Boiling Point - An Investigation Into Water Profits.
The overlooked water economics behind AI infrastructure — and the royalty empires poised to profit.
Welcome to the ROI Club.
It will be recalled that one of my key investment themes is the exponential growth of AI data centres.
My next few pieces will explore this theme and how I am positioned to express this thesis. Before I begin I should give you the following warnings:
Historically, investing around new technologies has been a losing gambit.
There are many smart people who disagree with me on this.
I will attempt the use of metaphors in this piece and make absolutely no guarantee of either literary, or investment success.
I will attempt calculations using what I call ‘ ‘Murican metrics’.. referencing mysterious measurements like gallons, barrels and other highly uncivilised barbarisms.
I am spectacularly unqualified in physics, consider me a humble entrepreneur with halfway-decent instincts writing to you from an overly large island in the south-pacific.
If you’re silly brave enough to listen more, I’ll flesh out what I consider the sine qua non of the advent of data-centres and their requisite power centres: water.
But first, a slight literary detour..
A favourite detective story of mine is‘El crimen casi perfecto’ by the Argentine author Roberto Arlt.
In the story, a murder occurs where the culprit ingeniously chooses a poisoned block of ice as the murder weapon. The genius of course is that the ice is hidden in plain sight and upon changing state, vanishes and thus taking with it all the evidence.
As I’ve been learning about power generation and the water loss associated through steam powered turbines the thought occurred to me that perhaps just as the change change in water’s state hid the evidence in plain view in the story, the water’s state change in power generation is obscuring profit potential from the view of uninformed investors.
Power and Water: The Hidden Economics of AI Infrastructure
High-order computation, be it AI, hyper-scale data centers, Bitcoin mining etc. requires an unbelievable amount of electricity.
This much is fairly well known.
Precisely because it’s well known people are fixating and speculating on what the input power source will be to provide the electric power to run these functions i.e: coal, natural gas or nuclear, in the belief that this will be the choke-point and therefore the most profitable aspect.
No doubt there will be opportunities in the aforementioned commodities but, in reality, all three are thermal generation methods which depends on one core input: water.
Here’s a highly sophisticated flow chart I prepared earlier:
Heat water → generate steam → spin turbines → create electricity.
The energy input to heat the water can vary but the water is irreplaceable.
No woman - no cry, no water - no power.
Here’s the part that is truly astonishing.
I have it on good authority that a 1 GW power plant circulates over 5 -10 Million gallons of water per day. Not all of it is consumed obviously, but a significant amount evaporates. Depending on the source, a 1 GW plant might consume 114,000 BBL/day.
(https://pubs.usgs.gov/sir/2019/5103/sir20195103.pdf)
(https://horizonkinetics.com/app/uploads/Horizon-Kinetics-Q4-2024-Commentary_Final.pdf)
Now, unfortunately due to factors like redundancy and equalising the power from the power plant to the data centre, I’m told in reality the above number needs to be multiplied by 3.08. So let’s call it 351,000 ish BBL/day vaporised in the power production alone.
Then of course, the data centre needs to be cooled and that alone requires ~150,000 BBLs/day according to Landbridge CEO David Capobianco.
Ergo, it’s entirely possible that the emergence of hyper scale data centres brings with it a demand of ~350-500,000 BBl/day of water just to replace the vapour lost through daily operations.
As rudimentary as this overview may be I think it’s reasonable to begin a working model on the premise that a 1 GW data centre and its associated power plant may need at a minimum 350-500K BBL/day to replace the water lost through vapour and cooling.
One may call me crazy, yet one may also wish to ask themselves a few questions:
How much might the water be sold for?
Who has both the water, the handling infrastructure and the transport capacity to supply these centres?
What might the associated fees for this transport and infrastructure be and what is that kind of recurring revenue worth?
What other requirements are needed to host such a centre and who has them?
By now you’ll know that I see TPL and LB as the companies in pole position to capitalise on this zeitgeist and I shall be delving deeper into what the potential water revenues may be for these two companies which, by the way, also includes the sales of sourced water and the handling of produced water used in fracking for oil which forms its own revenue stream independent of what happens with data centres.
For now, I leave you to ponder the potential of handling and supplying 350k BBl/day of water at various tariff rates.
At $0.50 per BBl/day, which LB estimates as the ave fee, = $64 million annually.
This, it should be recalled, is for one centre and before easements, lease fees and all that other good stuff.
If one values this at a 5% cap rate due to its indispensable demand and recurring nature, it might be fair to argue that 1 GW centre might be worth over $1 Billion for the water alone before even considering what might happen in the event of an increasing tariff rate above $0.50.
So to come full circle, just as the detectives missed the change of the water’s state in their investigation, investors I humbly believe, are missing the profits from the water value in their current calculations of TPL and LB.
Hopefully you enjoyed this introductory piece into my wonderings on water, up next will be a further look into Landbridge from a different lens.
Quick reminder grab your ticket to the Rule symposium here
Take care and do your own due diligence.
Benjamin.
PS: please share if you enjoy
Thank you for the article. I am having difficulty with your gallons and bbl (assumed to mean barrel = 42 gallons) calculations. You begin with 500,000 gallons needed for a 1GW facility. I assume this means 500,000 gallons are required to fill the vessels that get heated to generate the steam? Is this correct?
If so, you go on to state that some water is lost to evaporation during the day, with an estimate of 114,000 barrels. Now, 114,000 barrels is equivalent to 4,788,000 gallons, which is roughly 10 times the amount needed to fill the vessels at the start of the day. I am a novice, but that seems unreasonable to me. Further references to water and potential cost/value continue this theme, which makes those numbers also seem unreasonable. Where have I gone wrong?
I did enjoy your piece and have followed you in on both TPL & LB, so I have faith BUT I needed to ask with the whole DeepSeek/tech giants scaling back/ it’s all been overstated/tech probably going to drop as a rotation into value assets occurs type of thesis being fairly prominent, have you taken those arguments into consideration or ignored them or just don’t think they’re that relevant?