The impact of Artificial Intelligence and the Permian Basin opportunity.

In this essay we address:

1. The state of the concentrated US stock market, the factors that got us here, and the assumptions on which valuations are predicated.
2. What must occur to meet the expectations and the interrelated conditions.
3. Business services required for consensus projections to be realized.

Finally, we weigh which aspects are more or less likely to occur and explain why we own certain holdings. These themes are intertwined so we will start with why we feel you may want to swap mega cap tech stocks for land, water & power companies.

PROJECTED EARNINGS GROWTH

Mega cap tech stocks are expensive by every historical measure. This is in part due to the belief they will be a major beneficiary of Artificial Intelligence (AI). 2025 Consensus estimated earnings increase for this group range from 16- 18%.¹ Few have adequately explained how it will actually occur, only that it will.

We will start with cloud computing, which is fundamental to facilitate AI. The implementation of AI should lead to cost savings and higher earnings which support the lofty valuations. Cloud computing requires data storage. Data storage requires data centers. Data centers require power, water and zoning that allows them to be built. In order to meet projected earnings, companies need all of this to happen quickly, affordably, and at scale.

CLOUD COMPUTING

The business of cloud computing is an essential facilitator driving efficiencies for mega tech companies. MSFT’s Azure competes with Amazon Web Services, AMZN’s most profitable business, which compete with Google, Meta, Netflix, and others. Cloud, artificial intelligence, and GPU chips enable incredible computational speed. In order to have AI, companies must be able to store and access data.

Tech companies are obsoleting their data centers in urban areas because local utilities don’t have the capacity to provide the increased required power. Dominion Resources, as one example, admitted they don’t have the power to meet potential demand, and if they built it, it would require an annual rate increase of 15% per year for the next 15 years.² This is unacceptable to the local constituents. Microsoft recently said they are no longer “chip constrained,” they are “power constrained.”³ At any price, the lack of data storage to meet escalating demand is a big problem.

DATA CENTERS REQUIRE POWER & WATER

The most important thing to generate power is access to water. Almost every power plant is next to a river because of access to water. Whatever energy source is used, it boils water and the steam turns turbines, which creates electricity. To generate 1 megawatt (MW) of power, a utility boils 500k gallons water/minute. Much can be reused as the steam is recondensed. Some water is too hot and must be mixed with colder water. If it is reintroduced directly into a river at an elevated temperature, it creates a thermal plume where the heated water changes aquatic biology. That’s why utilities have “cooling ponds.”

The total amount of electric power generation in the US has not increased in 15 years.⁴ Since the early 2000s, power has stayed the same because manufacturing was outsourced to other countries. The amount of power required to facilitate the use of AI and the needs of data centers over the next five years, by many estimates, will triple.⁵ We believe in ten years it could increase by a factor of ten or more.

Like utilities, data centers require water to heat and cool. More and more data centers use “source” water which must be cleaned before it is used. Ground water can be consumed, source water is brackish, like brine, salty. (Intake of sea water is not possible because it hurts marine life and disrupts the shallows in the wetlands.) Filtering and recycling, “handling”, of water is a rapidly growing business.

TOTAL ADDRESSABLE MARKET OF DATA STORAGE

There are 5400 data centers in the US today.¹ It is estimated that the number of data centers needs to double in the next five years. The amount of energy each new data center will consume could be 10-20 higher. The average data center is 300megawatt MW, or 1/3 Gigawatt GW. In the future, they will need to be 4 GW each. The amount of power needed in the future is almost unimaginably large.

We believe that utilities are not in the position to supply the power. MSFT and Constellation Energy are reopening Three Mile Island nuclear power plant, but not until 2028, due in part to the large water needs required and regulatory red tape. Nuclear reactors need water also because the reaction itself takes place in water to slow down the neutrons which collide to initiate the chain reaction.

To make matters more challenging, data centers must operate 24/7 and always be “up.” To have a 2 GW power plant you need extra capacity to be able to take down a plant for maintenance. It may not be double, but it is some multiple higher than 1. This is not fully understood, nor priced into market expectations, in our opinion.

COST OF DATA CENTERS

The cost of building a large-scale data center is enormous.⁶ For one data center campus, you must lease 2000 acres of land. Then you need to build the data center, the power plant, and a water filtration system. Every Gigawatt of data center storage costs $30-40m in annual rent. On 6/28/24, David Capobianco CEO of Land Bridge on CNBC said if the data center is 10 GW annual rent could be $350m.⁷ For those not familiar, renting space at, or operating a datacenter, comes down to what you pay for electricity. The more energy you use, the higher your costs. Companies that own their own data centers and power plants will still need land, water, and fuel.

2000 acres is the approx. size of a data center campus and there may be many data centers on a campus – perhaps tens of gigawatts on a campus.⁸ Facebook 1.5G data center is 7.5m square feet. In two years, the size of the data center will not be any bigger, but as they install faster GPUs, the output will likely double, so the amount of power will go up, and with it, the water requirements.

RISING ELECTRIC BILLS OR LIMITS ON USAGE?

Two challenges exist. If industrial users use traditional utility power and force the utilities to build massive new plants, residential rates go up. Conversely, if the utilities lose the industrial baseload revenue, because they built their own powerplant, the cost is spread over fewer people and residential rates go up. The utilities need tremendous baseloads to spread out costs.

Alternatively, if industrial users need the power and the utility doesn’t build more capacity, the utilities could curtail retail use. That is to say your bill stays the same, but your usage is controlled. The utility could turn your power up or down – not a good outcome for residential users.

NIMBY

Much like an airport, no one wants a massive data storage center located near them. Large scale power plants are loud and expensive. These factors make remote locations such as Canada and the Permian Basin among the most attractive locations. Northern Virginia is home of the most data centers in the US and is not granting new permits. No one wants overhead transmission lines, nor do they want massive power plants to draw water from the Potomac River and alter aquatic ecology.

SOURCE WATER AND THE RULE OF CAPTURE

If you are not familiar with the term “Rule of Capture” you may be surprised to learn that in most places in the US, you are not free to do as you please with the water that falls on your property. Local law enforcement can prosecute you for gathering water to live on. Texas is the only western state with the Rule of Capture.⁹ Texas allows you to own water that you can pump from your land without regard to how it affects your neighbor. Access to water is critical. Texas is considered by most to be the best state.

As it relates to the energy industry, tens of millions of barrels of water is disposed of every day. Much of it had been pumped back into the ground. The amount that can be pumped back into the ground is based on the pressure, and by state statute that pressure cannot increase above a certain level. Curtailment, or stopping additional water being pumped into the ground, could come as soon as two years now.¹⁰ Some acres have already hit their limit.

Energy producers must get rid of the water in order to be able to pump more oil. The cost per barrel of water handled varies. It could be .25 cents a barrel or .50 a barrel. A barrel of clean water is .50, recycled is .76 barrel. If Texas imposes a curtailment of how much water can be pumped back into the ground, “handling water” will become a significant issue.

Whether water is used to make steam engines turn, cool data centers, or to frack oil wells, the price of disposing of produced water is going up.¹¹ Right now, we are just scratching the surface. Companies competing for water could become a Big Thing.

HYPERSCALERS AND EXASCALERS

Wholesale data center electricity usage is 20megawatts to 50megawatts, enough energy to power a small city.¹² A Hyperscalers data center can use 100MW. An exascale data center can use 1GW.

Hyperscalers data center costs $2b.¹³ Microsoft is spending

$80B on data centers alone. Their cap ex is greater than their income and they don’t have money to do it.¹⁴ Blackrock and Microsoft plan to raise $100b to build data centers but much more money is needed. Other PE firms are raising funds to build data centers. A foreign investor recently committed to $20b in the US to build data centers.¹⁵

Regardless of where data centers are built or by whom, access to water is critical. The amount of Gigawatt heat to be dissipated overwhelms commercial water coolers. This is a significant problem. A regular data center requires 18,000 gallons of water per day to cool. A Hyperscale data center requires 500,000 gallons of water per day to cool it. An Exascale data center requires 1.5m gallon/day to cool it. You

could put 1000 data centers in the Permian Basin if you could filter the water. An alternative being discussed is liquid nitrogen. However, it takes a lot of energy to get nitrogen to zero.

ADD THAT TO THE WWATER NEEDED BY BIG OIL.

8mm barrels of oil is produced in the Permian Basin every day. For every barrel of oil, 5 barrels of water come up. That equals 40m barrels of water a day. 42 gal per barrel equals

1.6b gal of water a day to handle the oil. Data centers will add to the demand for producing and recycling. Mercifully, the Permian Basin data centers will sit on top of the El Capitano Reef, providing a limitless amount of source water. Texas based water companies have a total addressable market that is enormous.

AI IS COMPOUNDING GLOBAL ENERGY DEMAND

Europe has the same problem – a lack of resources and capital. If they cannot produce the water, it might mean curtailing other usages, such as the closing of farms. This may be common in Third World countries, but Parisians may not like it. Powering data centers in Europe is expensive. Natural gas (nat gas) in Europe is traded on the Title Transfer Facility (TTF). The price is multiples higher than in the US. If you don’t have access to nat gas it will be hard or impossible to operate data centers, not to mention expensive. The efficiency gains from AI, already priced into the mega cap stocks, may be very expensive to deliver.

Demand for nat gas from Asia is also very high. Japan pays about four times more for nat gas than what is sells for in the US. Nat gas could be exported if it can be piped to the US west coast. California, Oregon, and Washington states won’t allow it. One pipe in Canada does allow it, but its capacity is a fraction of what Asia needs. A few companies (Prairie Sky) have pipelines that run north south. If Canada allowed a pipeline to go east to west, a few royalty trust companies’ revenues could increase dramatically.

OPTIMAL DATA STORAGE SITES

As we have eluded, locations that have energy production ability and few people are ideal. The first location is Canada, where energy is cheap, and the temperature is cold. The Permian Basin is the other. Data that must be secure has to be in North America, and probably in the USA for national security. This becomes an investment problem not just a technology issue. Even the biggest companies are far too small to fund it themselves. This is outside a historical context. As Nvidia raises the amount of GPU output, the wattage to power it will increase.

RIPPLE EFFECT

Many companies should benefit from this trend, not just the GPU makers (Nvidia, Broadcom) and mega cap technology companies. Nvidia is clearly the winner and should be for some time. Whether or not its stock price is higher in five years is up for debate. Why? Most of their largest customers are developing a competing product. Nvidia sell at 18 times sales when “chip” makers have traditionally been very cyclical and have traded at the market multiple or less with big swings between. Curiously, Nvidia insiders sold more than $1.8B in stock a few months ago.¹⁶ The key to successful investing is understanding how to participate without buying companies that are extremely highly valued and have long term risk.

LandBridge (LB) Is a company that has 273,000 acres of land, most of which is adjacent to Texas Pacific Land. LB price went up from its IPO price of $18 to above $68 likely because it appeared that few things were as good to capitalize on this trend. In November a property became available – Wolf Bone Ranch in Reed County, near the Pecos River in Wahajalara Hub (Waha Hub) – where gas sells for zero.^{17 18} LB did a 144 PIPE sale.¹⁹ through Goldman Sachs (GS). LB went down to 60 prior to the offering as is normal. Most believe GS shorted TPL to hedge their risk. TPL went down to 1150. The deal closed and the hedging appears to be unwinding.

Over time, they will build a power plants, which needs water, power lines and people. People need homes, hospitals, restaurants. Few companies have better operating leverage. If you can attract data storage facilities, and power plants with cheap gas and face no opposition you have a good business. Land is permanent and water is permanent. Neither requires cap ex, nor faces technological obsolescence. If you own the land, would you ever sell it?

Land, Water & Power, Part II

We started this discussion with the narrative that the earnings of largest companies in the US were going to be much higher due to the yet to be fully explained benefits of AI. We then detailed the component parts required to fall into place in order for those prognostications to come true. If they do, companies that provide the services will prosper. Before we explore what might happen if the benefits are delayed, let’s review how the mega caps got so big.

ASSET FLOWS, INDEX INVESTING AND MEGA CAP TECH

Passive investing through index funds has dominated for the last two decades. A disproportionate amount of new cash flows is being invested without any thought of valuation. As of Sept 30, information tech was 31% of market cap. Actually, when you add in Meta, Alphabet, Amazon, because they are effectively tech stocks, it increased to 41%.²⁰ These are companies that compete for data storage centers.

As of December 17, 2024, the top 5%, or 25 companies, were 50% of the weighting of the SP500 index.²¹ As money comes in it buys the biggest names, pushing their price higher. Can their prices keep going up? We have never witnessed anything close to this in history.

EVOLUTION OF SP500 INDEX COMPOSITION

At the start and historically the SP500 index composition was true to sector weight contribution to the US GDP. Exxon could never get to be a 7% weighting in the SP500 index because the entire energy sector was only 3% of US GDP. Prior constraints of sector definition by contribution to US GDP have been lifted. It would have never happened if the benchmark had not become a product.

In the 1990s, the SP500 index evolved from an index and became a product. The most profitable business of SPGI is not the research or debt rating service. It is the royalty from the index, by far. The goal of index providers today is to have the most capacity to attract additional investment dollars in order to drive higher royalty fees.

This development, the “Nifty Twenty-Five”, has no historical precedent and has introduced a hidden systemic risk. This risk is widely known but, in our opinion, underappreciated. People keep buying index funds.

IMPLICATIONS TO MEGA CAP

If the mega cap companies are to meet their earnings projections, they need land, water, and power. When there is a massive “land grab” (pardon the pun), everyone wants a piece of the action. The cost of land, power, water, and labor will almost assuredly keep going up until someone won’t pay anymore. This effectively should drive down the profit margins of the mega cap tech companies to a “normal” level. If this were to happen, it could be highly problematic for the valuations of these companies and the S&P500 index.

WITHOUT POWER, EARNINGS DISAPPOINT, PE MULTIPLES COME DOWN, AND PRICES COULD TUMBLE

If the largest companies can’t get enough water and power, they will be unable to incorporate AI to the extent promised. If earnings miss by even a little, the PE multiple could shrink.

The next 10 years projected returns of SP500 range from 2% to 6%. Morgan Stanley and Shiller CAPE expect 2-3% next 10 years. ^{22 23} We know from history, the average return of the “stock market” has been 9-10% per year. Periods of time when the “market” returned more than 10% are often followed by periods when the market returns were far less than 10%. The NASDAQ’s average annualized return for the last 10 years is 19.8%. The last 5 years is 36.6%.²⁴ We could be in that “well under 10%” period now.

We own several companies that have exposure to land, water, and power. One might ask if the data storage bonanza doesn’t occur, will our companies be hurt. Some companies perhaps, and certainly short-term prices will gyrate. However, if energy companies pump less oil and gas, the price per barrel normally goes up. That may hurt the oil drillers but has a limited impact on the landowners who receive royalties based on the value of what is extracted, not the volume.

RELATED BENEFICIARIES TO AI

Other vectors are being created which may benefit. Easements, roadways, people, fiber optic cable, leases go for 50 years. Oil royalties can deliver gas in kind, selling more gas. Natural gas price is, in our opinion, why the Delaware Basin part of the Permian Basin is so attractive. Nat gas largely replaced coal in the last 15 years. However, coal plant closing have stopped because every molecule of power generating hydrocarbon will be needed. We need the plants to generate the power needed by the huge mega cap tech stocks.

We believe Texas Pacific’s fundamentals are great. With the IPO of LandBridge (LB) – strategic importance of water is becoming better understood. In the Permian Basin, natural gas as a byproduct of oil production. In some regions, there is no access to pipe it out and nat gas trades for zero. Currently, these companies get limited revenue from produced natural gas. If they could get paid for it, say $2.00/btu, it could be very profitable.

LandBridge is well suited to take advantage. TPL and LB are very similar. A joint venture was put together near where New Mexico and Texas meet. New Mexico has lot of fracking but the fracked water, due to New Mexico law, is shipped to Texas to be processed. Having them align, one has to cross either LB or TPL land.

DIVERSIFICATION AND REBALANCING

If you own stocks that could generate rising cash flow for decades, let’s discuss how it could be handled within the context of a portfolio. A portfolio of 30-40 stocks will have returns that should be “normally distributed.” That means most of the company’s returns will be average, a few will do

terribly, and a few might do really well. Three standard deviations from the mean will be 2-3 stocks in a 100- stock portfolio. In  our  lifetime,  we  will  be  lucky to find and recognize 4-5 companies like that. They are easy to identify years after the fact, nearly impossible to identify before the massive gain. Even more difficult is resisting  the  urge to sell and claim victory.

In our opinion, if one is lucky enough to find one or two, be very reluctant to sell it. It borders on arrogance for me to believe that I will have another investment that is so much better than the “10 bagger,” to sell it, pay the tax, buy another, and believe it will outperform. The probability of being able to do that is effectively zero.

This begs the question, if I own an SP500 index fund with a 75% long term capital gain, should I keep it or trade into one of these companies? It may depend on one’s time horizon and overall cash flow. If your time horizon is less than 10 years, probably not.

IS MISPRICING POSSIBLE?

In an age of immediate access to unlimited data, could there be companies whose businesses will increase dramatically, and yet their share price does not reflect that outcome? Yes, because people use measurements that are familiar and easily accessible. Conventional metrics are not always the best measurements. As industries evolve, traditional measures rarely embrace qualities of unique companies. (Think about the cable company empire that John Malone built by using debt and the free cash flow to pay for it.)

Our goal is to own shares in companies which are likely to benefit from trends that may last for decades. We want to avoid owning shares in companies whose stock prices are highly valued and risk limited appreciation, even if the economy does well.

A Golden Age of true active management could just be re- emerging. We own many of the companies mentioned in this essay.

Christopher F. Poch January 9. 2025

Important Disclosures

1 https://www.statista.com/statistics/1228433/data-centers-worldwide-by- country/

1 https://markets.businessinsider.com/news/stocks/mega-cap-tech-stocks- earnings-chart-magnificent-7-nvidia-ai-2025-1

2https://www.google.com/search?sca_esv=d5800e4590f8577b&sxsrf=ADLYWI

KwHs6QSJthh-LuoqtIuH- pjrAz0Q:1736295380466&q=dominion+resources+will+need+to+increase+rates

+by+15%25+for+15+years+to+accommodate+data+center+power+needs&spell= 1&sa=X&ved=2ahUKEwjjv4Kr7OSKAxXRKFkFHQAyNMAQBSgAegQICxAB&biw

=1480&bih=771&dpr=1.5

3 https://www.cnbc.com/video/2024/11/19/watch-cnbcs-full-interview-with- microsoft-ceo-satya-nadella.html

4 https://decarbonization.visualcapitalist.com/animated-70-years-of-u-s- electricity-generation-by-source/

5 https://www.mckinsey.com/industries/private-capital/our-insights/how-data- centers-and-the-energy-sector-can-sate-ais-hunger-for-power

6  https://310value.substack.com/p/data-centers-in-the-permian-basin

7 https://www.youtube.com/watch?v=yZrwMmXN-0A

8 https://insideclimatenews.org/news/27122024/new-mexico-oil-gas- produced-water-reuse/

9https://www.twdb.texas.gov/publications/reports/numbered_reports/doc/R36

1/1%20CH%20Potter.pdf

10 https://www.reuters.com/business/energy/rising-curtailments-texas- magnify-grid-storage-shortfalls-2023-10-19/

11 https://insideclimatenews.org/news/27122024/new-mexico-oil-gas- produced-water-reuse/

12 https://www.iea.org/commentaries/what-the-data-centre-and-ai-boom- could-mean-for-the-energy-sector

13 https://www.ibm.com/think/topics/hyperscale-data-center

14 https://www.cnbc.com/2025/01/03/microsoft-expects-to-spend-80-billion- on-ai-data-centers-in-fy-2025.html

15 https://economictimes.indiatimes.com/news/international/world- news/trump-announces-20-billion-investment-in-us-data- centers/articleshow/117032067.cms?from=mdr

16 https://www.gurufocus.com/news/2541389/nvidia-nvda-insiders-sell-over-18- billion-in-stock-amid-ai-spending-concerns

17 https://www.eia.gov/todayinenergy/detail.php?id=44278

18 https://www.eia.gov/todayinenergy/detail.php?id=63044

19 https://d18rn0p25nwr6d.cloudfront.net/CIK-0001995807/1d54b262-ef4e-4711- 8706-70cbe6ab10ec.pdf

20 https://insideclimatenews.org/news/27122024/new-mexico-oil-gas- produced-water-reuse/

21 https://www.investopedia.com/best-25-sp500-stocks-8550793

22 https://www.businessinsider.com/sp500-returns-outlook-decade-flat- returns-mike-wilson-morgan-stanley-2024-12

23 https://www.advisorperspectives.com/articles/2020/07/20/the-remarkable- accuracy-of-cape-as-a-predictor-of-returns-1

24 https://curvo.eu/backtest/en/market-index/nasdaq-100?currency=eur

Investment advisory services offered through Promethium Advisors, LLC, a Registered Investment Advisor with the U.S. Securities and Exchange Commission.

This material is intended for informational purposes only. It should not be construed as legal or tax advice and is not intended to replace the advice of a qualified attorney or tax advisor. These materials have been prepared exclusively for a presentation and may not be distributed in whole or in part to any other person.

The information contained in this presentation has been compiled from third party sources and is believed to be reliable. This presentation neither intends nor constitutes an offer to sell any securities to any person nor is it a solicitation of any person to purchase any securities. Past performance is not indicative of future returns.

Diversification neither assures a profit nor guarantees against loss in a declining market. Different types of investments involve varying degrees of risk, and there can be no assurance that any specific investment will either be suitable or profitable for a client or prospective client’s investment portfolio. Historical performance results for investment indices and/or categories generally do not reflect the deduction of transaction and/or custodial charges, the deductions of an investment management fee, nor the impact of taxes, the incurrence of which would have the effect of decreasing historical performance results.