Data Centres Terminal value: Beyond 2034
This modelling exercise is for the boosting of Spark's 22MW data centre capacity out to 140MW by EOFY2031, with nominal data processing capacity remaining constant from that point. If Spark in the future revises their growth plans to build out beyond the planned 140MW of capacity, then that potential 'subsequent growth expansion' is not being modelled. Nevertheless I do expect some growth even if the data centre capacity remains nominally constant, to take account of inflation, as some inflation adjustment pricing is typically prided into longer term rental contracts. I am setting my future annual 'inlation adjustment' at 3%, which is at the top end of reserve bank targets. This is principally because I expect energy prices to rise faster than general inflation and the contract charging regime to reflect that, plus I am assuming some marginal rise in data handling efficiency within the data centres themselves as rack equipment is tweaked and updated.
I am going to use the 'perpetuity method'. The perpetuity growth model doesn't assume the company will be liquidated after the terminal year. It instead assumes that cash flows are reinvested and that the firm can grow at a constant rate into perpetuity.
'Terminal value' is calculated by dividing the last free cash flow forecast (FCF x growth rate) by the difference between the discount (d, typically the weighted average cost of capital) and terminal growth (g=3% assumed) rates. The terminal value calculation estimates the company's value after the ten year forecast period.
The selection of a discount factor is not clear, because the Spark data centres are part of a much larger Spark entity. How you determine the right discount factor is dependent on how Spark costs are allocated amongst their different operating units. But this is a similar dilemma to the Infratil/CDC scenario, where separate accounts for CDC datacentres, not a listed entity, are not fully publicly disclosed. Here is what Infratil said about this in their 30th June 2024 CDC valuation update:
https://infratil.com/news/cdc-indepe...-30-june-2024/
-------------
"The blended cost of equity used in the valuation has increased from 11.25% to 11.50% between March and June 2024. This primarily reflects an increase in gearing as a result of higher forecast debt levels as CDC continues investment in its expanded development pipeline. The increase in gearing is partially offset by a decrease in the asset-specific risk premium, driven by the valuer’s assessment of the status of CDC’s customer discussions and their overall view of CDC’s ability to deliver on its forecast growth. The risk-free rate has remained constant at 3.90%."
Risk free rate 3.90%
Asset beta 0.55
Cost of equity (blended rate) reflecting the assessed risk of the spectrum of CDC’s activity, from operating data centres with contracted revenues through to developing projects without contracted revenues. 11.50%
Terminal growth rate 2.5%
--------------------
On this basis I am declaring my 'blended cost of equity' of 12% and terminal growth rate of 3% reasonable, given that from an overall datacentre portfolio perspective, Spark are a little earlier in their development cycle compared to CDC.
(As an aside, over FY2024 we know that data centre revenue was $37m, (PRFY2024 slide 5) producing profits of $19m. If we use the same 'revenue to profits ratio', that means our forecast revenues for FY2034 will be:
$167.5m x (37/19) = $326m)
Meanwhile the terminal rate free cashflow has already been estimated to be $203.1m (refer post 2460).
Assuming that cash flows will grow at a constant rate forever, the formula to calculate the data centre's terminal value is:
Terminal value = FCF(1+g) / (d – g) = ($203.1m x 1.03) / (0.12 - 0.03) = $2,324m
But this value is calculated 11 years into the future. So to bring it back to a 'present day valuation', we must bring it back to PV using the factor: 1/ 1.12^11 = 0.2875
So the present day valuation of the terminal value of Spark data centres is: $2,834m x 0.2875 = $814.7m
SNOOPY
Data Centres: Discounted cashflow valuation: FY2024 perspective.
OK I am making a confession. As a rule I do not like discounted cashflow valuations. Why? Because there are a lot of assumptions involved. And small perturbations in those assumptions can make a large difference to the final result. So I come from the position that a discounted cashflow analysis is generally the worst method you can use to put a dollar valuation on the prospects of a high growth company unit, except for all other methods (which is why I reluctantly still use discounted cashflow analysis!).
To get the full valuation of Spark's datacentres we have to add up the totals from posts 2460 and 2461:
$19.1m + $814.7m = $833.8m
Straight away you should be able to sense a potential problem. Nearly 98% of this valuation depends on events happening more than ten years out into the future! How can there be any certainty in such a valuation? Well, there isn't certainty which is one reason why a relatively large annual discount factor of 12% has been used. I am mentioning this because although the $833.8m valuation sounds precise, it is better to think of it as quite a fluid number. But $833.8m is the number we get, so on a 'per share' basis this amounts to:
$833.8m / 1,814.2m = 46.0cps
This valuation should be considered in terms of nominal data centre earnings of FY2024 of:
$19m / 1,814.2m = 1.0cps
That means the discounted cashflow analysis is valuing the Spark data centres on a Price to Earnings ratio of 46. That sounds high. But it is probably not unreasonable, considering where we sit in the wider state of the development of data centre markets in general.
The more interesting consequence of this is that if you believe that:
a/ Spark shares currently trading at $3.60 represents a massive discount to fair value in terms of their dividend stream AND
b/ The actual contribution to that dividend stream from Spark data centres as of today is next to nothing. THEN
c/ It means that you are buying into the potential value of Spark data centres at no cost. IOW you are buying those Spark data centres, for free!
This is how I like to organize my 'tech' investments: Buying a promising but uncertain tech investment for no money. If it all works out for we shareholders, then we shareholders can look forward to a useful gain in the downstream reflective Spark share price. If it all turns to custard, well we bought in for nothing, so we haven't really lost anything. It all makes for a pretty attractive risk/return equation.
SNOOPY