Welcome to Episode 136 of Special Situation Investing.
AI and datacenter growth are the new hot topics in investing circles. I hear datacenter growth discussed on quarterly earnings calls from companies you would expect like Meta, Microsoft, Amazon, and from industries as unrelated as auto part manufactures. It seems that everyone expects to benefit from a wave of new infrastructure in the next five to ten years, but how can we know if that build-out will materialize or if it’s all hype?
If the build-out of data centers does become a reality, which companies will truly benefit and to what degree will they profit? Will TPL and LandBridge provide water, land, and energy for datacenter operations, or will nuclear power and other so called carbon free alternatives absorb the increased need for energy production?
Ultimately, the answers to these questions are unknowable. After all, we’re dealing with the future and there are many ways in which the future can play out. Even if we gather and analyze all of the facts in a logical manner, we aren’t guaranteed to come up with the correct conclusions about what’s going to happen next. On the other hand, as active investors, we’re paid to analyze information and draw reasonable capital allocation conclusions. So, as messy as the future is, we have to do our best to position ourselves favorably for what’s to come.
Turning these questions over in my mind for the last several weeks led me back to the ACH process that we described in Episode 55 of the podcast. Author, Richards J. Heuer, described the Analysis of Competing Hypothesis process in his excellent book, The Psychology of Intelligence Analysis, and I highly recommend the book to you if you’d like to learn more about how intelligence analysts sort information. There are, after all, many similarities between the job of intelligence analyst and investor. Both are charged with sifting through incomplete information and using that information to draw probabilistic conclusions that direct future actions.
To help clarify my thinking, I created two separate ACH matrices. The first deals with the future of datacenters, and the second deals with nuclear energy. To keep things organized, I’ll review the datacenter topic first and then move on to nuclear energy.
Datacenters
For LandBridge or TPL to benefit from growth in the datacenter industry, the total number of datacenters must also grow. If the total number of datacenters is increased, then someone is going to have to foot the bill for the increased infrastructure. Based on current projections, the infrastructure build will noticeably impact energy requirements at a national level and will demand sums of money that only the largest publicly traded companies could ever hope to provide.
Are we to believe that an infrastructure build which challenges the capacity of the US power grid, along with the balance sheets of global tech giants, is to be financed by efficiencies created for much smaller companies? In other words, are products who’s construction costs challenge the financial resources of companies that make up the top 30% of the S&P 500 to be financed by revenue generated from the bottom 70%? I suppose it’s possible but it doesn’t seem like the best strategy. Absent another source of funding, you wonder if the promise of hyperscale data centers is more hype than scale.
Enter the only entity in the US with a bigger budget than big tech, Uncle Sam. In a sense, the government holds a royalty option on every asset in the country. By adjusting the corporate tax rate, property tax rate, sales tax rate, or nearly any other tax rate, the government can lay claim to whatever income stream it chooses. Revenue generated from taxation, and the correspondingly massive ability to spend, results in economic distortions where some industries benefit and others do not. The military industrial complex is just one example of this type of distortion, and as an investor, there is money to be made in government spending adjacent industries.
Tax incentives and subsidies are additional ways that government policy can have an outsized impact on industry development. It’s difficult to make a free market case for ethanol in gasoline, and the same could be said for wind and solar power projects. Each of these industries has benefited from a tailwind of tax incentives for decades.
My point is not to argue for or against government spending, tax incentives, or subsidies, but rather to acknowledge their impact on economics. It reminds me of what is probably an apocryphal story, where in the lead up to a Berkshire annual shareholder meeting, Warren Buffett quipped to a fellow shareholder that, “Between the two of us we run this place.” In similar fashion, when it comes to you and the US government, you run this place.
With that context in mind, consider the following excerpt from the Readout of White House Roundtable on U.S. Leadership in AI Infrastructure from September of this year:
Today, as part of the Biden-Harris Administration’s comprehensive strategy for responsible innovation, the White House convened leaders from hyperscalers, artificial intelligence (AI) companies, datacenter operators, and utility companies to discuss steps to ensure the United States continues to lead the world in AI. Participants considered strategies to meet clean energy, permitting, and workforce requirements for developing large-scale AI datacenters and power infrastructure needed for advanced AI operations in the United States.
In addition to the above announcement, the following action steps were taken at the roundtable event.
1. The White House is launching a new Task Force on AI Datacenter Infrastructure to coordinate policy across government.
2. The Administration will scale up technical assistance to Federal, state, and local authorities handling datacenter permitting.
3. The Department of Energy (DOE) is creating an AI datacenter engagement team.
4. The Department of Energy will continue to share resources on repurposing closed coal sites with datacenter developers.
5. The US Army Corps of Engineers (USACE) will identify Nationwide Permits that can help expedite the construction of eligible AI datacenters
6. Industry leaders at today’s convening committed to enhance cooperation with policy makers and explore further solutions, through ongoing dialogue and collaboration.
7. Hyperscalers at today’s convening reaffirmed their commitments to achieving net zero carbon emissions and to procuring clean energy to power their operations.
To summarize this point, there are many reasons why a dramatic uptick in datacenter build-out might or might not happen, but government partnership in the endeavor is without a doubt the key factor. The US government sees AI superiority as a national security objective and views datacenters as critical infrastructure. This public private partnership is the single greatest driver I see behind the coming industry boom. While there will certainly be commercially viable and profitable AI applications that come from industry expansion. I do not think that those use cases alone are enough to justify the massive capex that the industry is forecasting. Just as the railroads, the interstate highway system and the fiber optic network that underpins todays telecom industry need government support to develop, so too will AI and datacenter development require government support.
Nuclear Power
Establishing that a boom in US datacenter construction is likely because of its connection to national security objectives is useful, but still doesn’t guarantee that Permian Basin land will be used for the project. The final action step issued from the White House reaffirmed all participants commitment to achieving net zero carbon emissions, and to procuring clean energy to power their operations. Because the Permian Basin is a prolific oil & gas basin that is worked by petroleum companies the world over, it doesn’t seem to fit the clean energy bill. Perhaps the Permian is not the place to put your money after all?
But what happens when we take a deeper look at the situation? It’s common knowledge, that datacenter projects are already running into power constraints. Wind and solar are too intermittent to provide the baseload power that hyperscale projects demand, and coal is a non-starter politically. Both industry and governments want reliable and carbon free baseload power, which leaves nuclear as the only option.
Natural gas is far from zero carbon, and natural gas & oil are what the Permian has to offer. If we stop the analysis here, then Permian Basin land companies will probably not benefit from the datacenter boom. As you would expect, however, there is more to the situation.
First, nuclear power today takes years to bring online. The Vogtle nuclear powerplant in Georgia, is the first nuclear powerplant constructed in the US in more than two decades. The project began in 2009 and suffered multiple cost overruns throughout its 15 year construction lifecycle. To some, nuclear power is the panacea to all our energy woes, and that might well be the case. But the reality of today’s regulatory and technological environment indicates we aren’t going to solve our near term power problem with the flick of a nuclear switch. Nuclear projects today simply take too many years to bring online and aren’t a viable near term power solution.
There are other problems with nuclear as well. For starters, it’s seen as environmentally friendly because it doesn’t emit CO2. This, of course, ignores the fact that uranium mining, plant construction, and plant de-commissioning all create CO2 emissions. Nuclear power generation, as it is used today, also consumes a tremendous amount of water. Reviewing documents that advocated for the successful shutdown of the San Onofre nuclear power plant in California reveal the extensive impact the plant had on the surrounding ocean life, impacts which are not unique to that location.
Additionally, nuclear power advocates celebrate the low accident rate and minimal loss of life associated with nuclear power generation. To be fair, nuclear power does boast of a low accident rate, but the rate at which something happens is not the only factor that matters. The degree of damage that results from an occurrence is also relevant.
Using Fukushima Daiichi as an example, you can view the mishap in several different ways. From one perspective, not a single person died as a direct result of the meltdown. From another perspective, 160,000 people from within a 40 kilometer radius were evacuated due to high radiation levels, and though many have returned home, access restrictions are still in place to this day. Additionally, over 500,000 tons of contaminated water escaped into the ocean prior to the plant operator pumping contaminated water into storage tanks for filtration. In the end, over a million tons of contaminated water was stored for filtration and, as of 2023, is slowly being released into the ocean in a process that may take over 30 years to complete. Finally, millions of square meters of waste was generated in the multi year effort to decontaminate soil from the surrounding area. By some estimates, portions of the soil will be impacted for centuries.
I’m not qualified to make a judgment regarding the best source of power for society. Rather, what I’m attempting to convey is that all energy solutions represent trade-offs. There are no free lunches. If you ramp up nuclear power production from the current 440 plants globally to 4,400 plants, then nuclear accidents will likely occur ten times more often. Instead of three major disasters since three mile island in 1979, we might have had 30 accidents. One mishap every year and a half for 45 years is hardly a non-event.
Yes, nuclear power is carbon free, but it’s not without environmental impact and it’s not readily available to meet power demands in the next one to five years. Perhaps as technology improves and regulatory structures are modernized, nuclear power will pick up the load. Will that happen between now and 2030? Perhaps, but I’m not betting on it.
By contrast, according to the EIA, a natural gas power plant takes 12-18 months to complete. Fourteen new plants were added with 1.9 gigawatts of capacity in 2022 and 2023 alone, and an additional 2.9 gigawatts is scheduled to come online between 2024 and 2025. But the speed at which you can build a natural gas powerplant isn’t the only factor working in their favor.
When oil is pumped out of the ground, gas comes up with it. If you don’t transport, store, and use the gas that is produced alongside the oil then it’s wasted. This is exactly what’s happening in the Permian Basin today where gas is produced but not all of it is brought to market. There is nothing environmentally friendly about wasting gas that was already pumped out of the ground.
If the datacenter industry needs power now, and if abundant cheap base load power is available in the form of produced but unused natural gas, then I believe natural gas will be the solution. At least in the near term, say one to five years, it’s reasonable to believe that physics will trump politics. This is exactly what happened in Germany between 2021 and 2023 when after losing access to Russian oil, Germany turned back to coal to meet their base load power requirements.
This was a long write up just to conclude that datacenter growth forecasts will probably play out as predicted, and that in the near term natural gas will be the power of choice for the industry. That doesn’t mean that things will play out that way, just that it is a likely outcome. As we covered earlier, intelligence analysts and investors have similar tasks. Both of them must sift through incomplete information and formulate actionable plans for unknown future outcomes. The ACH model was developed for the intelligence community but it serves investors just as well, allowing them to clarify the relationship between facts and hypothesis that drive decisions in the real world.
Because LandBridge and TPL are low to no debt royalty companies, they will likely be successful even if datacenters never add a dime to their bottom line. That is the beauty of their business model. By owning land, mineral, and water rights, they retain a perpetual call option on all forms of technology, both current and yet to be developed. Profitably investing in either company isn’t dependent on the data center thesis, it’s simply a free call option on one more technological development.
Share this post