Industrials & Energy Group

21Vianet — A Turnaround Story at the Heart of China’s Cloud

Rare opportunity where an inflection on an S-Curve meets significant undervaluation

Tiger Capital Management’s Industrials & Energy Group: Ryan Jiang ’23, Autumn Tan ’23, Katie Kolodner ’24, Yash Parikh ’24, Grace Zhuang ’23, and Muhammad Umar ‘22

Executive Summary

21Vianet is the Industrials & Energy group’s highest conviction idea for 2021. We believe 21Vianet is a 3-bagger in the next two years as the multi-year overhang continues to clear, hard catalysts play out, and the company successfully transitions to a more attractive business model. This is a classic case of a company’s messy history resulting in investors overlooking a clear turnaround story. 21Vianet is a Chinese mid-cap that faced fraud allegations in 2014, saw consistent management misexecution from 2014–2017, failed a privatization transaction in 2015, and until Dec. 2020, certain institutional shareholders were expected to dump large block-sales onto the market. Thus, the #2 data center provider at the center of one of the great investment themes of the decade trades at a 45% discount to peers.

This investment has two premises:

1. The overhang around the name has finally cleared and VNET will continue to re-rate to peers (VNET currently at 20x 2021 EBITDA vs $GDS and $CD at >35x 2021 EBITDA).
2. Sell-side is severely mismodeling VNET’s incremental margins as it transitions from retail to wholesale data centers. We are 40% above sell-side consensus for 2023 EBITDA due to our modeling of operating leverage.

Our price target is $126.37 by the end of 2022 from $36.83 today, reflecting a 95% IRR.

Company Overview

21Vianet is a carrier- and cloud-neutral internet data center provider, enabling customers to store their servers in “cabinets” inside of their data centers. They provide hosting services (storing a customer’s data) and value-added services such as VPN or cloud management, charged mostly on a recurring monthly revenue basis (>90% of revenue is recurring). It manages 51,476 cabinets inside of 85 data centers in China, primarily concentrated around tier 1 cities like Beijing and Shanghai.

There are two types of data centers: retail and wholesale. Retail data centers serve multiple customers, each using a few cabinets for their servers to run their own operations. Wholesale data centers are built to the specifications of a particular client who uses the entire data center. Because wholesale clients source bandwidth separately and utilization rates are higher, wholesale cabinets have higher incremental margins (>70% EBITDA) than retail cabinets (~55% EBITDA). 21Vianet is currently transitioning from retail to wholesale, with wholesale expected to make up 80% of net cabinet adds through 2023.

Industry Dynamics

China’s cloud is three to five years behind the United States, and they are nearing the inflection point for widespread cloud adoption.

Our build for China’s IDC Demand

On the demand side, both major public cloud providers and emerging technology leaders are significantly ramping up data center capital expenditures. Alicloud expects to spend RMB 200B on data centers in the next 3 years (>100% of their revenue per year), and Baidu is targeting 333k cabinets by 2030 (from 50k est. today). Moreover, emerging technology players like ByteDance and PinDuoDuo are increasingly demanding private/hybrid cloud servers, and wholesale data providers are the beneficiaries of this demand. We expect 70% of this CapEx to be outsourced to data center operators rather than in-house, which will result in demand significantly outpacing near-term supply and subsequently pressuring pricing upwards. Note that we do not model in any pricing increases — it is a free call option on the investment.

Our build for China’s IDC Supply

On the supply side, there are seven relevant carrier-neutral data center operators and three major telecom data center operators. While the three major telcos make up >60% of the market today, these telcos do not meet the needs of growing enterprises in China and thus will lose market share to carrier-neutral operators. We expect telco data centers to be market share donors as they grow at a ~3% CAGR over the next three years. Comparatively, the consensus industry projection is a significant y/y topline acceleration to 41% in 2021 vs 37% in 2020 and 27% in 2019 for the seven top carrier-neutral data center operators. Among carrier-neutral operators, we believe that 21Vianet is a strong #2 (behind GDS), with 15% market share among the seven players. Our projections for the seven players indicate that 21Vianet will gain significant market share in the next 3 years, increasing to 21% market share by the end of 2023.

Investment Thesis

Thesis Point 1: Sell-side is significantly mis-modeling operating leverage as 21Vianet transitions to wholesale.

The unit economics of 21Vianet’s transition is incredibly messy to model, and sell-side consensus does not understand the incoming inflection upwards in EBITDA margins.

1A: Wholesale cabinet incremental EBITDA margins are ~70%, significantly higher than retail (at ~58%)

We estimate the unit economics for wholesale cabinet costs to be 20% depreciation, 21% utilities, 5% rent/operations/maintenance, and 6% labor, reflecting a 68% incremental EBITDA margin. On the other hand, we estimate the unit economics for retail cabinet costs to be 18% depreciation, 2% operations/maintenance, 4.5% labor, and 35% utilities/bandwidth, reflecting a 58% incremental EBITDA margin. The bridge between these margins is higher retail bandwidth costs (20–25% of retail costs), offset by slightly lower labor and maintenance costs as well as slightly higher pricing. Moreover, wholesale data centers tend to have higher utilization rates (>90%) versus retail (80–85%), spreading fixed costs over more units. Thus, retail data centers have ~40–50% EBITDA margins fully ramped versus wholesale at >65%.

Wholesale unit economic breakdown per GDS & our estimates.

Retail unit economic breakdown for a Chinese peer.

Our unit economic breakdown is derived from a mix of filings in Chinese from competitors (AtHub, BJ Sinnet, Baosight) and industry reports that have a much clearer understanding of the on-the-ground economics when compared to American sell-side analysts. We’ve also confirmed these margins with management comments (Q4 2019 — indicated adj. EBITDA margins >70% for wholesale), GDS’s (a wholesale competitor) cost breakdowns, and project IRR numbers (10–15% unlevered, >20% levered given 70/30 debt/equity financing) from multiple players in the space.

1B: Sell-side’s projections of 34.7% EBITDA margins in 2023 do not make sense given the mix of net cabinet adds during the next three years. We project a 40.9% EBITDA margin in 2023 based on the project pipeline.

During the last earnings call, management raised guidance from 15k net adds annually from 2021–2023 to 25k net cabinet adds annually during the same period. This is a significant acceleration of cabinet growth — their current cabinet count is ~51k, indicating that 21Vianet will more than double cabinet count by the end of 2022. Our thesis is that 21Vianet will be ~60% wholesale by 2024 (from 20% today), which bridges our delta to the Street consensus given higher wholesale margins. Such an assumption drives our ~40% 2023 EBITDA margin (vs wholesale peer EBITDA margins of ~47–50%).

Cabinet Breakdown by Project through 2023

Corresponding utilization rates given project breakdown, ramp-up trends, and economics

Wholesale vs Retail Mix by Project

Corresponding GP given unit economics depicted above

Our conviction comes from the following:

• We noted that management guidance for 25k cabinets & equivalent 180MW of capacity added annually implicitly guides for ~80% of these net adds to be wholesale (21Vianet’s wholesale cabinets are 8kw vs. 5kw for retail).
• Our analysis of the projects in their pipeline (which they do not cleanly disclose wholesale/retail splits) indicates that ~80% of their pipeline through 2023 is wholesale, and management comments indicate they are prioritizing a wholesale split. Thus, we believe the pipeline will shift even more towards wholesale as they land more deals.

Our projections vs. Sell-side consensus

Why is sell-side so off?:

• Sell-side notes often reference the former management guide from Q2 2020 for a 50/50 to 60/40 split between wholesale/retail among new cabinet adds. We believe these numbers are outdated and do not reflect the pipeline.
• Sell-side is consistently conservative with 21Vianet due to its historical misexecution problems. We do not believe this to be a problem, as we’ll cover more in the second thesis point.

Thesis Point 2: The company’s history has created a significant, multi-year overhang that has resulted in investors overlooking the company’s turnaround story. We think the stock will re-rate to 25x NTM EBITDA from 20x today.

From 2014 to 2019, the stock fell nearly 80% due to a series of missteps for the company and, despite a >350% rally in 2020, trades at a >45% discount to peers despite equivalent financial profiles and better or near-equivalent competitive positioning.

A largely refuted 2014 short-seller report alleging fraud, significant management misexecution from 2014–2017, and a failed privatization transaction in 2015 prompted the majority of sell-side to drop coverage, destroyed the long-term shareholder base, and created a significant overhang in the share price.

2A: This overhang has begun to clear and investors will continue to wake up to the turnaround story in the works.

We went line-by-line through the short-seller report and believe the allegations are false — the consensus among sell-side and our own view is that much of it has been disproven by 21Vianet’s rebuttal and subsequent financial performance by 21Vianet (see appendix). Moreover, structural changes in the company have successfully turned 21Vianet around in the past three years.

1. Management has largely been replaced: critical board members and executives were added, and financial disclosure has improved steadily since the fraud allegations and management missteps. After the short-seller report, Tus-Holdings, Tsinghua University’s Venture fund, took a position for 51% of voting rights and has overhauled management after consistent guidance misses & misexecution. The CFO from 2014–2017, Terry Wang, was pushed out and replaced by Sharon Liu, who has not missed guidance since Q4 2018. Moreover, Alvin Wang, a former President of a Tus-Holdings subsidiary, was instated as Wholesale IDC CEO, and alongside Sharon Liu, has begun to successfully implement their shift to wholesale data centers. These are only two examples — nearly the entire former management team has been replaced with well-respected individuals in the industry.
2. Blackstone TacOps Investment: In the Summer of 2020, Blackstone’s TacOps team invested $150 million in preferred shares for 21Vianet. We believe Blackstone’s quality of diligence is a major vote of confidence for the quality of management and helps to alleviate potential concerns about fraud. Moreover, we were told from a member of the team that the group targets 40–50%+ IRR, which also helps us build conviction in our model.
3. Block shareholders have exited their investments: Kingsoft and Xiaomi, which together took a 15% stake in the company after the fraud allegations to stabilize the company, both sold their stakes in late 2020. Investors had been anticipating their exit given that the two companies invested to preserve the stability of their own relationships with 21Vianet, and they were looking for an exit upon recovery.

2B: Clearing of the overhang has already started to catalyze a re-rating towards industry peers, and we believe this will continue in the near future as the story continues to clear up and the turnaround becomes clearer.

We view the stock’s 80% rally in late 2020 as an indication that investors are warming to the story. There is significantly more room for 21Vianet’s stock to re-rate. We see no reason that VNET should trade at a 45% discount to $CD (35x EV/2021 EBITDA vs VNET’s 21x) given that $CD is competitively behind 21Vianet and only growing 5–6% faster in 2022. We believe VNET’s multiple will continue to expand to 25x NTM EBITDA by 2023 (still a 25% discount to peer multiples). Moreover, wholesale data center multiples tend to be higher at >30x EV/NTM EBITDA and VNET should re-rate as it transitions.

Valuation & Core Assumptions

Returns calculations

Our P&L Statement in our base case

Base: We are modeling:

1. Revenue acceleration to 37.8% in 2021 and 41.4% in 2022 for 21Vianet due to its project pipeline, guidance raise to 25k net cabinet adds through 2023, and the broader industry dynamics at play.
2. Operating leverage from the wholesale transition driving a ~40% EBITDA margin in 2023 versus ~28% today.
3. 25x 2023 EBITDA multiple to underwrite our $126.37 12/31/2022 PT. We believe this multiple is fair given that it corresponds to a 3.5% 2023 steady-state FCF yield (subtracting only maintenance capex) with 40%+ steady-state FCF growth.

For more color:

Assumptions — built from a project-by-project summation of current cabinets & pipeline

Bear: There are two separate bear cases — one where the company is fraudulent and one where it misexecutes.

1. If it is a fraud, we assign the company a value of $0. We believe the probability of this is minimal.
2. If it isn’t a fraud, there is nearly no downside. Given a 16.5x 2023 EBITDA multiple (down from 21x today and corresponding to a 6% SS FCF Yield), significant revenue deceleration (to LDD), and decremental margins (4.5%) yields 0% downside from today’s price.

Catalysts

1. Hong Kong Dual Listing: 21Vianet has hired Credit Suisse & Morgan Stanley to consider a dual listing in Hong Kong. We believe such a move would catalyze significant share price appreciation as 21Vianet gains a domestic shareholder base that understands the value of the business, lowers its cost of capital, and increases sell-side coverage in Asia.
2. Landing of Additional Wholesale Contracts: 21Vianet has already landed large wholesale contracts and MOUs with leading cloud providers in China. We believe more of these contracts will solidify the transition towards wholesale.
3. Continued Management Execution: 21Vianet has a historic management credibility problem. Continuing to execute as they have for the past year will extend the rally that has started to re-rate the company back to industry peers.

Risks

1. Management Misexecution/Fraudulent Financials: self-evident, read rest of note.
2. Oversupply of Data Centers: See macro research in the appendix as well as the industry overview.
3. Rising Interest Rates: This is a company that frequently raises capital across the capital structure, and rising interest rates will raise their cost of capital and subsequently make it harder for the company to aggressively grow.