The planned InterXion (INXN) IPO, scheduled for Friday, January 28, will draw investors' attention to the European network-neutral colocation industry.
IPOdesktop, on Seeking Alpha, has an interesting analysis of InterXion's main financials and metrics. The expected market cap, about $ 780 million using the mid point of the offering, speaks for the interest generated by the fast growing data center sector.
In this article we would like to put under the spotlight one little technical aspect related to the existing different approach toward interconnection services in Europe and U.S., which is often undervalued by investors, but very important, in our opinion, for a proper understanding of the European colo industry.
The potential implications on margins and churn are, as we will see, huge, and may justify using different modelings and metrics when evaluating European and North American colocation companies.
One of InterXion's main competitors in Europe, Equinix (EQIX), with operations both in North America, Europe and Asia, can be used as a bellwether to explain the different colocation business models existing in these different markets.
If we examine Equinix's results, we'll immediately notice that U.S. revenues are derived primarily from colocation and interconnection services, while revenues in Europe are derived from colocation and, marginally, managed infrastructure services.
Interconnection services represent a very high margin revenue source for North American colocation companies (up to 90%). As we will see, these services mean a small upfront cost for the company, but ensure a very profitable recurring monthly revenue stream for the whole length of the colocation contract (usually a few years).
Historically, each European country developed a not-for-profit interconnection point in order to facilitate the interconnection between local and foreign telecoms. These facilities gained such traction that they now represent the place to be for connectivity. Colocation companies, like InterXion, TeleCityGroup or Equinix, relate, in part, to these facilities to gain access to the widest number of telecoms, ISPs, etc. in each country where they have a data center. However, this makes their offering weaker as the same level of inter-connectivity is often available for their competitors, through the same partnership.
In Q3 2010 interconnection services represented for Equinix about 20% of revenues in North America, while they were only 3.5% in Europe.
These services also constitute a differentiators sales tool for U.S. providers, and insure a higher “stickiness” to the facility, as customers most probably wouldn't find the same level of connectivity at a competitor's place. Another reason why prices per cabinet are usually higher in North America than Europe, as good facilities benefit both from the revenues generated by the additional interconnection services they can offer and from being capable to sell an “un-commoditized” product.
Churn is not really very high in the colocation industry, but it has recently become one of the most watched for metrics by analysts and investors. Stickiness, as we will see, is not only the result of being capable to offer good connectivity, but we believe it is fair to say that European colocation players cannot really differentiate their offering from each other (from a connectivity point of view) and do not enjoy the high margins generated by these services in North America. Reason why, in our opinion, smaller multiples should be applied to European colocation providers.
Coming back to InterXion, it is interesting to underline that the company has so far enjoyed a very low level of churn (about 0.6% per month), but warned in its IPO filing that more than 50% of its revenues will be coming for renewal within the next 12 months.
For a proper technical explication and better understanding of the background of European not-for-profit peering points, and what it means for the industry, we decided to contact the best expert in the sector, Bill Norton.
Bill was co-founder and chief technical liaison for Equinix until 2008, spending 90% of his time working closely with the Peering Coordinator Community. In this period he created the white paper process, identifying interesting and important Internet Peering operations topics, and documenting what he learned from the peering community.
It was during his years at Equinix that the company started assembling its peering community, adding content providers and enterprises like IBM (IBM), Google (GOOG) and Microsoft (MSFT) to its customer list, and convincing all the major U.S. networks to exchange traffic at the Equinix data centers in North America – an impressive win, and the foundation of the company's success. With AT&T (T), UUNet, Level 3 (LVLT), etc. on its side, Equinix started building its leadership as the major network-neutral data center in the U.S.
Today, Bill is a highly sought after public speaker and international recognized expert on Internet Peering. He is currently engaged as the Executive Director and Consultant for DrPeering.net, a leading Internet Peering portal and consultancy.
Bill can help us understand the background that differentiates peering points between Europe and the U.S.:
The peering technologies (BGP, Ethernet Switches, some flavor of WDM between distributed switches) are generally the same but the business models are different in at least three key ways.
The U.S. Internet Exchange Points tend to be large-scale commercial operations that combine the operation of colocation services and the public peering and cross connect fabrics, while the European Internet Exchange Points separate out the colocation services from the operation of the switch fabric. The European colocation services are operated by commercial entities like InterXion, TeleCity, TeleHouse, etc. but the operation of the Internet Exchange fabric is generally done by a not-for-profit association of the peering participants. These associations have voting, a budget, staff to operate the infrastructure on behalf of the participants, etc. and seek to provide the best service possible for the lowest possible price. While they need to maintain enough cash for working capital and anticipated upgrades, there is also a requirement to not make a profit from the activities. This is why the European IXes tend to lead the charge in dropping prices as well.
Next, where the European IXes are spread across multiple colocation providers, the U.S.-based IXes are generally contained within the same colocation provider building or campus(es). One implication of this is that an ISP can choose a colocation provider with a different mix of price/product/service; there is an open market for colocation space. There is comparably little negotiation room with the colocation provider if you need to be at a particular IX in the U.S.
Finally, the aggregate volume of traffic over the public peering fabrics tends to much much greater in Europe than in the U.S. The DE-CIX (Germany), AMS-IX (Netherlands), LINX (Great Britain), etc. have always been growing on the bleeding edge of the public peering fabric technology. These largest European IX fabrics are breaking the 1Tbps barrier on a regular basis now while any individual public peering fabric in the U.S. might hit one tenth of that load. (On the U.S. side we actually like this because the Europeans are also very open and sharing of experiences on the bleeding edge. When the U.S. fabrics hit this point, the bugs will be well worked out and the switches hardened at this scale.)
What are the main financial implications of this difference for European and North American colocation players?
The U.S. based IXes make a lot of money from peering. A cross connect between two parties might cost $40 to run and generates maybe $250 in revenue every month. That's a good business. On top of that, the more meshed the customer base is, the more difficult it is for customers to leave. Ellen Hancock famously stated that IDCs are like roach motels - customers check in but they don't check out. Interconnection is indeed the glue that keeps the parties from leaving the colocation space, the so-called stickiness of the facility. The larger the participant base, the larger the interconnection potential, the stronger the glue.
Since the European peering model separates the colocation from the peering, the colocation business is much more of a commodity. The same stickiness though applies to the well established European IX, but as a not-for-profit, they are not in it for the margins. As a side effect of the success of the peering fabric, the colocation provider gets stickiness.
As most colo players in Europe connect to not-for-profit peering points, which are accessible to competitors as well, could it be correct to speculate that customers' stickiness is weaker in Europe than in the U.S.?
To a large degree colocation provider stickiness is weaker than the U.S. counterpart, and this is by design. European colocation-neutral IXes allow customers to choose the price-performance characteristics that best meet their needs.
Having said that, it is also clear that it is a bit of a pain to move equipment between colocation spaces so it is not done without a good reason. I asked the large content peers about this move and some of them said that, if they had to, plan the move to coincide with equipment upgrades. These guys said that servers get "forklifted" every 18 months or so, and when that happens they would install the new gear at the new colo and bring the new site up before moving out of the other one.
The other trend to highlight is for customers to move the large footprint into commodity data center space while maintaining a small network "umbilical cord" back to a well populated IX. This in my opinion can be a really good play and also helps maximize the value of the IX.
On a personal level, what are Bill Norton's future plans?
He is trying to finish up a manuscript for "Peering@TheCoreOfTheInter.net" that assembles what he learned over the decade working with the peering community. "When I am not working on the book I do peering and strategy consulting work for several international ISPs. I seem to get a lot of consulting gigs with the investment bankers and stock analysts. I will continue on this front as well, using proceeds also to maintain the DrPeering.net web site.
"But I am an entrepreneur at heart though - I expect to stumble upon an interesting opportunity and launch or help launch a startup."