4th Quarter 2025 Commentary

Theme and Background

This chapter addresses market skepticism regarding Texas Pacific Land Corp. (TPL) lagging behind in the AI data center wave. Clients repeatedly asked in October and December 2024: why doesn't TPL build its own data centers? Has the data center value of its land assets been disproven? By introducing the substantive partnership between TPL and Bolt Data & Energy, the report not only directly rebuts these doubts but also reveals the true driving force behind AI data center demand—NVIDIA's chip upgrade cycle.

Core Thesis

The author's central judgment is: TPL is not being marginalized; on the contrary, it has directly participated in private market investments in AI infrastructure as a cornerstone investor. More counterintuitively, the market's understanding of AI data center demand may be overly focused on the projects themselves, while the real driver is NVIDIA's chip upgrade cycle. This cycle is also a key factor in assessing the risk of current IT sector and S&P 500 index performance—meaning that judging AI computing power demand is essentially judging the pace of NVIDIA's product iteration.

Key Arguments and Data

1. TPL's Substantive Deployment: On December 17, 2024, TPL announced an agreement with Bolt Data & Energy to develop a large-scale data center campus on its land.

• TPL provided one-third of the $150 million financing (i.e., $50 million).
• TPL will be responsible for supplying water for potential projects.

2. Partner's Strength: Bolt's chairman is former Google CEO and chairman Eric Schmidt.

• Schmidt mentioned potential tenants include: Google, Microsoft, Meta, Amazon, Oracle, OpenAI, Anthropic, xAI, Palantir, and even the White House's "Genesis Mission for AI."

3. Project Scale Planning:

• Initial plan: A campus with 1 GW capacity, powered by a gas-fired power plant.
• Ultimate target: 10 GW.

4. Comparative Benchmark: The Vast Gap Between Fermi Inc.'s Stock Price and Reality:

• Fermi was co-founded by former Texas Governor and U.S. Energy Secretary Rick Perry, planning to build the world's largest private AI data center power grid with a total capacity of 11 GW.
• Its early market capitalization after IPO (October 2025) was as high as $19 billion; it has now fallen to $6 billion, a decline of approximately 68%.
• As of December 2024, Fermi's actual progress: only 300 acres of land cleared (out of 5,200 total acres), several miles of gas pipelines and water pipes installed, and 11 miles of fencing. The report sarcastically notes that the fence's purpose is "to prevent others from entering the aforementioned land."

Comparison Data Table:

Company/Project	Planned Capacity	Early Valuation	Current/Latest Valuation	Actual Construction Progress
Bolt Data & Energy (TPL Partnership)	1 GW (planned), 10 GW (long-term)	Private	Not listed	Agreement just reached
Fermi Inc. (Listed Comparison)	11 GW	$19 billion	$6 billion	Cleared 300 acres, installed pipelines and 11 miles of fencing

Companies/Assets Involved

• Texas Pacific Land Corp. (TPL): Bullish. As a landowner, it is no longer just a "land supplier." By providing capital (one-third of financing) and infrastructure (water supply), it has become a cornerstone investor and key participant in the AI data center project. This move clearly responds to market doubts about its lack of data center presence.
• Bolt Data & Energy: Unlisted company, TPL's partner. Led by former Google CEO Eric Schmidt, which serves as the main credit endorsement of its technical strength and resource acquisition capability.
• Fermi Inc.: Bearish (as a cautionary example). Its stock price collapsed from $19 billion to $6 billion, vividly demonstrating the fragility of valuations supported solely by blueprints and grand visions when actual operational progress is lacking. The report uses this to illustrate that market enthusiasm for the "data center theme" may run ahead of fundamentals.
• NVIDIA: The author believes its chip upgrade cycle is the underlying logic for understanding the entire AI data center demand. The report does not directly make a bullish or bearish judgment on NVIDIA, but it clearly points out that its product cycle is the core variable for assessing risk in the IT sector and S&P 500 performance.

Investment Implications

1. Focus on Basic Facts, Not Market Sentiment: Before the Bolt deal was announced, market concerns that TPL "had no data center" were genuine. However, the deal reveals the company management's active behind-the-scenes positioning. Investors should be wary of the risk of making buy/sell decisions based solely on public market rumors or lagging information.

2. Beware of Valuation Traps in "Concept Stocks": The Fermi case is a classic warning. A project with only land clearing and fencing as progress once had a market cap of $19 billion. The actual commercialization process of AI data center construction may be far slower than market expectations, and hype around related themes may experience sharp corrections. Investors should remain cautious toward companies that lack actual operating revenue and derive high valuations merely from plans.

3. Re-examine the Driving Force of Investment Logic: The report implies that judging the investment value of the AI data center theme should not focus solely on how many GWs several companies have announced. Instead, it should more prospectively look at the upstream chip (NVIDIA) upgrade cycle. If chip iteration slows, downstream computing power demand and construction pace may also adjust accordingly. This provides a deeper perspective for analyzing leading IT sector companies.

---

Theme and Background

This chapter explores a key yet overlooked contradiction in the frenzy of AI data center construction: "planned obsolescence" driven by rapid technological iteration. The report argues that while the market generally believes IT giants' build-out of AI infrastructure is inevitable, many fail to recognize that data centers themselves are rapidly becoming obsolete due to chip upgrades, which will profoundly affect the capital returns and competitive landscape of related companies. Texas has replaced Northern Virginia as the new epicenter of U.S. data center construction.

Core Thesis

The author's core investment thesis is: In the AI data center race, first-mover advantage may turn into "first-mover disadvantage," and the true winners will be asset owners with strategic resources such as water, natural gas, and land—not chip buyers or sellers.

Contrarian judgments:

1. Many existing data centers, including relatively new ones, are already technologically obsolete.

2. NVIDIA's chip upgrade strategy is essentially "designed obsolescence." The capex cycles of chip buyers (e.g., Microsoft, Google) may never end, causing their high cash flow expectations to fall short.

3. The stock price logic of both chip sellers and chip buyers contains a fundamental contradiction—they cannot both be correct simultaneously, posing a major risk to the S&P 500.

Key Arguments and Data

The report supports the "planned obsolescence" argument with exponential growth in chip power consumption and related data:

Chip Model	Release/Planned Time	Power per Rack	Power vs. H100	Performance Gain
NVIDIA H100	October 2022	14 kW (700W per chip)	Baseline	Baseline
NVIDIA Grace Blackwell (GB200)	March 2024	120 kW	8.6x	30x (LLM inference)
NVIDIA Vera Rubin	H2 2026 (est.)	600 kW (est.)	43x	5–10x (inference) & 45% efficiency gain

The report further states:

• Endless capex: IBM CEO Arvind Krishna estimates that building 100 GW of data centers would require hyperscalers to invest $8 trillion, with interest costs alone reaching $800 billion, and "old chips must be scrapped and racks refilled within five years."
• Structural contradiction in the S&P 500: Chip sellers (NVIDIA, Broadcom, Micron, AMD) account for 11.5% of S&P 500 weight, while chip buyers (Microsoft, Alphabet, Amazon, Meta, Oracle) account for 18.3%. The author believes the core business logic of these two groups is mutually contradictory—they cannot both achieve expected profits and cash flows.
• Scarcity of land and resources: The report emphasizes that the Delaware Basin in Texas uniquely offers "ample remote land, natural gas, and water," and the scarcity of these resources is critical. As a case study, Fortune mentions two companies: one data center project saw its market cap collapse from $19 billion to $6 billion, with actual progress limited to clearing 300 acres of land, installing several miles of gas pipelines, and 11 miles of fencing.

Companies/Assets Involved

• Texas Pacific Land Corp (TPL): As one of the core asset owners in the Delaware Basin, TPL is seen as a potential beneficiary. Its land hosts Bolt Data & Energy's planned 10 GW data center project (not detailed in this report, but noteworthy).
• LandBridge Co.: Owns land in the Delaware Basin. Recently announced a partnership with NRG Energy to develop 1.1 GW of natural gas-fired power generation, and a battery storage project with Samsung C&T.
• Apple Inc.: Cited as a negative example. The report notes Apple chose to pay Google $1 billion per year to lease AI models and cloud capabilities, adopting a "tenant" model rather than building its own—reflecting a fundamentally different assessment of capital returns.
• NVIDIA & Google (Alphabet): Defined as representatives of "chip sellers" and "chip buyers," respectively. The report argues that NVIDIA's chip upgrade strategy is, by design, detrimental to buyers like Google.
• WaterBridge: Mentioned as another resource owner in the Delaware Basin.

Investment Insights

1. Beware the "two-sided logic" trap in tech: Investors should recognize that the chip-buyer and chip-seller groups, which together account for nearly 30% of the S&P 500, cannot both succeed. If chip buyers' capex cannot converge, their high-valuation logic faces challenges; if chip buyers cut capex, chip sellers' revenue growth stalls. Either scenario implies fundamental risk in current tech-sector valuations.

2. Focus on "late-cycle" beneficiaries, not "first movers": The AI data center build-out has only just begun. "First movers" that mimic NVIDIA's chip upgrade cadence (e.g., early H100 data center operators) risk rapid asset depreciation. Save time, wait for technology standards to stabilize, or favor operators who control "irreplaceable" natural resources (water, land, electricity)—a better strategy.

3. Go long on "digital economy landowners": The report's core insight is that companies owning land, natural gas, and water in the Delaware Basin (e.g., TPL, LandBridge, WaterBridge) are becoming the "sellers of shovels" in the AI model race. No matter which tech giant wins, they all need these physical resources to build and run data centers. The value of these assets has not yet been fully priced by Wall Street and may be multiples of their current market caps.

---

Theme and Background

This chapter delves into the water usage issues of data centers in the Permian Basin, specifically focusing on the types, availability, and treatment costs of water sources used for power generation cooling and direct data center cooling. The report meticulously distinguishes the chemical differences among fracturing source water, produced water, and high-purity cooling water. It assesses the feasibility and economics of large-scale water treatment, revealing the complexity of water supply as a key bottleneck for AI data center expansion.

Core Thesis

The author's central argument is that the water demands of large-scale AI data centers (>1GW), especially for high-purity cooling water, cannot be economically met onshore at scale using existing sources (e.g., saline source water) or conventional desalination technologies (e.g., reverse osmosis). This will make water a hard constraint on data center construction, creating unique investment value for companies that control land, water rights, and differentiated water treatment technologies (such as Texas Pacific Land).

Counter-consensus / contrarian views:

• The market generally assumes "water is water," but in reality, the cheap, saline source water (brine) used for fracturing and the high-purity water required for data center cooling are two entirely different resources, with vastly different treatment costs and technical hurdles.
• Most people misunderstand "water abundance." The report notes that the Chihuahuan Desert, where the Permian Basin is located, receives less than 10 inches of annual rainfall, aquifer recharge is extremely limited, and water sources are not infinitely renewable.
• Although produced water is widely "recycled" for re-fracturing, its salinity is up to 10 times that of seawater and it contains heavy metals and radioactive materials. Large-scale treatment into usable water is currently considered infeasible.

Key Arguments and Data

The report supports its "water hard constraint" thesis with detailed hydrogeological data and engineering logic.

• Supply-Demand Imbalance for Water: A high-producing oil well in the Delaware Basin may produce approximately 1.562 million barrels of oil equivalent (BOE) over its lifetime, but the associated produced water could reach about 6.25 million barrels (based on a conservative 4:1 water-to-oil ratio). In contrast, fracturing the same well requires only about 875,000 barrels of water. The volume of produced water is more than 7 times the fracturing water demand.
• Massive Cost Differences by Water Quality:
• Source Water: Prices in parts of the Delaware Basin exceed $1/barrel. Under a long-term, large-volume supply contract, a conservative estimate is $0.5/barrel. For a 1GW power plant, this translates to annual revenue for the water supplier of over $20 million, and this is for untreated source water.
• Treated High-Purity Water: To meet standards for data center cooling or agricultural use, treatment costs (including desalination and waste disposal) could approach $2-3/barrel. Among these, waste (concentrated brine) disposal (e.g., via SWD injection or landfilling) is the primary cost and scalability constraint.
• Treatment Technology Bottlenecks: The most common reverse osmosis technology, when applied to produced water, causes rapid membrane saturation and frequent replacement due to extremely high total dissolved solids (TDS), making the process "impractical." In contrast, the freezing desalination technology developed by Texas Pacific Water Resources (TPWR) may have potential for cost reduction, but without initial subsidies, costs remain a limiting factor.
• Water Consumption Estimates:
• A 1GW combined-cycle gas turbine power plant, running at a 100% capacity factor (though uncommon in reality), requires 5 million gallons (approximately 120,000 barrels) of water per day for cooling. About 70-80% of this is lost to evaporation.
• Due to the massive direct cooling needs of modern hyperscale data centers, their water consumption will "meet or exceed" that of power generation.

Data Comparison Table:

Water Source Type	Source / Characteristics	Cost (USD/barrel)	Treatment Difficulty / Use
Source Water	Saline formation brine from aquifers	0.50 – 1.00+	Low, directly usable for fracturing
Produced Water	Salinity up to 10x seawater, contains heavy metals, radioactive substances	Treatment cost 2-3	Extremely high, currently mainly disposed of via injection, almost unusable for data centers
Treated High-Purity Water	Desalinated (freezing or reverse osmosis)	2-3+ (excluding waste disposal)	Extremely high, theoretically usable for data centers or agriculture, but scaling is expensive

Companies/Assets Involved

• Texas Pacific Land Corp. (TPL): As a broad owner of land and water rights in the Permian Basin, TPL is a direct beneficiary of source water sales. Its subsidiary Texas Pacific Water Resources (TPWR) is developing freezing desalination technology, aiming to use treated produced water for agricultural irrigation, livestock watering, and aquifer recharge. The author believes that given the massive capital expenditure of AI data center customers, they are relatively insensitive to water treatment costs, creating a potentially large private market demand for TPWR's technology commercialization, even if initial costs rely on subsidies.
• Fermi Inc. (not directly mentioned in this chapter but part of the report's broader context): As a contrasting case, it illustrates that without reliable water support, an ambitious data center construction plan (11GW) could face enormous execution risk, with its market capitalization collapsing from $19 billion to $6 billion.

Investment Implications

• Bullish Direction: Bullish on Texas Pacific Land Corp. (TPL). The report clearly indicates that the surging water demand from AI data centers will directly translate into value for TPL's land and underlying aquifers. TPL can not only generate stable revenue as a source water supplier but also, if its subsidiary TPWR's differentiated desalination technology achieves scale, it could fundamentally solve the "water shortage" problem for data centers, becoming the "water seller" for the industry with substantial long-term value.
• Risk Warnings: Be cautious about projects dependent on "water purchase agreements" for building large-scale data centers (e.g., Fermi Inc.). The author suggests that the technical feasibility of water sources and the economics of long-term supply contracts are the most critical prerequisites for data center commissioning. Projects lacking stable, low-cost water supply may have fundamental flaws in their investment thesis.

---

Theme and Background

This chapter responds to a late 2024 Wall Street Journal article on water stress and pore space depletion in the Permian Basin. The article raised market concerns that the Delaware Basin might be forced to halt production due to wastewater injection pressure issues. The report’s author argues that this concern is overblown and, from the perspective of land and water infrastructure owners, is actually a positive.

Core Thesis

The report’s central judgment is: Water stress will not halt production in the Permian Basin; instead, it will boost demand for long-distance water treatment and land assets, benefiting land and water rights owners. The author presents a counterintuitive argument: even in the worst-case scenario (a shutdown of the Delaware Basin), global oil supply would drop by roughly 4%, sending oil prices soaring to $150–$250 per barrel, and energy companies’ valuations would increase significantly. Furthermore, regulators have already taken substantive measures, and companies are resolving bottlenecks through long-distance pipelines and out-of-basin disposal.

Key Arguments and Data

• Abundant Land Resources: TPL’s surface land area is 33% larger than Rhode Island, providing ample space for water storage—only requiring longer-distance transport infrastructure.
• Supply Shortage Risk: A Delaware Basin shutdown would remove approximately 3.5 million barrels per day of oil supply (equivalent to UAE production), or roughly 4% of global supply. For inelastic oil demand, a 1% supply-demand imbalance has historically caused price swings of ±50% to +100% (citing Q3 2020 commentary).
• Regulatory Responses Already in Place: The Railroad Commission of Texas has established Seismic Response Areas that strictly limit injection volumes; it has banned permits for deep injection (below oil and gas zones); and it imposes pressure and flow rate caps on shallow wells.
• Infrastructure Under Construction: Phase I of the WaterBridge Speedway pipeline project is underway, comprising 70 miles of 30-inch diameter pipe with a capacity exceeding 1 million barrels per day, expected to begin operations by mid-2025. An additional 3 million barrels per day of potential capacity is in development. Out-of-basin disposal is not “years away”; it is happening now.
• Cost Pass-Through: A 70-mile transport route will significantly raise producers’ water handling costs, but for companies that own land and long-distance pipeline systems (such as TPL, LandBridge, and WaterBridge), this represents incremental revenue.

Item	Data/Status
TPL land area vs. Rhode Island	TPL is 33% larger than Rhode Island
Potential impact of Delaware Basin shutdown	Loss of ~3.5 million bbl/day (4% of global supply)
WaterBridge Speedway pipeline Phase I	70 miles, 30-inch diameter, >1 million bbl/day, mid-2025 start
Historical price elasticity reference	1% inventory swing historically caused ±50% to +100%

Companies/Assets Covered

• Texas Pacific Land Corp. (TPL): Core beneficiary. With a large land portfolio, the expansion of long-distance water transport infrastructure will directly increase demand for its land and associated usage fees. Bullish.
• LandBridge: Similar to TPL; as a Permian Basin land asset owner, it benefits from land lease and pipeline access revenue driven by water disposal demand. Bullish.
• WaterBridge: An infrastructure company providing water treatment services. Its new Speedway pipeline project, expected to be operational by mid-2025, is a direct beneficiary of bottleneck resolution. Bullish.

Investment Implications

1. Go long on Permian Basin land and water rights assets: Energy production and regulatory pressures will not subside; they will only increase reliance on “pipeline distance” and “permitted capacity.” Companies with already-permitted disposal well capacity (e.g., TPL, LandBridge) have upside valuation potential.

2. Go long on long-distance water treatment infrastructure: The longer the pipeline, the higher the fee and the tighter the capacity. Companies like WaterBridge, with pipelines under construction or already operating, will benefit from rising water treatment fees.

3. Take a contrarian view on energy production risk: The market’s overreaction to “pressure cooker” news may offer patient investors an opportunity to acquire land and infrastructure assets at a low cost.

---

Theme and Background

This chapter focuses on the scale of AI data centers' water demand and its pricing logic. While current market attention centers on data center electricity supply, there is little awareness of the supporting cooling water requirements. The report points out that both power plants and data centers are major heat loads, and their combined water consumption far exceeds common perception, a demand that is reshaping the value chain of the water resource industry in the Texas Basin.

Core Thesis

The report's central judgment is: Data center water usage is not a distant future concept but a realized source of revenue growth that is already occurring. Texas Pacific Land Corp. (TPL)'s water supply revenue already accounted for over 35% of total revenue in 2024, rising from nearly zero eight years ago. The author argues that the market significantly underestimates the value of water assets — current water prices are only about $1 per barrel, yet a 1 GW data center campus could incur annual water costs of over $125 million, and if expanded to 10 GW, that translates to billions of dollars in recurring, high-margin annual revenue. The counterintuitive aspect is that while the market debates "AI power bottlenecks," the scarcity and pricing power of water may be systematically undervalued.

Key Arguments and Data

1. Power Plant Water Efficiency Differences:

• U.S. power plants account for 40% of total national water withdrawals, as natural gas, coal, and nuclear are all thermal generation (boiling water → steam → turbine).
• Natural gas plants use only 15% of the water per MWh consumed by coal plants (EIA data); excluding evaporation and maintenance losses, natural gas plants consume about 35% of nuclear and 44% of coal plants' water per unit of generation (USGS).
• A 1 GW natural gas combined-cycle plant uses approximately 120,000 barrels of water per day (44 million barrels annually).

2. Data Centers Themselves Use Even More Water:

• A 1 GW data center has an annual cooling load of about 8.76 TWh (equivalent to 800,000 U.S. homes or 95 skyscrapers).
• Assuming a 50% wet cooling + 50% adiabatic cooling split, annual water consumption is approximately 67 million barrels [Lawrence Berkeley Lab: wet cooling WUE 1.8 L/kWh; Microsoft 2024 Sustainability Report: adiabatic cooling WUE 0.30 L/kWh].
• This is about 70% higher than the water usage of the supporting power plant.

3. Combined Water Cost Calculation:

• Power plant water: At $0.50/barrel, annual water cost for 1 GW is about $22 million (adjusted to $25 million after considering a PUE of 1.15).
• Data center water: At $1.50/barrel (including treatment and disposal), annual water cost is about $100 million.
• The total annual water cost for a 1 GW data center is at least $125 million; if Bolt's 10 GW plan materializes, water costs alone would exceed $1.25 billion per year.

4. Industry Comparison:

• WaterBridge has a market cap of $2.8 billion, Landbridge $4.7 billion, and TPL $24 billion (as of January 23, 2026).
• TPL's water supply business has grown at a 45%-50% compound annual growth rate over the past eight years, with an asset-light operating model (source water infrastructure generates no direct operating expenses, and saline water resource fees are pure royalties).

Water Usage Component	Type	Water Consumption/Usage	Price Assumption	Annual Water Cost (1 GW)
Power Plant	Thermal cooling	120,000 bbl/day (44M bbl/year)	$0.50/bbl	$22M–$25M (after PUE adjustment)
Data Center	Cooling	67 million bbl/year	$1.50/bbl	$100M
Total		~111 million bbl/year		$125M–$130M

Companies/Assets Involved

• Texas Pacific Land Corp. (TPL): The report's core target. Owns the largest source water infrastructure network in the Northern Delaware Basin, covering aquifers. Water supply revenue already exceeded 35% of total revenue in 2024. Asset-light operating model, akin to a "financial casino" (high margin, low capex). The author is implicitly bullish.
• WaterBridge ($2.8B market cap) and Landbridge ($4.7B market cap): Comparison targets, primarily focused on produced water treatment, but source water is viewed as higher-margin yet more cyclical assets. The author implies their valuations may not fully reflect source water value.
• Bolt Energy & Data: Plans to build a 1 GW data center on TPL land, expandable to 10 GW, chaired by Eric Schmidt. Its water demand serves as the base case for the above calculations.
• Chevron Data Center: Plans to build a 5 GW campus.

Investment Implications

• Direct Focus on Water Supply Assets: Landowners with untapped water sources (such as TPL) will benefit from the dual water demand of data centers and supporting power plants, generating a sustainable, high-margin revenue stream. Current market pricing for "water fees" is only a fraction of actual value.
• Beware of Lagging Consensus: When financial media begins to hype "water ETFs," the theme may already be overheated. However, we are still early — water source assets have not been broadly revalued, yet demand is already a reality (TPL water supply revenue CAGR 45%+).
• Infrastructure-Type Valuation Logic: Data center water supply contracts are long-term, recurring, and high-margin (similar to pipeline tolls), not cyclical "water sales." Such revenue should command valuation multiples above traditional utilities.

---

Theme and Background

This chapter explores the investment value of the "Croupier" business model. By comparing two typical croupier businesses—exchanges and asset managers—the author argues why the former is a superior long-term compounding tool and points out a systematic mispricing of this business model by the market. The report traces Horizon Kinetics’ research on exchanges dating back to the 1990s and connects it to emerging croupier investment themes such as today’s AI data centers.

Core Thesis

The core investment thesis is: Exchanges are more ideal croupier investment targets than asset managers. The author believes exchanges possess a near-monopolistic business model, a fixed cost structure, no need to assume client capital risk, and performance that does not depend on market direction (they can charge fees whether markets rise or fall). This judgment was formed 20 years ago (in 2005) and remains unchanged. A contrarian observation: as early as 2005, the author proposed that holding exchange stocks could be a superior way to participate in the market compared to traditional index funds—because buying a few monopolistic exchanges might outperform an index composed of thousands of stocks.

Key Arguments and Data

1. Historical Comparison Verification: Horizon Kinetics’ "Money Managers Index" (initially 11 asset management firms), compiled since 1990, achieved an annualized return of approximately 15% through 2025, demonstrating that asset managers as "croupiers" did outperform the broad market. However, the author argues this is far from optimal.

2. Business Structure Comparison: Exchanges vs. Asset Managers:

Dimension	Exchange (Ideal Croupier)	Asset Manager (Suboptimal Croupier)
Capital Risk	Extremely low, almost no capital deployed	Relatively low, but may co-invest (e.g., AMG)
Leverage Needs	No debt leverage	Typically no leverage, but revenue is volatile
Cost Structure	Fixed cost dominant, variable costs extremely low	Variable costs (compensation) high, ~50% of revenue
Revenue vs. Market Direction	Charges fees in both up and down markets	Revenue shrinks in down markets, redemptions increase
Permanence of Client Capital	Not applicable (transactions non-custodial)	Non-permanent, high redemption risk
Capital Accumulation Ability	Strong, can retain and reinvest earnings	Weak, earnings drained by management bonuses/dividends

3. Data Case Studies:

• Affiliated Managers Group (AMG): Manages $800 billion in assets, but compensation expenses account for ~50% of revenue; net equity is $4.3 billion, and after deducting $4.2 billion of intangible assets (goodwill), tangible shareholder equity is nearly zero. This illustrates that asset managers cannot effectively accumulate capital.
• Alliance Bernstein LP: Annualized stock price appreciation over the past 10 years was only 4%, but adding a ~9% dividend yield brought total return to 14%. This indicates the model relies mainly on dividends, not capital appreciation.
• NYSE Seats (1986-2004): Superficial annualized price change was less than 4%, but reinvesting seat rental income yielded an annualized return exceeding 17%. This reveals the market systematically underestimates the true profitability of exchanges.

Companies/Assets Discussed

• Affiliated Managers Group (AMG): An asset management case used to illustrate compensation structure issues (50% of revenue for compensation). The report notes it is a holding in some strategies, pays no dividends, and its returns come entirely from stock price appreciation.
• Alliance Bernstein LP: An asset management case, structured as a limited partnership with a ~9% dividend yield, used to demonstrate a "capital recovery first" model—limited stock price appreciation but substantial dividends.
• Urbana Corp.: Used as an example of a "quasi-private" exchange investment vehicle. Market capitalization is only $286 million, average daily trading volume is about 19,000 shares, which institutions cannot build positions in quickly. The report argues this company resembles an early-stage Texas Pacific Land Corp., presenting a cognitive mismatch opportunity.

Investment Implications

For investors, the implication is: Prioritize finding true "croupier" models over suboptimal substitutes. Specific directions:

1. Focus on Exchanges: Although both exchanges and asset managers are viewed as "toll bridges," exchanges’ business structure (fixed costs, no leverage, permanent clients, fees in up and down markets) gives them higher intrinsic returns. Currently, the exchange sector makes up less than 0.5% of the S&P 500, making it a systematically undervalued category.

2. Beware of Hidden Costs in Asset Managers: High compensation ratios and profit-sharing mechanisms (insider bonuses) significantly erode shareholder returns, and these companies struggle to accumulate their own capital outside of bull markets.

3. Look at Illiquid "Quasi-Private" Opportunities: Companies like Urbana Corp., with small market caps and low trading volumes, may be overlooked due to institutional liquidity thresholds, yet these are often where the greatest cognitive mismatches lie.

---

Here is the analysis of the sequel to "Then and Now: Searching for Croupiers," continuing the style of earlier content with new arguments, data, and perspectives.

From "Casino" to "Digital Infrastructure": The Evolutionary Logic of Urbana and Exchanges

The sequel uses Urbana as an entry point to deepen the "casino croupier" metaphor. Urbana is not an ordinary investment company; its unique corporate structure (reclassified from an investment fund to a corporation) allows it to exert control over its assets (especially exchanges) and hold them for the long term, much like a "casino owner." This structural "privilege" is key to capturing the long-term value of exchanges.

Urbana’s "Croupier" Characteristics:

• Structural Advantage: By using a corporate rather than a fund structure, Urbana avoids investment fund position limits and can take controlling stakes in private exchanges like MIAX. This "semi-public/semi-private" model allows investment in assets inaccessible to ordinary investors (e.g., CNSX Global Markets and Blue Ocean Technologies).
• Value Gap: It has historically traded at a 40%-50% discount to NAV, which has narrowed recently but remains significant. This resembles a closed-end fund discount arbitrage opportunity, but Urbana’s asset quality (especially exchanges) provides a margin of safety.
• Unique Assets: Holding 44 gold interests in Quebec’s Urban Township embeds a real-asset hedge within the portfolio, increasing its appeal as a "store of value."

The "Free Call Option" Model in Exchanges:

The sequel explicitly introduces a core idea: exchange businesses inherently embed "free call options." This does not refer to financial derivatives but to the fact that exchanges, as trading infrastructure, passively grow in value as new asset classes emerge.

Historical Example	Option Nature	Path to Realization
VIX Index	Free call option	After CBOE developed the VIX business line, volatility itself became a tradable product, with annual volume reaching 200 million contracts.
Digital Assets and Blockchain	A much larger free call option	Blockchain technology not only gave rise to cryptocurrency spot trading but also opened up the broader arena of tokenization.

Tokenization: The "Second Growth Curve" for Exchanges

The sequel’s core argument is that tokenization is not simply "cryptocurrency trading" but a fundamental reconstruction of the post-trade processing of traditional securities. This presents exchanges with a transformative opportunity far exceeding that of the VIX.

1. Breaking "Post-Trade" Bottlenecks, Enabling 24/7 Global Trading

• Traditional Pain Point: Exchange matching engines can run 24 hours, but post-trade processes—clearing, settlement, fund transfers—rely on intermediaries such as banks and custodian banks, which are inefficient during non-business hours (weekends, holidays) and in cross-border scenarios.
• Tokenization Solution: Tokenizing assets like stocks and bonds unifies cash and securities on a single blockchain system, enabling 7x24 instant settlement. This bypasses the "siloed" systems of traditional intermediaries, allowing matching engines to operate without interruption.
• Historical Pattern: Any technology that lowers trading barriers and increases market liquidity ultimately leads to a significant rise in trading volumes. Tokenization is the modern embodiment of this pattern.

2. Enhancing Capital Efficiency and Risk Management

• Instant Settlement: The immediacy of blockchain drastically reduces counterparty risk. The default risk inherent in traditional T+2 settlement cycles is compressed to nearly zero.
• Capital Redeployment: Funds and securities become available immediately after a trade is completed, greatly increasing capital turnover. For high-frequency traders and market makers, this means higher capital utilization and lower funding costs.

Regulatory "Green Light" and Industry Response

The sequel highlights the joint statement by the CFTC and SEC at the end of 2025 as a turning point, signaling that regulators have paved the way for trading tokenized assets on regulated exchanges. This is not an isolated event but a collective action by industry giants:

• NYSE: Within months of the statement, announced a new platform for trading and settling tokenized securities.
• ICE: The parent company of NYSE, Intercontinental Exchange, has begun collaborating with BNY Mellon and Citigroup to plan 24-hour global trading. This shows that even traditional financial giants recognize that embracing tokenization is necessary to maintain their "croupier" status.

Conclusion: From "Casino" to "Digital Highway Operator"

Urbana’s investment logic and the evolutionary path of exchanges together paint a clear picture: The future exchange is no longer a simple "casino" but a "digital highway operator."

• Urbana: As an early investor in "highway operators," its structural advantage allows it to hold and control these key infrastructure assets long-term. Its NAV discount represents a market mispricing of the "new asset class option."
• Exchanges: Through tokenization, they expand their business from "matching trades" to "full lifecycle trade management," including issuance, recording, clearing, and settlement. This shifts their value from trading commissions to the rent from the entire financial infrastructure.
• Risks and Challenges: Tokenization poses a direct threat to traditional clearing houses (e.g., DTCC) and custodian banks. Whether exchanges can successfully "disintermediate" will depend on their negotiations with regulators and existing financial institutions. But as the sequel notes, this business model has adapted to centuries of technological change and is most likely to "stand above the market" through transformation.

Ultimately, investing in Urbana or similar exchanges is essentially betting on the digital upgrade of financial infrastructure. This upgrade process will generate enormous value, and the "croupiers" are riding the crest of this transformation.

---

Theme and Background

This chapter argues that the "royalty" business model enjoys a significant valuation advantage over traditional capital-intensive enterprises. The report points out that market consensus generally undervalues royalty companies and does not even recognize them as an independent investment category, possibly due to their lack of capital expenditure, strong earnings compounding effects, and neglect of the value of non-producing royalties.

Core Thesis

The report's core judgment is: one dollar of net profit from a royalty company is worth far more than one dollar of profit from a traditional enterprise. This stems from the fact that royalty companies do not need to reinvest profits into capital expenditures, so their P/E ratios are systematically overstated. The market commonly categorizes royalty companies as "high P/E" and rejects them, which is precisely a cognitive bias. Additionally, a large number of assets in royalty portfolios that have not yet generated revenue are essentially zero-cost call options, whose value is severely underestimated.

Key Arguments and Data

1. Profitability Comparison: Taking Google as an example, it needs to spend over 50% of its net profit on capital expenditures such as plant and equipment. Therefore, $100 of net profit from a royalty company is worth at least $200 of net profit from Google in valuation, and possibly even more.

2. Franco-Nevada's Asset Structure: The company's 119 producing assets contribute all revenue, but this only accounts for about 28% of its total royalty portfolio. The remaining 311 assets (about 82%) contribute no revenue, including 38 mines at the "advanced development" stage. The market treats these "dormant" assets as having zero value, but the report argues they are essentially zero-cost long options. Their strike prices (initial investment costs) were based on low prices from years ago (gold around $1,000/oz, silver around $15/oz), while current prices (gold around $4,800/oz, silver around $94/oz) have caused these options to soar in value.

3. Altius Minerals Transaction Case: In July 2025, Altius Minerals sold a 1% royalty interest in the Silicon and Merlin gold mines in Nevada to Franco-Nevada for $275 million. The implied value of this transaction is enormous, as shown in the following data:

Data Item	Value/Description
Royalty percentage sold	1%
Total transaction price	$275 million
Implied value of 100% royalty	$275 million × 100 = $27.5 billion
Comparison benchmark	The implied value of 100% royalty ($27.5 billion) exceeded the market cap of mine operator AngloGold Ashanti at the time ($26 billion)
Net asset size for Altius in this transaction	The transaction value was about 30% of Altius's market cap at the time ($900 million)
Initial investment cost (2015)	Approximately $300,000
Investment return multiple	Approximately 1,375x (annualized return of about 106%)

4. Altius Stock Performance: In the year following the transaction, Altius's stock price rose 75%. Copper prices rose 33% (accounting for 40% of its revenue in the first nine months of 2025), and potash prices rose 22% (accounting for 30% of its revenue).

Companies/Assets Involved

• Wheaton Precious Metals (WPM): No bull/bear view mentioned. As a gold/silver royalty company, its stock has risen more than 10x in 10 years, with a market cap exceeding $60 billion. However, due to its Canadian domicile and the "passive" nature of royalty business, it has not been included in the S&P 500 index and is overlooked by the market.
• Franco-Nevada Corp. (FNV): Bullish. The report's core beneficiary. The market only values it based on its 28% of producing assets, ignoring the huge call options constituted by 82% of non-producing assets.
• Altius Minerals (ALS) / Silicon & Merlin Gold Discoveries: Case study target. Demonstrates how holding a 1% royalty can yield a 1,375x return over 10 years. The transaction itself is a strong market pricing validation of the royalty business model's value.
• AngloGold Ashanti (AU): Comparison target. The overall market cap of its global mining operations ($26 billion) is surpassed by the implied value of a single mine's 1% royalty ($27.5 billion), highlighting the valuation difference between traditional and royalty models.

Investment Implications

• Reject using traditional P/E frameworks to evaluate royalty companies. Investors should proactively correct their cognition and grant a higher valuation premium to royalty companies with the same profit level, because the "quality" of their profits (no need for reinvestment) is far superior to that of traditional enterprises.
• Focus on "dormant" assets in royalty portfolios. These non-revenue-generating, seemingly zero-value assets are actually low-risk, high-return "convex" options (moonshots). When mining development succeeds or commodity prices surge, returns are astonishing. They should not be ignored.
• View royalty companies as structural tools to hedge against inflation and currency depreciation. The "capital-light, hard assets" strategy mentioned in the report is a core driver of long-term compounding. Investors may consider allocating to such assets as core holdings, leveraging their low capital expenditure and high inflation-benefit characteristics for long-term returns.