In my previous article I tried to examine the landscape surrounding 21Vianet's (NASDAQ:VNET) IPO, which can be resumed, using a quote from a recent Seeking Alpha article by Screener.co, as an environment where "high growth tech companies in emerging markets are trading at sky high valuations."
These valuations may be partially due to very high growth expectations for the Internet sector, especially in China. However, the specific sector of 21Vianet's activity (colocation) doesn't really seem to enjoy, in China, a much higher forecasted growth than it is expected to have in the U.S., Europe or the whole Asia Pacific area. I will point out later that the company itself is setting a growth target in line with most western competitors and slightly below market expectations.
21Vianet was also smart enough to acquire, just six months before the IPO, two companies, Beijing Chengyishidai Network Technology Co., Ltd. and Zhiboxintong (Beijing) Network Technology Co., Ltd., that represented about 30% of its revenues in Q4 2010. These acquisitions were made at very conservative multiples (about 1 time Q4 annualized revenues), while the multiples enjoyed by the combined company, after the floating, are now much higher (7 to 8 times Q4 annualized revenues) -- an additional reason to wonder if today's valuation incorporates unrealistic expectations for the company.
Time to dig a little bit deeper into 21Vianet's business model.
The company describes itself as "the largest carrier-neutral Internet data center services provider in China as measured by revenues in 2009."
However, if we examine 21Vianet's activity better, we find that the company is operating both its own data centers and hosting customers at partners facilities, which are mostly owned by telecoms.
This quote is taken from the IPO filing (pg.14):
As of December 31, 2010, we leased 3,105 cabinets from China Telecom and China Unicom that are housed in our 44 partnered data centers, accounting for 54.0% of the total number of our cabinets under management. We also rely on China Telecom and China Unicom for a significant portion of our bandwidth needs and lease optical fibers from them to connect our data centers with each other and with the telecommunications backbones and other ISPs. Our agreements with affiliates of China Telecom or China Unicom usually have a one-year term with automatic renewal option. In addition, China Telecom and China Unicom also provide data center services and directly compete with us for customers.
A couple of comments.
21Vianet has only 2,645 cabinets (year's end data) in its own facilities, a number that doesn't really seem impressive [just as a reference, Equinix (NASDAQ:EQIX) started billing about 2,000 cabinets in the North American market alone in Q1 2011].
The fact that the partnered data center are all under one-year contracts also raises a red flag to the risks of having more than 50% capacity that might be recalled at very short notice. I will not comment on the fact that it is hard to imagine that these facilities may be considered "network-neutral," as the company might have found a way to guarantee its customers diversified networks offerings in spite of the original limitations.
The recent acquisitions strengthened the company in managed network services, which represented about 44% of revenues in Q4 2010.
At the end of the day, in spite of the company's self description, it is probably safer to compare 21Vianet to Internap (NASDAQ:INAP), as both companies rely on owned and partnered facilities for data center services and offer high quality IP Services.
Those who are familiar with Internap know that the limitations of this business model relate to the fact that margins are much lower in partnered data centers (Internap is now targeting 20% margins), as opposed to the 50% plus margins usually enjoyed by pure network-neutral colocation players.
21Vianet's churn rate is also interesting. The company reports a very low monthly churn rate in 2009 and 2010 (0.8% and 0.9% respectively), in line with comparable businesses in Europe, Asia or the U.S. Strangely enough, it experienced a 3.3% monthly churn rate in 2008 (that's in excess of 39% on a yearly basis).
The company had about 1,300 customers as of December 31, 2010. Recurring revenues from top 20 customers represented about 40% of 21Vianet's top line. From these data it seems that a large number of customers may have just one or two cabinet deals, on average, with the company.
Future plans are quite ambitious (from the IPO filing, pg. 13):
We plan to increase the aggregate number of cabinets under our management from 5,750 cabinets as of December 31, 2010 to more than 10,000 cabinets by the end of 2013 through adding new self-built data centers and partnered data centers. In addition, we plan to expand our private optical fiber network to cover all of our major data centers throughout China and plan to increase our network services capacity from over 295 gigabits per second presently to over 1,000 gigabits per second by the end of 2013. To achieve this expansion plan, we will be required to commit a substantial amount of operating and financial resources.
If we look at these numbers, however, it seems that the company is pursuing an average 20% growth in cabinets for the next three years, which is slightly less than the forecasted growth for the network-neutral colocation market. More info about the company's building strategy is available in the filing, although there are not a lot of details (timing, costs, cabinets number, size):
We are currently building six additional data centers in China’s major Internet cities: Beijing, Shanghai, Shenzhen, Hangzhou, Xi’an and greater Guangzhou metropolitan area, which will both help us meet increased customer demand. We also plan to build and utilize container-based data centers as a rapid way to add more cabinets under management.
The company also experienced a very good occupancy rate in the last few years (about 79.6% in 2008, 80.9% in 2009 and 78.8% at the end of 2010), which is a positive sign (as it has been able to manage well its inventory, especially in partners facilities), but may limit future growth unless new cabinet additions, both in own or partnered sites, are foreseen in a short time frame.
Cisco (NASDAQ:CSCO) is among 21Vianet's shareholders, and has committed to buying more shares at the IPO (pg. 153):
Our existing shareholder, Cisco Systems International, B.V. has subscribed to approximately US$1 million of ADSs offered in this offering at the initial public offering price by itself and/or through its affiliates (collectively, "Cisco"). Cisco has agreed, subject to certain exceptions, not to transfer or dispose of, directly or indirectly, any of the ADSs acquired in this offering for a period of 90 days after the date of this prospectus.
Existing shareholders have a 180 day lock up period.
There are other data which might be interesting to comment on (like average revenue per cabinet, that came in at a very strong $ 6,350 at year's end, but may not be comparable with the western data that range at about $2,000 per cabinet, for pure colocation companies), but this brief write up might be a good starting point for helping take an informed investing decision.
To summarize, it's an interesting company that would need some further diligence to be understood properly. Its business model is a combination of network-neutral colocation, both in its own and partner sites, and managed network services. It operates in a very interesting market with great growth prospects. However, it trades at multiples I do not feel comfortable investing in as there are several other western players, both in the U.S. and Europe, that have been performing pretty well so far, and may still offer good prospects going forward, with more limited risk (at least because it is possible a deeper analysis of their assets and their EBITDA margins appear more mature).
Disclosure: I am long EQIX.