Datacentre profitability (FY2024 perspective)
Quote:
Originally Posted by
kiwikeith
I do wonder if the market is taking account of Spark's data centre potential. The development pipeline could increase total capacity from 22MW to 93 MW.
I have just had another look at the HY2024 presentation material from Spark, kiwikeith. Slide 11 on "Data Centre Growth Ambition" is of particular interest.
"• Targeting returns of ~9-10% over time as utilisation scales."
And what does 'returns' mean in this context?
Quote:
Originally Posted by
Snoopy
10/ Spark is looking for a 'return' of 9-10% ('return' meaning, a 'net operating profit after tax' over 'invested capital') on data centres once operating at capacity. Third party partners may be brought in to help fund these data centres. A 9-10% return ensures that the IRR is above WACC. The signed customer contracts include an inflation adjustment element This is a higher return that global hyperscale data centres are generating. But remember, global hyperscale is on a 'next level' compared to Spark 'hyperscale'.
Slide 6 from HYP2024 suggests that data centre revenue for Spark is on track to be $35m for FY2024. Exactly how much of that revenue flows straight to the bottom line (to NPAT) is another matter.
Spark's 'flagship' datacentre opened in Takanini in 2014 with a mere 2.3MW capacity (albeit still significant back in the day) and at a cost of $60m.
https://aecom.com/nz/projects/projec...i-data-centre/
A major update has been completed, lifting capacity at that site to 12.3MW, out of a total Spark built capacity to date of 22.3MW. 22.3MW is up from a total built capacity of just 10MW at EOHY2023 one year earlier. The built capacity is continuing to increase. But for the purposes of this exercise, I am going to assume that the total billed load for Spark datacentres amounts to 22.3MW per year over FY2024.
Spark noted that from 8th April 2023, they had budgeted between $250-$300m to expand their data centre capacity over three years.
https://www.datacenterknowledge.com/...5g-in-new-plan
That is what it will take to complete a 70MW expansion of the data centre portfolio, with both greenfield and brownfield additions and amounts to a cost of $3.57m-$4.28m per MW. If I backdate the 'higher cost bound' of that 'cost range' across the 22.3MW of data centres fully built, I get an historic cost to build of $4.28m/MW x 22.3MW = $95.6m. Lets add a 20% odd factor of safety to that total (because I suspect on a per MW basis historical build costs were higher) and call it '$120m spent on data centres to date', in round figures. We know that Spark estimates they will earn 9-10% on their initial outlay. I will call it 9% because 10% is when the equipment is fully utilised, (and we know even with Sparks diligent pre-selling these centres are never 100% utilised on day 1) . So expected annual earnings (NPAT) from the Spark data centre portfolio over FY2024 is to be: 0.09 x $120m = $10.8m.
If the data centre revenue prediction of $35m is accurate, this represents a net profit margin for data centres over FY2024 of: $10.8m/$35m= 30.9%
SNOOPY