What is holding me back from buying Keppel DC REIT? (I applied for it during the IPO but did not buy any post-IPO as I thought that the price does not offer me sufficient margin of safety. 7% is so-so, 6% is not high enough...)
1. Lease expiry
When you look deeper into the distribution of rental income vs lease expiry, 66.8% of the leases are 2.3 years, which is very short. In terms of distribution of rental income vs tenants, 62.1% belong to internet enterprises and IT services. My rough guess is that the 62.1% of tenants form the majority of the 66.8% short leases.
2. Demand and Supply
Keppel DC REIT appears to suggest that the growth in global data centre traffic (which is something like internet traffic) will provide the demand for data centres. To me, there are certain assumptions for this correlation to hold true:
- Size of servers remain the same
- Servers will still be in existence and will not be overtaken by a newer breed of hardware (much like how servers overtook mainframes 20 years ago)
- No consolidation of server infrastructure across the globe (hence demand keeps growing)
- Wired network infrastructure will still be in existence and will not be overtaken by a newer wired/wireless/satellite technology (much like how copper wires were replaced with eternet cables 20 years ago and then fibre cables 10 years ago)
If you believe in those assumptions, then perhaps the growth in data centre traffic will indeed continue to contribute to the demand for data centres. Personally, I feel that the size of servers is already shrinking at a very fast pace. For example, if we need to use 5 server racks 5 years ago, we can easily get the same computing power with a fraction of a server rack today. (This is purely based on my own observations which will need to be further validated to be believed.) As servers become more energy and space efficient, data centre space (physical space which is what Keppel DC REIT is about) requirements will definitely be reduced. Global data centre traffic can grow, storage requirements can grow, and data centre requirements can still be reduced at the same time if any of the assumptions I stated above suddenly become irrelevant.
3. Valuation model
The valuation of the data centres assumes that the space is worth a 1000 times more than the raw cost of renting the physical space (at industrial REIT rental rates) because of the specialised manpower and equipment required to manage the data centre, and high barrier for entry because of the high capital investment required.
While this is probably the best valuation method now, we need to remind ourselves that this method of valuation is not an assurance of its asset value. In the container shipping industry, we had seen how software and machines automated the entire sorting and movement of containers. Ships that were valued based on the pre-automation days of specialised manpower rates become overvalued. In telecommunications, we had seen how cable TV had become obselete in a matter of 20 years. If the coaxial cables were valued based on the cable TV subscription fees then they would be not worth anything now because fibre cables had fully replaced the network infrastructure.
Telecommunications as an industry will not disappear because people still need to communicate. Data centres businesses will not disappear too because people still need the infrastructure for internet applications. We just need to ask ourselves how much we are willing to pay as an income/value/growth investor. As an income investor, I will expect a minimum of 8% yield to consider an entry. I personally do not see any growth or value in this, so ask someone else about that.
References:
