Keppel T&T was able to finally stabilize their data center to realize their investments by bundling them as a real estate investment trust. This is something rather new in Singapore so I decide to take a look at it.
Summary of Keppel DC REIT
- WALE of 7.8 years based on NLA (compare to retail malls which are usually 3 years, logistics tenants 5-6 years, Australia offices 10-15 years, healthcare assets 10-15 years)
- 93.5% occupancy
- Mixture of co-location, double and triple net leases
- Forecasted Yield of 6.8% and 7.1% for FY2015 and 2016 respectively
- Most have build in rental escalation of 2% to 4% as well as variable rent as a basis of % of EBITDA (Gore Hill 3.5%, S25 and T25 3%, iSeek 2-4%,
- Leverage 27.8% with an average duration of 4.4 years
- Base fee: 0.5% of REIT’s deposited property, 3.5% of the year’s net property income, 1% acquisition fee
- Cornerstone: AEW Asia, DBS Bank, Eastspring Investments, Fortress Capital, Gordon Tang (Sing Haiyi NED), Lim Chap Huat ( Soilbuild Group owner), Myriad Asset Management, Wellington Management Company
This REIT is interesting in that, it has a relatively long tenant average lease which creates a certain predictability in cash flow. Yet, the Singapore assets are with rather short land lease of 40 years (10 years + 30 years). This looks very much like Cache Logistics with their typical 6 years contract but very short land lease. In this case, Cache was yielding 8% when the IPO came out versus this which is 6.8%. Gearing for these 2 are almost the same.
Typical long leases are embedded with rental escalation but in the case of Keppel DC, they seem much better than Cache.
One note is that, as part of co-location, the customers require Keppel to provide servers. In this case, there may need to be some future capital expenditure unlike other kind of REITs.
Different kind of co-location leads to different levels of rental rates. This can range from bare-bones co-location, telecom carrier based co-location and services based co-location. In the first instance, it is very commodity like, and commands the lowest value add, since the REIT is just renting the space based on per square foot or meters. For carrier based, the customer is renting based on connectivity and this as well as the serviced based co-location, is based on the level of capacity, connectivity, ,human factor, security.
Keppel’s main asset M1 Limited would obviously be one of the carrier based co-location business.
Typically the service and carrier based customers lease for a shorter duration perhaps 1 to 3 years. A bare bones customer could lease for 10-15 years.
Being previously part of situations where i deal with a bare bones data centre co-location service, there are hidden cost to the customer such as design, capacity, power and security and space management, not to mention that staff costs have to cater for routine maintenance, deployments as well as fast turn around.
Equinix, a US based data centre provider can be seen as a large service-based co-location customer. In the case of their engagement with Maple Industrial Trust, they have a design, build and lease deal with them, leasing for 20 years.
The way i look at these carrier based and service based its more like the master tenant relationship that a few of those industrial REIT have, such as Sabana and Aims Amp.
Singapore good location to host regional data centers
Due to the political safety, communications infrastructure, Singapore makes a good location for businesses who wants a strategic choice to host their IT infrastructure. Due to these factors, it means the cost of ownership might be lower than a place where the land price or lease could be much lower. A ready educated group of IT professionals also make Singapore a good choice to host your data centre for long term.
Defensive nature of data centers
This buzz word have been thrown around a lot. And in a certain sense people want to think that they put their money in an asset that is not going to lose them money.
Just remember that business may go on as usual, but you are buying the group of assets. If you pay SG$5,000 for an IPAD Air, your resell price is much lower than that. This is why price is what you pay value is what you get.
The way I look at it, data centres are defensive in that, they are rather uncorrelated to the normal economic cycles. It infrastructure are considered capital investments, for some like DBS, Facebook and Google, are critical for their business. For the small business enterprises, it has come to a stage where failure or downtime can lead to loss of revenue.
A typical economic down cycle could last around 3 years (my estimation based on what i read in the past, no data to back it up) but really, businesses invest their IT infrastructure strategically. There are usually a baseline of capacity required to run businesses. You do not say I will not use my IT infrastructure because business is bad, you still need to keep your payroll, resource planning, decision support systems going. If you don’t have that, then when the business comes in do you set them up again? That’s not possible. You reduce less strategic core IT projects but likely do not reduce them entirely.
I look at this in different tiers:
I have just explain from the most downstream tier the customers.
In the middle tier there are the service providers and carrier providers, these are the tenants of the industrial properties or the data centre REITs. In an economic downturn, they might see their customers reducing their capacity. Whether the service providers can garner new business depends more on the longer term outlook of business in the region, whether they see a strategic investment in this region.
As these service providers are in the business to provide value added services and capacity, even if the economic cycle is not looking good, they would still need to maintain that capacity. Without the capacity, a big client comes along, they don’t even have the capacity to offer them, they cannot secure the customer in the first place.
World wide service providers like Equinix for example, need to keep a footprint in different countries because their end customers sees that as their edge. These players have plan longer term and would invest in a data centre here only when they hold a good strategic long term view.
The bare bone industrial property owners and the bare bone data centre REITs are the last tier, and if there are much co-locations, it depends on the overall state of health of their service providers. The lease of these are usually very long.
The problem here is that the trend is such that, It departments are getting smaller, functions are provided by cloud providers such as Symantec for Security, Microsoft Office for productivity, ESRI’s cloud solution for geo spatial. Virtualization have in my opinion, also make back end infrastructure very fluid.
All this means that the power lies in these large providers. Large providers tend to work establish players that they can ramp up and ramp down their back end servers based on demand for their services. So suppose that demand for their products are weak in Singapore and they see a pick up in business in Japan, they can easily spin down their capacity and relocate to Japan. This concept is how Netflix can spin up and spin down their virtual servers on demand riding on Amazon AWS (Equinix)
This impact is more felt with the service providers, and in order for this fluid movement of back-end, the service providers must be establish, reliable not to mention a large footprint. The large software companies will likely go with the international supply chain (Digital Ocean, Amazon AWS, Azure which should all ride on Equinix). The small up and coming start ups are usually not Singapore centric ( small addressable market), or they can start using Singapore during the incubation stage and then deploy their services elsewhere, making use of services like Digital Ocean and Amazon AWS.
Equinix in this case is unique in that they are a step up over a bare-bones co-locations, providing inter-data center connectivity, internet exchange. Their economic moat is that the guys in their data centers are the Google, Facebook, Amazon, Digital Ocean. If you want seemless intergration and the best quality of service with these applications, you got to be in the same neighbourhood, and this means making use of Equinix facilities.
Not everyone puts international connectivity high on their priorty. You will still have your government companies that are looking to outsource their data centres. There are certain advantages of this to them, not that I agree from a strategic angle this is a wise thing to do. When it comes to the financial services companies and government related companies, security and reliability are more important, thus given the ease to which servers can be move about, this is seldom done due to security, processes and reliability.
Data centres and providers can cater to a more local context in this case. It is not that they are cut off from the outside world, rather your customers might not have the best connectivity to where they want.
“Connectivity was an issue in Australia,” said Fagan. “Japan is a tough place to start out doing business in Asia. You’re limited to the Japanese market, and we saw a lot of barriers. One of the key things we look at in terms of data protection and customer privacy, and that’s the main reason we crossed China off the list pretty quickly, bringing it down to a choice between Hong Kong and Singapore.”
Fagan said Hong Kong offered the best choice on the company’s key selection criteria. “In Singapore, from a conductivity standpoint, you’re pretty much stuck with SingTel,” said Fagan. “One of the other questions was, how do you get into mainland China? Clearly, we weren’t going to start there, and Hong Kong was the best gateway.”
Fagan said some Rackspace customers based in the Asia/Pacific region wanted to move from the US or UK data centers to the new Hong Kong facility, but said most of the Hong Kong clients are expanding their presence with Rackspace.
“We’re bullish on our business,” said Fagan. “Clearly the goal is not just to serve Hong Kong to increase our penetration in the Asia-Pacific market. One of the things we are looking at is what is the right strategy for data center infrastructure.” –Rackspace 2008
While, it is true that computer equipment and racks invested are sunk costs and creates switching costs, there is no reason why a customer after 3 years or 10 years cannot move capacity to a rival data centre. I have work in cases where the data centres are own by 3 different parties. The value here is more towards the service provider rather than the REIT or property owner.
Keppel DC strength lies in their 2 data center locally with tight integration with their partner M1. Perhaps out of Singapore, they are not really that competitive if you are looking for a data center provider with a global or regional footprint. This does not discount that majority of the companies that are co-located are financial services and companies affiliated with Keppel T&T. This should provide a certain level of defensiveness.
Because everyone thinks it is defensive….
Everyone wants a piece of this pie. The question is whether this pie is big enough for everyone. Digital Realty has a footprint here in Singapore, they are a landlord and work with their own service based tenants to come into this region.
Equinix is more of a service based provider but they are also in this region and for Digital Realty and Equinix they would have brought some better practices over from USA. They also have their economic moats as previously stated.
Keppel is not the only local player here. In fact, if we are talking about bare bones, how difficult is it for a property company such as Boustead to build up their ability to build data centres.
Keppel cites the expertise required to build a data centre but I can envision an international player + Boustead having the competency to fix up one just as nicely.
I won’t be surprise you start seeing some REITs converting some of their properties to data centers if they have a long term customer who wants to set up something like this renting on a triple net lease.
This space is competitive, but I see that this is a segment that is likely to grow due to the progression of how IT infra have become more critical, more part of our lives. There may be space for a few players.
If I were to own something along these lines, I would rather be Maple Industrial Trust, who worked with a global player like Equinix, getting them to lease triple-net bare bone for 20 years with an option for a further 5+5 years with a yearly 2% escalation. The trend looks to be that global service providers forming a web of well connected infrastructure and industrial REITs like MIT lock in a very long predictable lease and a strong tenant which reduces risk of not invested long enough.
For this reason, I felt that data centers are not exclusive to Keppel, despite their great expertise.
I’m honestly hoping that Keppel surprise me with a positive spin off. Most of their spin-offs look like a dumping exercise. You will be much better off owning their parents. Still, if you purchase them at the right valuations, they can perform decently. This is usually during periods where there are greater systematic risks in the market.
To get started with dividend investing, start by bookmarking my Dividend Stock Tracker which shows the prevailing yields of blue chip dividend stocks, utilities, REITs updated nightly.
- Defensive Plays in real estate [WSJ]
- Digital Realty buys Singapore Data Center 
- Rackspace outlines Asian Expansion 
- Data Center REITs outperforming markets in face of bearish street