment mode.

Likely Cloud Industry Evolution in China

Of particular note, the business model shift towards SaaS that the US software industry is currently undergoing has yet to begin in China, a market that, at full maturity, could reach total sales anywhere between two to three times that of the American SaaS market. Even before the development of China's SaaS market, Cisco projects that Asia, led by China, will lead the world in cloud usage and have 36% more cloud-based traffic than the US by 2016.

Source: Cisco

Bank of America Merrill Lynch projects that the US SaaS market will account for 20% of the overall US software market, or $220 billion in total sales, within five years (source: "The Cloud Wars Part V: The New Cloud Nine", 13 May 2013, Bank of America Merrill Lynch). To get a sense for the scale of the opportunity in China, consider that if the Chinese SaaS market achieves just 10% of the size of its US counterpart in five years, the cloud industry's total sales of $22 billion will be as large as Baidu's revenue last year. And these figures only estimate the potential size of the SaaS segment of the cloud market in China.

The Chinese cloud market's structure should develop similarly to that of the US, which is comprised of three segments: a) cloud-based software (Software as a Service, SaaS), b) cloud-based platform (Platform as a Service, PaaS), and c) infrastructure (Infrastructure as a Service, IaaS). (By PaaS, we mean companies that provide hosted cloud platforms that offer computing services-from shared hosting services to application-level services such as payment or storage-with which customers or third parties can develop and host their own cloud applications. The preeminent example in the US is Amazon Web Services.)

Although we believe the Chinese market will segment similarly to the US market, we expect the competitive environment to develop quite differently in China. VNET stands to benefit from each of those differences:

  1. Software: Local companies will dominate consumer SaaS, which will achieve massive adoption but struggle to generate profits, while foreign (mostly American) companies will dominate the enterprise and finally realize profits commensurate with a software market as potentially large as China's
  2. Platform: Ultimately, an oligopoly of two to three players, comprised of the VNET/Microsoft partnership and the largest Chinese Internet companies, will dominate this segment, but the VNET/Microsoft partnership's implementation will be the only one to reach scale until China's backbone network congestion issues are resolved several years from now because nobody else has the network efficiency, reliability and scalability, to compete with VNET
  3. Infrastructure: Infrastructure market share gainers will be even more differentially advantaged as cloud models proliferate because cloud's inherent network effects reward scale. VNET will dominate this segment and achieve substantially higher market share than any US counterpart has been able to achieve in the US

The consumer-facing SaaS strategy of China's incumbent Internet giants (Alibaba, Baidu, Tencent, Qihoo 360, et al.) is clear: offer effectively free productivity software as a loss leader to retain Web users on their platforms, hoping to subsidize these offerings with increased monetization from core businesses such as search or online gaming. We don't believe this strategy will change; therefore, we expect widespread adoption of these free or nearly free services, which will benefit infrastructure companies indirectly as the rise of consumer SaaS means increasing demand for their services.

The advent of SaaS will be a huge boon to the enterprise software market. Historically, rampant piracy has limited the enormous potential of the Chinese market for software companies. Our interviews with various local software executives as well as VNET's management team suggest a consensus view that only 5-10% of all software use in China is officially licensed and the rest is pirated. One of the benefits of SaaS over traditional packaged software is that it is much more difficult to pirate, since the software is delivered from a centrally managed server cluster that is activated through Web-based access controls (and not software keys that can be hacked locally). Illicit software patches that bypass registration requirements are obsolete in a SaaS environment. Furthermore, the SaaS on-demand billing model makes software more affordable and reduces pricing barriers to adoption.

Large American enterprise software companies will be the biggest beneficiaries of enterprise SaaS in China, since the vast majority of enterprise software used in China is from the US, with very few local competitors of note. It is much easier to localize US software than it is to replicate a local version from scratch, so even developed Asian markets like Japan mostly use American enterprise software. China will be no different.

By a large margin, the most popular and most pirated enterprise software is Microsoft's. As the Windows operating system itself is adjusted to become increasingly reliant on cloud services and new versions of flagship software such as Microsoft Office are offered exclusively via the cloud, we believe Microsoft will be the biggest winner (as measured by sales) in Chinese enterprise SaaS. VNET, as Microsoft's exclusive partner, stands to benefit not only as the infrastructure provider to Microsoft's cloud initiatives, but also as a revenue-sharing partner of both Microsoft Azure, Microsoft's PaaS offering, as well as Office 365, Microsoft's SaaS offering.

Microsoft Partnership

On November 1, 2012, VNET and Microsoft announced a partnership to bring Windows Azure and Office 365 to China. This partnership is as significant for Microsoft as it is for VNET.

From VNET's vantage point, the partnership provides VNET, an IaaS company, entry into the PaaS and SaaS markets to which it was only exposed indirectly in the past. From Microsoft's vantage point, the partnership, which is exclusive, effectively blocks any foreign competitor such as Amazon or Google from partnering with the only Chinese infrastructure provider that has the network capabilities required to support a nationwide large-scale implementation with the developed market service levels that larger customers demand.

Key terms of the partnership include:

  • Three-year two-way exclusivity. VNET cannot operate any competitor's public cloud offering and Microsoft must exclusively distribute its cloud offerings through VNET
  • Revenue sharing arrangement for both Azure and Office 365. How much VNET gets is unspecified, but we estimate conservatively that at least 25% of total revenue is a safe figure
  • VNET invested primarily by increasing headcount of approximately 800 by 30%, or approximately 240 engineers and support staff dedicated to the partnership
  • Microsoft invested all the capex required and will pay VNET for all data center hosting and infrastructure services on an ongoing basis (Microsoft is not only a partner, but also a large customer)

After one year of preparation, several quarters of up-front investment, and several months of beta testing with a limited number of unpaid customers, VNET expects the partnership to start to monetize in late September of this year with the first paid Azure customers. VNET expects to on-board the first paid Office 365 customers in mid-November.

How much revenue can we expect this partnership to generate for VNET?

Microsoft performed all the financial projections for this partnership and applied local market idiosyncrasies (e.g. lower pricing, lower subscription rates, etc.) observed during their past software launches to a model developed from their global ex-China rollout of cloud services to date. They estimate that total partnership revenue will be $160 million in 2014, $450 million in 2015, and $1 billion in 2016. At a conservative revenue share rate of 25%, VNET is projected to generate $40 million of 90%+ EBITDA margin revenue in 2014, $112.5 million in 2015, and $250 million in 2016. We believe the majority of this revenue is not properly modeled by the sell side.

A bottoms-up calculation of Office alone supports Microsoft's estimates, which appear conservative:

  • Microsoft believes there are approximately 400 million Office users in China, and only 5%, or 20 million, are monetized
  • In the long run, per VNET management, the partnership expects a SaaS version of Office to convert between 30-40% of all Office users. The rest would face a choice of either giving up Office altogether or continuing to use an outdated and unsupported pirated version
  • Global pricing for Office 365 is a reasonable $6/user/month ($72/user annually)
  • If Office 365 can simply convert the lowly 5% rate of total Office users who pay to be licensed, even with no user growth, the partnership would generate $1.44 billion of annual sales
  • If the nearly impossible to pirate SaaS model of software delivery can help Microsoft improve that rate from 5% to a still very low 10%, the partnership would generate $2.88 billion of annual sales

It is not surprising that Microsoft CEO Steve Ballmer stated during the Q3 ceremony for the Office 365 beta launch that he expects Office 365 to be the most profitable product ever offered by Microsoft in China.

Several of the CEO candidates rumored to replace Ballmer at Microsoft have a strong cloud technology background and should push Microsoft's evolutionary transition to the cloud with much more urgency than Ballmer did in the past. Foreshadowing this priority, Microsoft recently announced that it will change the way it reports its financials to break out cloud-based sales explicitly.

On 10/10/13, VNET hired Wing-Dar Ker as President of a newly created division named Microsoft Cloud Operation. No stranger to Microsoft, Ker had a 19-year career at Microsoft, most recently as General Manager of Customer Service and Support for the Asia Pacific and Greater China region. Having overseen customer service and technical support in Asia as well as web support for the Americas and the EMEA regions at Microsoft, Ker is uniquely qualified to ensure that VNET's industry-leading service level extends to its partnership with Microsoft.

Investors who buy VNET at today's prices are effectively getting the future earnings, cash flows, and various strategic benefits of this partnership with Microsoft for free.

IX. Valuation

As is the case with any company deploying a significant percentage of its enterprise value in expansion capital, the most appropriate multiple to use as a measure of value relative to normalized earnings is the EBITDA multiple. And since 2014 is the year in which the largest capacity expansion in VNET's history as well as the Microsoft partnership will start to monetize meaningfully, it is the most relevant year to use for valuation.

Comparable Company Analysis

VNET is significantly undervalued relative to peers, particularly when its much higher growth is considered:

Carrier Neutral Colocation and Managed Hosting companies are VNET's closest comps, while wholesalers who operate primarily as REITs are least comparable to VNET.

VNET's growth-adjusted multiples compare very favorably to comps: 2014E PEG ratio of 0.19 and EV/EBITDA/G at 0.15.

When EBITDA and growth are considered together, VNET is literally off the chart relative to peers:

Precedent Transaction: EQIX Acquisition of Asia Tone

On May 1, 2012, EQIX acquired the assets and operations of Hong Kong-based data center company Asia Tone for US$230.5 million. As a result of the transaction, EQIX gained five operational data centers, one data center under construction and one disaster recovery center across three key markets: Hong Kong, Shanghai and Singapore. Of note, the Shanghai-based cabinet inventory is not directly owned, but partnered.

Although VNET should command a substantial premium to any Asia-based data center, assuming Asia Tone is an exact comp, the 5.8x sales multiple that EQIX paid for Asia Tone still suggests that VNET is significantly undervalued:

VNET-Specific Considerations

A simple multiple-based analysis of comparable company valuations ignores several factors that should be considered when VNET's financials are compared more thoroughly.

Bandwidth reselling. While the vast majority of customers in developed markets buy bandwidth directly from ISPS, as discussed in section VI - "Competitive Differentiation", VNET procures and resells bandwidth to most of its customers. Bandwidth reselling is not a core business for VNET, and with approximately 20% gross margins, is dilutive to overall margins.

Interconnection. As discussed in section IV, VNET is not legally able to offer the high-margin interconnection services that some of its peers offer. An analysis of EQIX's revenue by service illustrates the potential impact that regulatory change could have on VNET:

Depreciation. VNET depreciates its capital expenditures more aggressively than its peers, so the Company's earnings during expansion years are depressed on a relative basis by depreciation charges. The chart below compares the depreciation schedule for its Carrier Neutral Colocation peers. Note that VNET does not directly own any real estate at the moment, so the effective depreciation period for VNET's data centers is only five years.

Microsoft-related head count. For most of 2013, VNET's margins are burdened by the 30% increase in total head count attributed to preparation for the launch of the Microsoft partnership. Clearly, the hiring of that many incremental employees 3-4 quarters in advance of revenue is not a recurring situation.

Key Financial Drivers

So far, 2013 has been an eventful year that should set up a long period of outperformance. We note several key drivers that should complement the massive cabinet expansion of 2013 to allow VNET to grow aggressively while expanding margins.

N.B. Although the Company intends to build an incremental 7,000 cabinets in 2014, we have assumed cabinet expansion of 4,000 units for conservatism.

10% annual pricing increase. As noted previously, the Company has consistently raised prices by approximately 10% per annum, and 2014 should be no different.

Higher concentration of cabinet inventory in Beijing. Beijing is the most robust Internet region in China, home to the majority of Internet companies and China's most sophisticated Internet users. Demand for cabinets nationwide peaks in Beijing, where VNET will deploy the majority of its cabinets in 2013 and 2014. Utilization rates generally reach the low 90s% in Beijing, where the monthly recurring revenue (MRR) per cabinet is between 12,000-13,000 RMB compared to the total average of approximately 10,500 RMB. Coming into 2013, approximately half of all cabinet inventory was in Beijing. About 5,000 of the 8,000 cabinets deployed in 2013, with a largely 4Q13-weighted deployment schedule, will be based in Beijing. VNET's 2014 financial profile should improve due to the incremental regional mix shift to Beijing, which has higher utilization rate, higher price, and much faster sell-through.

Mix shift towards self-built cabinets. Self-built cabinets have ~35% gross margins, 10-12% higher than partnered cabinets. VNET will primarily expand inventory through self-built cabinets going forward, using partnered cabinets only in less developed Internet regions where the ROI of building its own data centers does not pass the 30% bar the Company sets for itself. The table below shows cabinet mix by quarter (3Q13 and beyond reflect our estimates):

1. d

Return to historical utilization levels. VNET's utilization has historically been ~80% but began to decline in Q3 of 2012, bottoming at 66.3% in Q4 of 2012, then rebounding to 70.2% in the 2Q 2013. The decline in utilization levels was primarily attributable to three factors:

1. Nationwide fiber backbone upgrade. In 3Q 2012, VNET began a nationwide upgrade of its fiber backbone capacity in preparation for its integration of Fastweb (its CDN business) and the launch of its cloud partnership with Microsoft in the second half of 2013, which is expected to significantly increase network loads. The upgrade took approximately four months and cost VNET RMB 100 million in capital expenditures. It also resulted in temporary business disruptions throughout the installation process, which led to a temporary slowdown in new cabinet deployment. The upgrade was completed in 1Q 2013, and the Company does not expect another upgrade for at least 3-4 years.

2. Chinese political transition. Though no Chinese management team will openly discuss this issue, it is no secret that the majority of companies delayed their IT spending due to the political uncertainty.

3. Low Utilization in new Southern data centers (Shanghai, Shenzhen and Guangzhou). Unlike Beijing, which has sold out even before cabinets are deployed, data centers in the Southern region are tougher to fill because of its relatively underdeveloped Internet economy. Suboptimal sales leadership exacerbated these issues in late 2012 and early 2013, prompting the installation of a new sales leader to remedy the problem. During its 2Q 2013 earnings conference call, management noted its expectation that utilization in these Southern data centers should rise to ~70% from ~25% utilization levels.

The first two events were clearly one-time. Regarding the underperforming Southern region, the management team has stated that they should be able to raise utilization from ~25% to ~70% by the end of this year. Our checks bear out the progress, as the company has recently signed major customers such as China Mobile, Citic, Guotai and Taikangin the region.

The sensitivity table below shows the impact that improving utilization will have on 2014 results. We estimate that a 5% increase in utilization yields an incremental ~RMB 80-100 million in EBITDA.

Microsoft partnership. Cloud revenue will be by far the largest contributor to margin expansion, with estimated 2014 gross margins of 60% after adjusting for one-time startup costs. Future gross margins should trend upwards towards ~80%.

Another major benefit of Microsoft cloud revenue is the likely re-rating of VNET's shares as it becomes China's first infrastructure provider to vertically integrate cloud services into its suite of offerings. The market should bid VNET multiples higher as cloud revenue ramps, as was the case for RAX in the US:

Increase in high-margin value added services. Fastweb, the CDN business, has ~35% gross margins and will grow sales ~60% this year, taking greater share of VNET's overall revenues. We expect Fastweb to outpace VNET's ~45% overall growth next year as well.

We expect that VNET will increase EBITDA over >50% in 2014 (analysis below). Meanwhile, global peers are growing EBITDA 15-25%.

Given the significantly higher level of growth, we believe VNET should trade at a premium to its global peers. Our 12-month price target of $32 is underpinned by 20.5x 2014 EBITDA multiple (based on our 2014 Adjusted EBITDA estimate of US$104.9 million).

Despite all this analysis, valuing VNET's on 2013 or 2014 results alone is short-sighted. The Company is undergoing the most aggressive expansion in its history to try to meet demand while new Cloud and CDN offerings are just beginning to ramp. Numerous financial catalysts abound as VNET's aggressive capacity addition moderates and the key growth catalyst becomes the addition of new value-added services that drive operating leverage. VNET is on the path to multi-year margin expansion during a period of peak growth that we have estimated solely based only on the facts we know today in a very dynamic industry whose evolution is highly likely to repeat what occurred in developed markets. As in China currently, growth in developed markets was initially driven by raw capacity expansion before value-added service-led growth took over and drove margins and multiples much higher.

Interconnection is one such service that we believe VNET is likely to introduce sometime in the next 18 months given its recent partnership with the Dongguan government to address the inefficiency of the current backbone framework. The following analysis shows the impact to margins that interconnection services alone could have on VNET's business:

With interconnection at just 10% of future revenue, VNET would see a 35% increase in adjusted EBITDA and a 5-point increase to its EBITDA margin.

X. Key Risks

No investment is without risks, and this one is no exception. As with all US-listed Chinese companies, VNET exposes investors to governance risk related to the VIE structure that ADRs use to allow foreign investors to invest in protected Chinese sectors. This is not a risk unique to VNET but is noteworthy nonetheless.

We believe the following risks are the most important company-specific risks to be aware of.

Unanticipated changes to government regulation of the telecommunications industry and/or deterioration of VNET's relationship with governmental authorities. The Chinese telecom sector is one of the most heavily regulated. VNET's business depends on cooperation with a government which has imposed restrictive licensure policies at the local, regional, and national levels. As we have discussed in previous sections, VNET is unique not only because of how critical its success is to the overall Internet economy, but also due to its best-in-class relationship with the government. We don't see that relationship souring, but regulatory risk is nevertheless the most notable in our view.

Macroeconomic slowdown in China. As with any growth sector, China's TMT industry relies on the health of the broader economy, which has begun to moderate from historical peak GDP growth rates. Macro softness will impact demand for premium hosting and network services and potentially slow the trend towards outsourcing as potential customers are forced to downsize IT spending.

Execution. Data center businesses are operationally intensive, require very high service levels, and demand ongoing capital investment. VNET's management is simultaneously almost doubling data center capacity, introducing new products and services (i.e., cloud) and helping China modernize its outdated and overloaded Internet infrastructure. Delays and unforeseen problems will arise, such as the delayed launch of Azure/Office 365 beta from 1Q 2013 to 3Q 2013. The company also mis-executed by over-building outside of Beijing, resulting in low utilization due to a failed go to market partnership with China Telecom. We believe most execution risks ahead of the company are temporary in nature and imminently addressable because the backdrop for growth and the blueprint to navigate it are so clear. Unlike many emerging industries, VNET is exposed to minimal business model or invention risk.

Customers in-sourcing their data centers. Over 60% of VNET's revenue comes from Internet companies such as Tencent and Alibaba, which also operate their own data centers. These customers could attempt to reduce their reliance on VNET by managing their infrastructure directly.

Industry oversupply. Excessive data center supply has occurred periodically in the US, resulting in pricing degradation and financial distress for data center companies, most notably in the early 2000s. China's data center supply constraint amid heavy secular ramp in demand provides a much more favorable macro backdrop for VNET, but as with any capex-intensive business, data center companies are exposed to unforeseeable fluctuations in supply and demand.

XI. Conclusion

We believe that the potential rewards of investing in VNET far outweigh any risks.

VNET enjoys unfair competitive advantages protected by robust, sustainable, and self-reinforcing structural moats as the Company continues to take share of the world's largest and highest-growth Internet infrastructure market, a market that must flourish for a long time in order for Chinese Internet to progress to its potential. The combination of undervaluation, competitive superiority, and enormous greenfield opportunity makes investing in VNET a unique opportunity that might occur again only a handful of times over the next decade.

Disclosure: I am long VNET. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.

Additional disclosure: Refer to additional disclosures here: toroip.com/disclaimer-vnet.html