IFT - Infratil

[Ferg]

Snoopy - I am not interested in debating this with you for 3 reasons: 1) you don't take counsel and I don't think that is about to change, 2) you are the only person I know who will argue black is white despite evidence to the contrary and 3) your argument about NPAT is a 'strawman' argument given no-one here (other than you?) has said NPAT is an indicator of operating profitability.

Some things for you to think about given my previous anecdotes went over your head:
1. Can land value change when you change its use? (hint: answer = yes, but this often requires capex)
1a. And would such revaluations form part of NPAT but not EBITDAF? (hint: yes)
2. Can interest costs for long term capex projects be capitalised as a cost of that project? (hint: yes).
2a. And are such capitalised costs excluded from the P&L and also NPAT? (hint: yes)
2b. Are interest costs the only costs incurred for such long term projects? (hint: no)
3. Is CDC undertaking long term capex projects? (hint: yes)
4. So is it reasonable to assume that interest and other costs are not showing in the P&L for good reason? (hint: yes)
4a. Or do you think you know more about the CDC financials than the Accountants and Auditors despite possessing less information than them? (hint: no)
5. Why would someone calculate their own EBITDAF when the actual EBITDAF is disclosed on page 25 of AR2024? (see 4a)
6. Lastly, how do you solve a mathematical equation with multiple missing variables?

As I said earlier, better to ask "what am I not understanding?" than to go off half-cocked and claim the numbers are wrong or not 'real'.

Hog: IMO this is a better approach - first ask about the bits you don't understand and then develop & present your analysis. Not the other way around.

Let's have a look at these 'real numbers' shall we? AR2024 p80, referred values for FY2023. CDC Revenue for FY2023: $345m. Net Profit After Tax for FY2023 $762.7m. Now I am not saying the numbers aren't correct. But they obviously are not related to the operational performance of CDC. How can the net profit of a company after all taxes and expenses are taken out equal more than twice the revenue taken it? If it were not obvious before that the declared net profit at CDC was not related to operational profit, then surely this is the proof?

Now go back to slide 27 of the capital raising presentation. CDC EBITDAF for FY2023 was $215m. Take out Depreciation and Amortisation, net interest charges and tax and you will end up with a net profit after tax much lower than $215m from an operational perspective. However the actual CDC NPAT for FY2023 was $762.7m, more than three times higher than the published EBITDAF figure!

The disclaimer notes of the presentation offer an explanation:
"EBITDAF represents net earnings before interest, tax, depreciation, amortisation, financial derivative movements, revaluations, gains or losses on the sales of investments, and excludes acquisition and sale related transaction costs and International Portfolio Incentive."

It seems obvious to me that the net gains from these other sources I have highlighted in bold have absolutely monstered any gains from operational performance. And from the description of these items, it is equally obvious they are one offs and not expected to be repeated. So to claim that a CDC profit figure boosted by 'financial derivative movements', 'revaluations', and 'gains or losses on the sales of investments' is proof that managing data centres is a very profitable business activity seems to be a gross distortion of the word 'proof'.

SNOOPY

[Ferg]

Snoopy - the CDC accounts are not much of a mystery at all. I gave you the necessary references:

Known CDC values are for FY2024 per AR 2024 p80 and A$m unless otherwise stated:

• EBITDAF NZ$292m (per p25) {edit: this translates to A$271m @ 0.9271}
• Revenues $412m
• NPAT $202m
• Other comprehensive income ($12m)
• Drawn debts $4,055m
• Non-current assets (these are the income producing assets subject to D but will also include capital WIP) $6,666m
• year end balances revalued using a spot rate of 0.9181
• trading values revalued using an average rate of 0.9271
• p5 "EBITDAF is an unaudited non-GAAP measure of net earnings before interest, tax, depreciation, amortisation, financial derivative movements, revaluations, and non-operating gainsor losses on the sales of investments and assets."

Let's assume there are no non-operating gains or sales of investments or assets. Asset revaluations would be part of 'revaluations'. I will wrap 'A' into 'D' given both relate to non-current assets. Financial derivative movements are IFRS adjustments for things such as currency hedges and/or interest rate hedging instruments. Such instruments usually pertain to debt so we can look at I & FDM together. Interest 'I' in the NPBT calculation is non-capitalised interest i.e. 'I (P&L only)'. Capitalised I is part of the capex numbers.

We know:

• EBITDAF = Revenues - Opex
• NPBT = EBITDAF - I +/- FDM - D +/- R
• NPAT = NPBT - T
• TCI = NPAT +/- CI

IFT's share of CDC values further detailed as follows (NZ$m per p80):

• share of NPBT $156m
• share of tax ($51m)

So for CDC we can derive:

• Opex = A$141m (412-271)
• Tax = A$98m (0.9271 x NZ$51/.482) - note this will include deferred tax given FV items are generally non-taxable
• NPBT = A$300m (which is A$202 + A$98 OR it is 0.9271 x NZ$156/.482)
• Total comprehensive income = A$202 - A$12m = A$190m
• I +/- FDM - D +/- R = +A$29m (being $300 NPBT - $271 EBITDAF)
• R cannot be known but we can get the value of (R - D) by looking at the annual change in non-current assets (p80 says +A$904m) and deducting known capex of (p26 0.9271 x NZ$292/0.482) A$561m leaves a value for R - D of +A$343m. This assumes there are no other transactions impacting non-current assets such as an acquisition of a new subsidiary by CDC which could muddy the waters if there was such a transaction.
• EBITDAF + (R - D) - NPBT = A$314m which gives the value of 'I +/- FDM'
• FDM is the value of Financial Derivative Movements which mostly pertain to debt
• A$314m / average debt of $3,741 gives an implied debt cost of circa 8% (note I say 'debt cost' and not average interest rate.....debt cost being I + FDM). 8% costs assumes no interest is capitalised. However, if some interest is capitalised as part of the capex number, then the average debt cost will be higher - note this is more than the circa 7% IFT get on their advance to CDC but that would be expected if there are FDM losses. This 8% also assumes a gradual increase in total term debts; if however a graphed monthly profile of total debts drawn during the year was either a U shape, or an inverted U shape, then that would change the average debt cost.
• Average MINIMUM debt cost of 8% seems a touch high, especially if holding costs are being capitalised. In the event interest costs are not being capitalised then IFT need to have a word to CDC about correctly accounting for Capital WIP costs. Or it could be there are some financial instruments that are negatively impacting CDC.

If you want to solve for FV or D then I recommend present a range of values based on varying assumptions of depreciation rates as well as assumed splits between capital WIP vs income producing assets. Same for solving I vs FDM - present a range of values. But right now there are too many unsolved variables to meaningfully unpick these two unknown values any further.

It could be the USPP debt is costing CDC a lot. So how much of the IFT capital raise will be used for debt repayment versus capex....have they disclosed this?

[Ferg]

"CDC metric" of:

"A$2m+ EBITDAF / ICT MW p.a. (1)"

is techno-babble to confuse shareholders and get shareholders to believe those Infratil head office boffins are so much smarter than they really are. I would still love it though, if anyone can decipher what it is that Infratil are trying to say here.

EBITDAF / ICT MW = A$2m
EBITDAF / A$2m = ICT MW

FY24 EBITDAF NZ$292m (AR24 p25) converted at 0.9271 (AR24 p80) = A$271m (confirmed on CR p27)

ICT MW = A$271m/$2m = 135.5
Total capacity = 302MW

ICT MW / Capacity = 135.5 / 302 = 45%

Rack utilisation = 89% but note says "Including white space and reserved."

So that leaves 44% white space and reserved but we don't know if 89% is a point in time value or an annual average.

The 135.5 is an annual average for FY24. So the 45% average could be say 35% growing constantly to 55% at year end. Too many unknowns.

So CDC is dependant on the building / development profits while they increase utilisation percentage.

[Snoopy]

"A$2m+ EBITDAF / ICT MW p.a. (1)"

is techno-babble to confuse shareholders and get shareholders to believe those Infratil head office boffins are so much smarter than they really are. I would still love it though, if anyone can decipher what it is that Infratil are trying to say here.

EBITDAF / ICT MW = A$2m
EBITDAF / A$2m = ICT MW

Firstly good on Ferg for putting his oar in to help decipher this. If Ferg's interpretation is right, and I am not saying it isn't, it highlights a grammatical error in the results presentation which has been throwing me off.

Consider the presentation wording:

"A$2m+ EBITDAF / ICT MW p.a.
(CDC currently generates a blended EBITDAF per ICT MW across all sites and customer segments of over A$2.0m)"

with this wording (which if Ferg is right is I believe what Infratil wanted to say):

"A$2m+ EBITDAF / ICT MW p.a.
(CDC currently generates a blended EBITDAF per ICT MW across all sites and customer segments, of over A$2.0m)"

The only difference is a single comma towards the end. Yet that tiny change completely alters the meaning of the text.




Original Text Meaning

The first text is suggesting that there are two broad categories of data centres under discussion.

1/ Those that bring in less that $2m per year (probably the original legacy sites that were no doubt much smaller than the mega data centres that are being built today). AND
2/ The aforementioned mega data centres of today.

It is quite clear there will be 'scaled efficiency gains' across the more modern sites. So the profitability of those more modern sites, on a per MegaWatt consumed basis, is likely to be higher. Thus when as a data centre provider, when you are looking at projecting your profitability into the future it makes sense to look at your profitability achievements of the recent past, not the much older stuff. So it makes perfect sense to look at what EBITDAF on a per MW basis across a blended collection of the more modern sites might be. But then the presentation doesn't say what that EBITDAF/MW figure might be! I had assumed this was for competition reasons (you don't want your competitors to know exactly how many dollars you are pulling from your latest installations). But then I thought, why are Infratil going to the bother of explaining all this, when the aren't going to tell us the answer in the end? Thus I concluded the original statement was gobbledygook and the result of 'nothing to do analysts' shining their bums on their office chairs, giving Infratil shareholders the impression they were telling them something useful, when they weren't.



Revised Text Meaning

In the revised text, all installation sites be they old or new, small or large, are blended together. Infratil/CDC know the 'EBITDAF they are taking in' and they know the collective power consumption of their data centre plants. So they can take all of this collective information and produce a 'whole of portfolio' EBITDAF return on a per MW of capacity used basis. Quite why they would want to do this and hang out their internal cost structure metrics for all their competitors to see I am not sure. I believe this is why I did not think to interpret the presented information in this way the first time I read it. But if Infratil are so keen to tell shareholders that they are earning EBITDAF of $2m for every 1MW of electricity consumed, what do they expect shareholders to do with this information?

a/ Stand up and applaud?
b/ Congratulate management for turning green energy into carbon responsible profits?
c/ Castigate management for wasting precious energy, and turning it into meaningless AI babble?
d/ Shame management for getting a much lower return per MW consumed than the likes of Spark do with their datacentres?

What are shareholders meant to garner from a presentation metric like this? I am really not sure.

SNOOPY

[winner69]

Suppose one day with all the energy needed to run data centres and supporting growth in AÍ nuclear will be way to go to make it happen.

Apparenty some US sites agave, or about to have, there own nuclear generator onsite

[kiora]

IFT now clearing a runway to $15 SP?
May be need to clear 1 or 2 business sales first to recycle capital?
If they can achieve 1 or even better the second one as well then?

Posted 6-4-24

OK ,they done a massive,well supported, capital raise instead
By X-mas? Just not sure which Xmas

Some investors make decisions on why they shouldn’t invest in a company while others make decisions on why they should invest in the same company

It's a dichotomy with two completely different outcomes for investors.

One group has achieved very credible compounding returns over 30 years.
And the compounding returns are highly likely to accelerate

While the other group……..

[Snoopy]

EBITDAF / ICT MW = A$2m
EBITDAF / A$2m = ICT MW

FY24 EBITDAF NZ$292m (AR24 p25) converted at 0.9271 (AR24 p80) = A$271m (confirmed on CR p27)

ICT MW = A$271m/$2m = 135.5

I am afraid there is a bit of a problem with Ferg's calculation above. Not in the calculation itself, the numbers look right. But the equation he has derived is not portrayed with dimensional accuracy. What do I mean by this?

If you have an equation with dollars on the right side, you can add or subtract those numbers up to get dollars on the left side. IOW dollars beget dollars.
Likewise you can manipulate a series of numbers depicted as MegaWatts and come out with an answer in MegaWatts.

However, what you can't do is start with some numbers in dollars and somehow convert that into an answer in MW. That is like adding up a pile of apples and calling out the total as oranges. So that last line in Ferg's calculation is wrong, and should actually read like this:

ICT MW = A$271m / A$2m per MW = 135.5 MW

This way the units balance (they are MegaWatts all the way through, as the A$s units cancel out), AND we get to see that the equalisation factor that we are meant to go all gooey over is "Two million dollars per MegaWatt" (not "Two million dollars").

Now, what shareholders are meant to make of this I am not still not sure. I would have thought that if shareholders could derive more income from less power burned, then that would be good. But it sounds like Infratil are promoting "$2m per MW" as a semi-immutable reality, and are leaving shareholders to decide whether this is bad or good. The only way to make relative sense of this is to look across at what other data centre providers might do with a similar power resource.

SNOOPY

[Snoopy]

Let's look at this calculation in a bit more detail now. I have added some extra explanatory information in bold.

EBITDAF / ICT MW = A$2m
EBITDAF / A$2m = ICT MW

FY24 EBITDAF NZ$292m (AR24 p25) converted at 0.9271 (AR24 p80) = A$271m (confirmed on CR p27)

ICT MW = A$271m/ A$2m per MW= 135.5MW

Total built capacity = 302MW (as at the end of the reporting period)

ICT MW / Capacity = 135.5MW / 302MW = 45% (based on end of year capacity)

From AR2024 p32:
"Performance highlights include delivering a record 200MW in new contracted capacity (includes reservations and rights of first refusal), its largest ever addition in 12 months"

This implies the CDC data centre capacity at EOFY2023 was just 102MW. Taking a linear average through the year would suggest an average annual data centre capacity of 202MW ( [102MW+302MW]/2 ).

ICT MW / Capacity = 135.5MW / 202MW = 67% (based on linearly averaged over the year capacity)

That sounds better than a 45% utilisation rate doesn't it? (or does it?, see last section of this post)

Rack utilisation = 89% but note says "Including white space and reserved."

So that leaves 44% white space and reserved but we don't know if 89% is a point in time value or an annual average.

This is the way I see the new build of a CDC data centre working, given they seem to be largely in the private cloud space. A green field site will be quickly internally fitted up with rows of interconnected cabinets all computer ready so that clients can 'fill the racks' with their own computing gear. Exactly the number of clients that put their gear into these cabinets on 'day 1' is not so important. From a fit out perspective, it would make much more sense to have all cabinets customer ready on the day the data centre opened. I am thinking of an unconnected by empty cabinet as 'white space' in this discussion context. This would lead me to think that the percentage of rack space utilised, while it is a measure of how quickly a particular CDC centre might be filling up, is not necessarily a statistic of particular concern (If CDC did not expect their individual cabinets to fill up in reasonably quick time, then they would not have built them).

The 135.5 is an annual average f)or FY24. So the 45% average could be say 35% growing constantly to 55% at year end. Too many unknowns.

So CDC is dependent on the building / development profits while they increase utilisation percentage.

With respect, I don't think this is how the data centre business model works. Businesses come to CDC with a data centre need. They agree to sign up to a certain computing processing capacity in kW or MW with CDC to fulfill that anticipated need. Thus CDC gets to be the counterparty to a series of contracts which add up to a total in MW that sums to the capacity of the data centre (100% data centre utilisation, at least in theory). However in practice not all clients would need all of their data processing capacity all of the time. For example some companies might run a small local server that 'backs up to the cloud' overnight. That means despite having signed up to 'x'MW of processing capacity, they may only 'need' that for a few hours a day. That particular customer cabinet at CDC could be turned off most of the time, saving CDC money.

I expect CDC to be profitable at the individual data centre contract level, no matter what the actual utilisation rate of their facilities is. But it strikes me that the most profitable customers might be the ones that sign up for x MW of data processing but then don't use it much of the time. This would show up in the data operating statistics as 'underutilisation'. But because certain customers pay for utilisation and then don't use it all, all the time this is actually a good thing for CDC. Because CDC get to bank the money, but they have reduced consummate operating expenses. Consequently I would suggest that a lower power utilisation percentage at a CDC data centre would likely be more profitable for CDC. Not less as you suggest. But then I don't work in a CDC data centre. So what do I know?

SNOOPY

[kiora]

From your analysis Snoopy then you would prefer to invest in Spark then?

https://finance.yahoo.com/quote/SPK....xpZXIiOjF9fQ--

[Snoopy]

From your analysis Snoopy then you would prefer to invest in Spark then?

Between SPK and IFT it is a case of 'horses for courses' kiora. Both companies have gone through huge changes in the make up of their businesses, particularly in the last ten years or so. Capital growth has been better at Infratil, dividend flow better at Spark. Both companies I think have been exceptionally well manged over the last 10 years or so. And right now, both find themselves at that confluence point of being pivotal players in NZ's telecommunications industry. I honestly hadn't realised how much Infratil had moved into that telecommunications space, with 56% of all assets under management now One NZ and CDC. And that industry percentage asset allocation is predicted to grow.

With datacentres being key strategic growth planks for both companies, I am interested in how the datacentre business works in practice. I find I can understand a lot more how a business works if I can study the competition. Hence my 'sudden interest' in Infratil. I give absolute respect as to how management has grown Infratil over the last ten years. And I fully understand how investors who have ridden that wave of success, such as yourself kiora, are quite happy to sit back and let Infratil 'do what they do'. Yet I have a hankering doubt of uncertainty that 'things are different this time.'

Infratil has a long history of exceptional management of key strategic assets. But datacentres don't quite fit into that category, because they are a brand new class of strategic asset. What infrastructure investment has Infratil been in before where the international behemoths, like AWS and Microsoft Azure, have the ability to come in and swamp the market with capacity? How will the market interplay, play out? I have to admit I swallowed the line that these giant hyper players would come in and take over and the existing bit players would be squeezed out blah blah blah.... Since then I have learned more about 'private cloud' and 'hybrid cloud' which seem to be game board spaces that the international hyperscalers don't want to move into. I now realise the datacentre market is far more fragmented that I believed and that there are 'niche opportunities' for the likes of both Spark and Infratil to exploit.

My reading of the developing datacentre market is that Spark have the more conservative strategy. Customers have to be lined up and committed before serious build money is put into any Spark datacentre project. But the Spark approach, with the company footing the bill for much of the equipment that ends up on the data racks, is more capital intensive. Infratil running a lower capital cost private data model, is more a 'build it and they will come' market player. But if they don't come, well Infratil haven't actually built that much! My feeling is that Infratil, via CDC, is running a higher risk higher reward game. But with a developing market like datacentres, who really knows how it will develop? I certainly don't, which is what I am 'here' (on the Infratil thread) trying to gain as many market insights as I can.

I think Infratil is at a crossroads, entering a dynamic market of a type they have never experienced in their thirty year history. Some who have ridden on the Infratil pigs back over the last 10,20 30 years are happy to ignore what is going on, and back management to 'do their thing' as the good ship Infratil sails on to new heights, in the lee of their past success. But I think new investment waters are potentially troubled investment waters. Best to keep an eye out for those potential investment rocks as the good ship Infratil sails along....!

SNOOPY