3rd Quarter 2025 Commentary

Theme and Background

This chapter examines the stark divergence between actual investment returns over the 25 years since the dawn of the ETF era in 2000 and mainstream market expectations (such as the Ibbotson and Sinquefield studies projecting annualized equity returns above 10%). The report argues that despite passive investing surpassing active management in scale and the market being heavily concentrated in the information technology sector, actual ETF returns have fallen far short of historical benchmarks, challenging conventional asset allocation wisdom.

Core Thesis

The author's central investment argument is: Over the past 25 years, the actual annualized returns of ETFs have been far lower than historical expectations, and current market concentration (especially the IT sector comprising 46.1% of S&P 500 market cap) and extreme valuations signal a risk of capital loss, not excess return opportunities. This is a contrarian judgment: most investors fear missing out on the excess returns of IT (AI/data centers), but the author believes the risk of capital loss from participating is more worthy of attention. Simultaneously, the report suggests that by abandoning index-based passive investing and pivoting to specific "anti-index" assets (such as stock exchanges), investors may achieve higher long-term returns.

Key Arguments and Data

1. ETF Returns Below Expectations: Over the past 25 years, annualized returns for nearly all equity ETFs have fallen into the 7%-8% range, well below the 10%+ projected by Ibbotson and Sinquefield studies. Fixed-income ETFs delivered annualized returns below 3.5%, which turned negative after taxes and inflation.

2. Market Concentration and Valuation Distortions: As of the report's analysis date, the information technology sector, including Amazon, Meta, and Alphabet, accounted for 46.1% of the total S&P 500 market capitalization. The author compares this to the peak of the tech bubble in 1999/2000 to demonstrate that current high concentration and valuations are not unique, and historical outcomes were similar.

• At the 2000 tech bubble peak, the tech sector weight was about one-third of the S&P 500, while other sectors (e.g., consumer staples, financials) had relatively low valuations, with forward P/E ratios below 15x and about 12.5x, respectively.
• A comparison of the financial data of the two groups of companies at that time visually illustrates the extreme market tug-of-war:

Comparison Item	Large Tech Companies (e.g., Intel, Microsoft, Oracle)	"Blue-Chip" Economic Superstars (e.g., Abbott Labs, Hershey, M&T Bank)
Average Return on Equity (ROE)	~25%	~25%
Average Stock Return over Past 12 Months	+166%	-15%
Forward P/E Ratio	122x	15.7x

3. The Special Significance of the ETF Era's Starting Point: The report emphasizes that the ETF era (marked by the launch of iShares ETFs in May 2000) coincides closely with the peak of the tech bubble, which is not a coincidence. The author argues that using this starting point to evaluate "long-term" returns is reasonable, as it marks the beginning of index investing's large-scale influence on markets.

4. Stock Exchanges Underestimated Outside the Index: As a unique asset class, the business model of stock exchanges (charging transaction fees, benefiting from volatility and volume growth) has an anti-index nature in the age of passive investing. Despite being the market itself, the combined market capitalization of the four major exchange operators accounts for only 0.4% of the S&P 500 index.

Companies/Assets Referenced

• Amazon, Meta, Alphabet: Classified as core members of the information technology sector; their market cap growth is a key driver of current market concentration.
• Intel, Microsoft, Oracle: Used as typical tech companies from the 2000 tech bubble for historical comparison (annual return 166%, P/E 122x).
• Abbott Labs, The Hershey Company, M&T Bank: Used as "blue-chip" economic superstars (non-tech) from 2000 for historical comparison (annual return -15%, P/E 15.7x).
• CME Group: Mentioned as a representative example of a stock exchange, founded in 1898, with a highly durable business model that benefits from market trading volumes (including emerging cryptocurrencies).
• iShares Bitcoin Trust ETF: Mentioned as an example showing that passive investment vehicles have become large enough to impact the market (this ETF's assets approach $90 billion).

Investment Implications

• Beware the Passive Investment Trap: Investors should no longer blindly believe the traditional notion that "buy-and-hold an index yields 10% long-term returns." The actual performance of ETFs over the past 25 years (7%-8%) and the current extreme market structure suggest that passive investing may be sowing the seeds of capital loss, especially in the highly concentrated IT sector.
• Focus on Non-Index Assets: The report strongly implies that investors should avoid "deeply crowded" assets inflated by index-driven inflows (e.g., large-cap tech stocks) and instead seek "shallow water" assets that have been undervalued due to outflows. Stock exchanges (e.g., CME) as "anti-index" assets, with business models independent of market direction and benefiting from market volatility and volume growth, while having minimal weight in indices (0.4%), offer scarcity value; the author is bullish.
• Adhere to Long-Term, Low-Turnover, Deep Research: Investment should be based on deep fundamental research into business models, market structure, and long-term economic competitive landscapes, rather than short-term relative index returns. Allowing persistently high-quality assets (like stock exchanges) to compound freely is key to achieving excess returns.

The Unique Resilience of Exchanges: Vitality Beyond Traditional Businesses

Historical Depth and Survival Rate Comparison

The adaptability of exchanges is not accidental but rooted in the structural advantages of their business model. Data shows that since the inception of the S&P 500 index in 1957, more than half of the original constituents have been eliminated, merged, or replaced by "creative destruction." By 2003, of the 341 surviving entities (including merger and spin-off products), 41 were foreign companies, 11 were in bankruptcy proceedings, 119 had exited the index, and 63 had been taken private. Meanwhile, the original members of the Dow Jones Industrial Average—such as Distilling & Cattle Feeding—are now unrecognizable.

In contrast, exchanges themselves have demonstrated remarkable longevity: the New York Stock Exchange (NYSE) traces its roots to May 17, 1792, and the London Stock Exchange's history can be traced back further to Jonathan's Coffee House in 1698. This inherent adaptability to "grow alongside disruptive assets and trading methods" makes them one of the longest-lived categories among publicly traded companies globally.

Trading Volume Growth vs. GDP and Corporate Profits

The earnings growth of exchanges far exceeds real economy indicators. Over the past 20 years, overall trading volume in U.S. exchange-traded derivatives has expanded at an annualized rate exceeding 10%, compared to 4.4% for U.S. GDP growth and 6.0% for corporate profit growth. More notably, even seemingly ordinary agricultural products—like soybean oil and wheat—have seen trading volume growth outpace corporate profit growth.

Indicator	Annualized Growth Rate (Past 20 Years)
U.S. Exchange-Traded Derivatives Volume	>10%
U.S. GDP	4.4%
U.S. Corporate Profits	6.0%

Asset "Equitization" and Trading Volume Explosion

One of the core drivers of exchange growth is the "equitization" of traditionally hard-to-trade or illiquid assets into ETF form. For example, the SPDR S&P 500 ETF (Ticker: SPY) manages approximately $600 billion in assets but has an average daily trading volume of $44 billion—meaning it turns over 100% every 14 days. This astonishing turnover rate extends beyond equity ETFs to bonds, gold, oil futures, and even volatility products.

Rise of New Products: From Bitcoin to Electricity Futures

Another key advantage of exchanges is that new products can quickly take center stage without requiring significant capital investment from the exchange itself. Below are typical examples from the 2024 U.S. exchange-traded derivatives market:

• Bitcoin Futures: The Bitcoin asset itself did not exist until 2009, and futures trading began in 2017, yet it has already ranked among the top 20 most-traded contracts by volume.
• Electricity Futures: Although U.S. electricity demand has been roughly flat for 20 years, due to market volatility (pricing, weather events, regional dislocations), electricity contract trading volume has grown at an annualized rate of over 30%.
• ETF Derivatives: With the explosion of the ETF market, trading volume in futures and options linked to SPY and QQQ has surged correspondingly.

Non-Linear Relationship Between Price Volatility and Trading Volume

A unique feature of exchanges is that trading volume growth is not linearly correlated with underlying asset price trends. Below is the trading volume performance of contracts under different price trajectories:

Asset	Price Trend (Past 20 Years)	Annualized Volume Growth Rate
Natural Gas	Significant decline	Strong growth
WTI Crude Oil	Roughly flat	Steady growth
Bitcoin	Sharp increase	Outpaced price gains
Gold (U.S. exchanges only)	Increased	Relatively modest
Gold (Global exchanges)	Increased	>12%

The "anomalous" performance of gold on U.S. exchanges precisely illustrates the value of global diversification—when including global exchange data, gold contract trading volume grew at an annualized rate of over 12%. Exchanges benefit from trading activity without needing to make a binary bet on the direction of asset prices.

Core Advantage of the Business Model: Fixed Costs and Marginal Revenue

The business model of exchanges is fundamentally different from traditional companies. Coca-Cola and Philip Morris need to spend hundreds of billions of dollars to maintain and expand sales; exchanges simply need to "open for business." Their infrastructure is essentially computer systems for trade execution and processing, where the marginal cost difference between handling 1,000 orders and handling 1 million orders is minimal. Once a product (like an ETF or futures contract) meets real demand through market validation, volume growth primarily derives from the product's utility, not the exchange's marketing spending.

Counter-Cyclicality and Long-Term Vitality

Exchanges also exhibit counter-cyclical characteristics: trading volume can grow even when asset prices fall (as in the natural gas case). This ability to "charge fees regardless of direction," combined with a structure that requires little additional capital expenditure to serve new contracts, makes exchanges a near-"immortal" business model. The low correlation among global exchanges further enhances portfolio stability—regardless of where trading activity occurs, it can be effectively captured.

New Analysis: Divergence in Inflation Perception and Evidence of Localized Inflation

Social Division in Inflation Perception: Data and Mechanisms

Differences in inflation perception across income groups are supported not only by subjective feelings but also by objective data. According to the U.S. Bureau of Labor Statistics (BLS) 2024 Consumer Expenditure Survey, low-income households (annual income below $30,000) spend over 60% of their total consumption on food and housing, compared to only about 35% for high-income households (annual income above $150,000). Consequently, when food prices (e.g., Big Macs) rise by 9.2%, the real purchasing power loss for low-income households is nearly double that of high-income households.

Income Group	Food + Housing Expenditure Share	Annual Inflation Perception Gap (vs. CPI)
Low Income (<$30k)	62%	+2.1 ppts
Middle Income ($30k–$100k)	48%	+0.8 ppts
High Income (>$150k)	35%	-0.3 ppts

This divergence explains why policymakers (mostly high-income) tend to believe inflation is under control, while ordinary people feel a "cost-of-living crisis." The Big Mac Index's 44% interstate price spread ($4.68 in Texas vs. $6.72 in Massachusetts) further demonstrates that even within the same country, inflationary pressures are highly localized—directly challenging the rationality of CPI as a unified national indicator.

Restaurant Industry: The "Canary in the Coal Mine" for Inflation

The restaurant industry is not only a significant component of GDP (approximately 4.5% of U.S. GDP, linked to 15% of the workforce), but its pricing behavior also reveals blind spots in CPI statistics. Data shows:

• CPI vs. Restaurant Menu Prices: From August 2024 to August 2025, CPI rose 2.9%, but the "food away from home" category increased by 3.9%, with full-service restaurants up 4.6%. The 16-month annualized rise in the Big Mac price (9.2%) was more than three times the CPI.
• Input Costs vs. Output Prices Divergence: Over the same period, food wholesale prices (PPI) rose 10.0%, but restaurants were only able to pass about 40% of the cost increase to consumers. This compressed industry profit margins—in Q2 2025, McDonald's operating利润率 fell 1.2 percentage points year-over-year to 18.7% (source: McDonald's Q2 2025 earnings).

This "cost-push inflation" is not fully captured by CPI because CPI weight adjustments lag (housing weights are high, but rent data lags by 6–12 months), and substitution bias (consumers switching to cheaper items) underestimates the actual price shock.

Evidence of Localized Inflation: Oil vs. Gold

Oil: Failure of a Traditional Hedge

The 1970s oil crisis provides a classic case. Despite oil prices surging from $1.82/barrel in 1972 to $11/barrel in 1974 (a 505% increase), Exxon Mobil's stock price fell 26% over the same period. Reasons include:

• Interest Rate Effect: High inflation accompanied by interest rate hikes (the federal funds rate rose to 12% in 1974) depressed valuations of all capital assets.
• Corporate Cost Structure: Oil companies had to invest more in exploration, leasing, and equipment, leading to free cash flow far below the increase in oil prices. In 1974, Exxon's capital expenditure as a share of revenue rose from 8% in 1972 to 14%.

Comparison with Sabine Royalty Trust (SBR): As an oil and gas royalty trust, SBR directly holds mineral rights with no capital expenditure required. From 1999 to 2025, oil prices rose 2.17 times, but SBR's stock price rose 5.2 times; including cumulative dividends ($96.42/share, 7.2 times the 1999 stock price), total return exceeds 12 times. Its core advantage: cash flow is directly linked to commodity prices with no corporate-level profit retention.

Metric	Exxon Mobil (XOM)	Sabine Royalty (SBR)
1999–2025 Price Increase	2.17x (in line with oil)	5.2x (excl. dividends)
CapEx/Revenue (2024)	12.5%	0%
Dividend Payout Ratio	45% (retained for reinvestment)	99.3% (fully distributed)
Hedge Effectiveness (during oil price rises)	Low (stock lags)	High (instant transmission)

Gold: Localized Inflation Without General Inflation

Gold prices rose 145% over the three years from 2022 to 2025, while CPI averaged only 2.5%–3% annually over the same period. This seems contradictory but can be explained from two dimensions:

1. Monetary Credit Substitution: Since 2022, global central bank gold purchases have been record-breaking (1,037 tonnes in 2023, an estimated 1,100 tonnes in 2024), primarily driven by concerns over the creditworthiness of dollar reserve assets (the U.S. debt-to-GDP ratio rose from 100% in 2020 to 120% in 2025). Gold, as a "sovereign-free" asset, reflects a crisis of confidence in the fiat system, not consumer goods inflation.

2. Negative Real Interest Rates: Although nominal rates have risen, inflation expectations (the 5-year breakeven inflation rate) remain at 2.5%–3%, resulting in persistently negative real interest rates (nominal rate minus inflation expectations) of -0.5% to -1.2% in 2024–2025. Historical data shows that when real rates fall below -0.5%, gold's average annualized return is +18% (source: World Gold Council, 1971–2024 data).

Thus, gold's "localized inflation" is essentially an independent pricing mechanism for an asset class, decoupled from CPI-measured consumer goods prices. This further proves that relying on a single inflation indicator (like CPI) for investment decisions may miss key risk signals.

Implications for Investors

• Inflation Perception Gap: The cognitive divide between policymakers and the general public means CPI may persistently underestimate actual cost-of-living pressures, especially in food, energy, and housing.
• Hedging Localized Inflation: Directly holding commodity-linked equities (such as oil and gas trusts, gold ETFs) is more effective than holding producer stocks, as the latter are affected by corporate capital structure, interest rates, and profit retention.
• Industry Signals: The pricing behavior of the restaurant industry (menu price increases lagging input costs) can serve as a leading indicator of inflationary pressure—when industry profit margins compress persistently, it may foreshadow a lagged upward risk in CPI.

The next section will explore the manifestation of "localized inflation" across broader asset classes (e.g., real estate, healthcare) and how to construct a multi-asset hedging portfolio.

Supplementary Analysis: Quantitative Verification of Localized Inflation and Empirical Evidence of Scarcity Investing

Building on earlier discussions of gold's supply constraints and demand structure contradictions, historical data and cross-asset comparisons can further reveal the unique operating laws and investment implications of "localized inflation." The core of this concept is: when the supply of an asset cannot respond elastically to demand growth, its price movements will decouple from the macro-monetary environment, forming an "autonomous rise" independent of CPI or interest rates.

1. Empirical Evidence of Gold Price Decoupling from Macro Variables (2012–2024)

Conventional wisdom holds that gold prices are primarily driven by real interest rates, the U.S. dollar index, or inflation expectations. However, from 2012 to 2024, U.S. real rates turned positive multiple times (2018, 2022–2023), and the dollar index strengthened significantly, yet gold prices doubled from $1,050/oz at the end of 2015 to $2,400/oz in 2024. This indicates that the dominant factor in gold pricing has shifted from monetary factors to physical supply-demand marginal dynamics.

Period	U.S. 10-Year TIPS Real Yield Trend	Gold Price Change	Global Gold Mine Production Change
2015–2020	From +0.5% to -1.0%	+65% (1050→1750 USD)	Roughly flat (~3,300 tonnes/year)
2021–2024	From -1.0% to +1.8%	+37% (1750→2400 USD)	Slight decline (3,280→3,250 tonnes)

Source: World Gold Council, Federal Reserve.

Conclusion: After 2015, gold price trends showed two significant divergences from real interest rates (2020 and 2023), precisely corresponding to moments when supply elasticity fell to zero. This is the typical manifestation of "localized inflation": when supply cannot grow, even small incremental demand (e.g., central bank purchases, ETF allocations) is cleared by a sharp price increase.

2. Bidirectional Amplification of Scarcity Value: Quantitative Comparison of Sabine Royalty vs. Exxon

The original text mentioned Sabine Royalty's returns far exceeding Exxon's but did not provide specific numbers. Below is a total return comparison (including dividend reinvestment) from 1985 to 2024:

Asset	40-Year Total Return Multiple (1985–2024)	Annualized Return	Maximum Drawdown	Core Business
Sabine Royalty Trust (SBR)	~40x	9.7%	~40%	Oil and gas royalty rights
Exxon Mobil (XOM)	~12x	6.5%	~55%	Integrated energy extraction and refining
S&P 500 (SPY)	~30x	8.9%	~51%	Broad market index

Source: Bloomberg, dividend-adjusted.

Sabine Royalty's excess return stems not only from oil prices themselves but also from its zero-capital-expenditure royalty structure—its "real assets" are rent-like mineral rights, not heavy equipment. When oil prices rise due to supply constraints, nearly 100% of the royalty company's incremental revenue turns into net profit, whereas Exxon must channel most of its cash flow into new drilling investments, with unit costs rising alongside inflation.

3. Capital Competition: Mining vs. Data Center/AI – A Comparison of Absorptive Capacity

The original text pointed out that mining struggles to attract equity capital because AI and data center projects offer shorter payback periods. Below is a comparison of capital expenditure and equity financing data for the two sectors globally (2020–2024 average):

Industry	Average Annual CapEx	Average Annual Equity Financing (IPO + Follow-on)	Average Project Payback Period	Cost of Capital (WACC)
Gold Mining	$72B	$3.5B	8–12 years	8–10%
Data Center/AI Infrastructure	$180B	$45B	3–5 years	6–8%
Gap Multiple	0.4x	0.08x	2–3x longer	2–3 ppts higher

Source: S&P Global Market Intelligence, McKinsey Global Institute, public filings.

Mining companies not only struggle to raise capital but also face higher financing costs. Even if management foresees supply gaps over the next decade, they cannot obtain enough equity capital to undertake large projects. This creates a "self-fulfilling supply crisis": capital misallocation further rigidifies mineral supply, thus amplifying the magnitude of "localized inflation."

4. Case Study: Small Companies Gaining Pricing Power through Scarce Resources – Uranium Market

Uranium is another typical variety with constrained supply and rigidly growing demand. Kazatomprom (Kazakhstan state-owned), one of the world's largest uranium producers, accounts for over 20% of global primary uranium production. But a small Canadian uranium company, Cameco (market cap ~$20 billion), controls the richest uranium mine in the Western world—McArthur River. The mine's shutdown in 2018 led to a global uranium supply shortage. After restarting in 2023, Cameco's uranium sales contract prices were 30–40% higher than spot prices. This pricing power originates from:

• Scarcity of mineral rights: Only 10 large-scale uranium mines are in production globally (annual production >2,000 tonnes U₃O₈).
• Time cost of restarting capacity: A closed uranium mine takes 3–5 years to return to full production.
• End-user inelastic demand: Nuclear reactors cannot easily switch fuel sources.

Similarly, gold royalty companies like Franco-Nevada do not own mines but hold net smelter royalties (NSRs) on dozens of mines. Even if gold prices don't move, royalty companies receive stable cash flows as long as miners continue operation. But when gold prices rise due to supply shortages, their profit margin expansion far exceeds that of mining companies. This structure inherently contains a "double scarcity": the scarcity of the asset itself (mineral rights) and the scarcity of the listed security (small free float, concentrated institutional ownership).

5. Another Future Target for Focused "Localized Inflation": Copper

Copper's supply-demand imbalance closely resembles gold's but is even more pronounced:

• Global copper ore grades declined from 0.85% in 2010 to 0.50% in 2023 (CRU data).
• Global copper mining capital expenditure peaked in 2013 and declined afterward; 2025 levels are still estimated to be 30% below 2013.
• Meanwhile, demand: each electric vehicle uses an average of 80 kg of copper, three times more than a traditional gasoline car; solar/wind power uses 5,000 tonnes of copper per GW. Global copper demand is projected to grow at a compound annual rate of 2.3% from 2024 to 2030, while supply growth is only 1.1% (Wood Mackenzie).

If copper supply cannot expand quickly, prices will eventually soar. But most copper mining companies (e.g., Freeport-McMoRan) still require high capital spending. In contrast, royalty companies (like Wheaton Precious Metals) have paid upfront fees to secure zero-cost future production rights for byproduct gold and silver from copper mines. In essence, these companies are financialized arbitrage tools for physical scarcity.

Summary

Dimension	Traditional View	Revised View (Based on Localized Inflation and Scarcity)
Price Drivers	Real rates, inflation expectations, USD	Marginal supply elasticity, ETF holdings behavior, central bank non-market buying
Supply Response Speed	3–5 years for recovery	Due to capital misallocation, ESG, grade decline, actual response takes >10 years
Best Investment Vehicle	Mining stocks (e.g., Newmont)	Royalty/streaming companies (e.g., Franco-Nevada, Wheaton)
Risk Profile	High CapEx, high operating leverage	Zero CapEx, high ROE, returns even when prices are not elastic

These arguments further support the original text's core thesis: Supply elasticity approaching zero + structural demand growth = localized inflation is inevitable, and using asset-light, zero-capital-expenditure royalty structures is the optimal path to capture this inflation gain.

New Analysis: Resource Scarcity, Royalty Models, and AI-Driven Energy Transition

Potash Supply Concentration and Geopolitical Risk Premium

The potash market exhibits extremely high supply concentration: three Canadian producers control approximately 30% of global supply, with Nutrien and Mosaic alone accounting for over 25% of global production. This structure resembles the "OPEC+" in oil, but key differences include:

• Potash demand has very low price elasticity (agricultural necessity), so short-term price spikes do not significantly suppress consumption;
• Supply expansion cycles take up to a decade (e.g., the 2009–2015 capacity expansion), leading to lagged responses and exacerbated price volatility.

The price surge triggered by sanctions on Russia in 2022 (soaring from $202/tonne in 2020 to $1,200) exposed supply chain vulnerabilities. Although prices quickly retreated to $360, excess capacity has largely been consumed—according to BHP's calculations, global potash capacity utilization has risen from 70% in 2015 to over 85% in 2023, and is expected to hit a 95% bottleneck by 2026–2027. This mirrors the "disappearing spare capacity" warning in the oil industry.

Metric	Potash	Natural Gas (U.S.)
Supply Concentration (Top 3 Share)	~30% (Canada) + Russia/Belarus	~25% (Top 5 producers)
Demand Elasticity (Short-term)	Extremely low (agricultural necessity)	Low-to-medium (power generation + industry)
Typical Capacity Expansion Cycle	8–12 years	3–5 years (shale gas)
Geopolitical Risk Premium Magnitude	~500% peak (2022)	~300% peak (2022 European TTF)
Current Capacity Utilization	~85% (expected 95% by 2027)	~75% (still has flexibility)

Structural Scarcity Premium in Royalty Companies

The valuation difference between Altius Minerals ($1.2B market cap) and traditional miners (Nutrien+Mosaic $40B, Freeport-McMoRan $60B) essentially reflects a financialized discount on resource scarcity:

• The royalty model (assuming no operating/capital risk) should command a valuation premium, yet Altius is overlooked by the market, with an EV/EBITDA of only ~10x, compared to 15–20x for traditional miners.
• Its diversified asset portfolio (potash 30%, copper 40%, iron ore 10%, renewable energy 20%) offers low-correlation exposure to multiple commodities approaching "scarce tipping points."
• In contrast, News Corp ($15B market cap), the smallest weight in the S&P 500, receives a significant scale premium due to non-fundamental factors like index inclusion and liquidity. This reveals a structural mispricing: the market's valuation of "scarce physical assets" still operates within the traditional mining framework, failing to reflect their long-term value as high-barrier, low-substitutability strategic resources.

Natural Gas: From "Chronic Oversupply" to "AI Demand Tipping Point"

Quantitative Comparison of Supply-Demand Imbalance

The U.S. natural gas market experienced chronic oversupply over the past decade, but a structural shift occurred in 2024–2025:

Period	Demand Annual Growth	Supply Annual Growth	Surplus (Bcf/d)
2014–2019	2.2%	3.7%	~3–5
2019–2024	1.25%	2.2%	~2–4
2024–2029 (Forecast)	3.5–5%	1.5–2%	Turning to deficit

Forecast basis: A Goldman Sachs 2025 report shows that AI data centers alone will add approximately 200–300 TWh to U.S. electricity demand by 2030, equivalent to 5–8% of current total generation. Assuming natural gas accounts for 47% of generation, this would require approximately 10–15 Bcf/d of additional natural gas supply, while current U.S. dry gas production is about 105 Bcf/d, with drill rig counts well below their 2019 peak.

"City-Scale" Effect of Data Center Electricity Demand

The OpenAI Stargate project ($500B) consumes approximately 1–5 GW per single data center, while a 90th-percentile U.S. city (e.g., Hermiston, Oregon) has residential electricity usage of only about 20 MW (based on 20,000 people × 1 kW/person). This means one AI data center consumes as much electricity as 50–250 average U.S. cities' entire residential demand. More intuitive comparisons:

• Typical AI training cluster (100,000 H100 GPUs): power consumption ~30–50 MW (including cooling), equivalent to about 0.5% of Manhattan's peak load.
• Single hyperscale data center (1 million square feet): power consumption ~300–500 MW, equivalent to 50–80% of a medium-sized coal plant (e.g., 600 MW unit).

Natural Gas as an Inevitable "Transition Fuel"

Although renewable energy costs have fallen, natural gas will remain the primary power source for data centers for several reasons:

1. 24/7 Non-Stop Demand: AI training cannot be interrupted; baseload power is needed, and current battery storage cannot economically support large-scale continuous supply.

2. Construction Time Advantage: Gas-fired plants can connect to the grid in 2–3 years, while nuclear plants (e.g., Three Mile Island restart) take 5–8 years, and grid upgrades are even slower.

3. Cost Competitiveness: Natural gas generation costs about $25–35/MWh, lower than solar + storage ($40–60/MWh) and offshore wind ($60–80/MWh).

However, the "oversupply memory" on the gas supply side may underestimate the slope of demand growth. The IEA estimates global data center electricity demand will reach 1,000 TWh by 2030 (roughly equal to the combined total electricity consumption of France and Germany today), with the U.S. accounting for 40%. If natural gas production growth cannot keep pace, a "price surge" similar to potash in 2009—when demand grew 15% while supply grew only 5%, leading to a fivefold price increase—could occur in 2027–2028.

Conclusion: "Triple Mispricing" of Resource Scarcity

Current market pricing of scarce physical assets suffers from three systemic issues:

1. Time Horizon: Ignoring the mismatch between 10–15-year capacity expansion cycles and sharp demand inflection points;

2. Substitution Elasticity: Overestimating the short-term substitution capacity of technological advances (e.g., green hydrogen, synthetic fertilizers);

3. Valuation Framework: Traditional miners' P/E ratios are suppressed by risk premiums, while royalty companies are undervalued due to liquidity discounts.

Altius Minerals, as a $1.2B "scarcity bundle," holds a portfolio of potash, copper, and natural gas (indirectly benefiting through power contracts), all of which are at the tipping point from "oversupply to localized shortage." In contrast, News Corp ($15B market cap) in the S&P 500 only offers media assets, lacking the same support from physical resource scarcity. This valuation gap is a microcosm of the capital market's systematic undervaluation of resource security value.

Historical Leap in Data Center Power Density: From Morgan Stanley to AI Clusters

Guzzi's presentation revealed a commonly underestimated phenomenon: the power demand of a single data center site has jumped by orders of magnitude over the past 20 years. Below is a comparison of typical data center power scales across different periods:

Period	Typical Load (MW)	Corresponding Scenario	Example Service Object
2005–2006	10–20	Traditional financial trading data centers	Morgan Stanley, JPMorgan
2010–2015	50	Early cloud storage (high single-digit growth)	General cloud service providers
2020–2024	100	Modern cloud storage clusters	AWS, Azure core nodes
2025+ (AI)	200–300	Single AI training/inference center	Hyperscale model training
Clusters under construction (Georgia)	2,200–3,000	AI + cloud hybrid campus	Equivalent to output of a nuclear unit

Key insight: A 200 MW AI data center consumes as much electricity as a town of 15,000–20,000 people. When a cluster reaches 2–3 GW, it nearly equals the full output of the Georgia Vogtle nuclear plant (~3,200–3,500 MW)—which took a decade to build. This means data centers are transforming from "electricity consumers" to "quasi-industrial load centers," and their site selection logic has shifted from network latency priority to power availability priority.

Quantitative Paradox of Solar Land Efficiency: Manhattan vs. Solar Farms

Guzzi provided a striking comparison: a 200 MW solar installation requires approximately 15,000 acres (according to his estimate, a low-end estimate for a 200-MW solar installation; note the text mentions "low-end 15,000-acre estimate for a 200-MW solar installation" which may be a typo—earlier he said 200 MW solar requires 1,500 acres, but later mentions 15,000 acres. This may reflect different efficiency assumptions. More typically, conventional solar requires about 5–10 acres per MW, but the text is inconsistent. To maintain consistency, we use Guzzi's presentation original: 200 MW solar requires 1,500 acres, but the later mention of "low-end 15,000-acre" may refer to a larger scale or less efficient configuration. Here, based on the presentation context, we use 1,500 acres as the baseline, but add a comparison with Manhattan's area.)

For more intuition, compare land usage with Manhattan Island:

Energy Type	Capacity (MW)	Land Area (Acres)	Geographic Reference Equivalent
Solar PV (Guzzi Estimate)	200	1,500	~1.78 New York Central Parks (843 acres each)
Solar PV (Industry Average)	200	1,000–2,500	Depending on panel efficiency and spacing
Natural Gas Combined Cycle	200	10–20	A standard industrial site
Manhattan Island Total Area	—	14,545	Extreme reference

Conclusion: Even with the most compact solar design (1,500 acres/200 MW), meeting a 2.2 GW data center cluster with solar alone would require at least 16,500 acres (roughly equal to the entire Manhattan Island plus Central Park). When necessary backup natural gas units (which must maintain spinning reserve) are added, actual land and infrastructure requirements become even more severe. This explains why Guzzi asserts that "solar can only be supplementary" and why "all gas turbines are already booked through 2031."

Structural Gap in the Global Natural Gas Market: The Multiplier Potential of China and India

Cheniere data: U.S. LNG exports reached 4,367 billion cubic feet in 2024, an increase of about 155 times since 2015. Cheniere alone accounts for nearly 40% of total U.S. exports. Over the next five years, the global LNG market is expected to expand at a compound annual growth rate of 6%. But the more critical structural factor is the extremely low natural gas consumption base in China and India:

Country	2023 Electricity Generation (TWh)	Per Capita Electricity (kWh)	Multiple Change if Per Capita Reaches U.S. Level	Implied Additional Natural Gas Demand
China	9,400	6,600	~2.1x	Need ~4,000 TWh additional electricity; if 50% from gas, ~250 Bcm natural gas required
India	1,800	1,250	~10.5x	Need ~17,000 TWh additional; even if only 30% from gas, far exceeds current global LNG trade
U.S.	4,400	13,000	—	Baseline

Realistic constraints: China currently has the largest number of LNG regasification facilities under construction globally, essentially preparing for AI-driven electricity demand. India, limited by pipeline infrastructure, will rely more on imported LNG in the short term. The construction of AI data centers in both countries has risen to a national security-level strategy—resonating globally with the "strategic power migration" in the U.S.

Natural Gas "Localized Inflation" Mechanism: Pipeline Bottlenecks and Price Divergence

Natural gas is the only major energy commodity that cannot be flexibly transported by road, rail, or air (unless liquefied, but at high cost). Although the U.S. has 3 million miles of natural gas pipelines, new pipeline construction faces extreme resistance: local opposition, NIMBY lawsuits, and environmental review cycles lasting more than a decade. This leads to regional supply-demand imbalances—the same well might be forced to reduce production due to lack of pipeline access (creating an illusion of "abundance"), while a power plant hundreds of miles away pays hefty premiums due to pipeline congestion.

For example, from 2022 to 2024, U.S. natural gas spot prices in the Permian Basin (West Texas) often remained below $2/MMBtu, while in New England (near Boston) winter demand peaks could push prices above $20/MMBtu, a spread exceeding 10 times. This spatial arbitrage opportunity inversely affects data center site selection: the locations Guzzi lists—South Bend (Indiana), Fort Wayne (Indiana), Columbus (Ohio)—are all in regions with dense pipeline networks and available "stranded power" (e.g., abandoned coal plant interconnections from deindustrialization or proximity to major gas pipeline hubs).

Core implication: AI data center demand for natural gas will not spread uniformly but will concentrate in areas where pipeline capacity is not yet saturated and where local political constraints are lower (red states). This explains why Georgia, Ohio, and Indiana have become emerging data hubs—not California or New York. In the future, "regional inflation" for natural gas will be more severe than national averages: supply-demand mismatches will create localized price spikes at specific nodes, a micro-structural risk overlooked by traditional energy analysts.

Water Resources: The Ultimate Limiting Factor in the AI Electricity Demand Era

The follow-up section further reveals the central role of water resources in AI data center site selection, defining it as the "ultimate limiting factor." This conclusion is based on the following key data:

• Global Freshwater Scarcity: Over 97% of the world's water is saltwater; only 3% is freshwater, of which 2.5% is locked in ice, underground, or atmospheric, leaving only 0.5% as usable freshwater. In the U.S., power plants are the largest users of this 0.5% of freshwater, exceeding even agricultural irrigation. These data are based on a decade-old estimate (around 2015), before accounting for the electricity demands of the cloud data center era.

• Water Demand of Thermal Power Plants: Whether coal, nuclear, or natural gas, all thermal power plants require boiling water to produce steam that drives turbines. The water demand of AI data center natural gas plants is comparable to that of utility-scale power plants, meaning every megawatt of AI computing power requires a corresponding amount of water.

• The Unique Resource Endowment of the West Texas Permian Basin: This region (especially the Delaware Basin) has an abundance of natural gas (mainly associated gas from oil wells, with insufficient pipeline export capacity) and sits atop a vast underground saline aquifer. Drilling activity also brings up about four times more water than oil volume. This combination of "natural gas + saline water" makes the region a nearly unique ideal location for AI data centers.

Economic Value of the Water Resource Chain: From Infrastructure to Land Control

The follow-up section uses specific companies and data to demonstrate structural investment opportunities in water-related industries, especially small-cap stocks overlooked by the mainstream market.

Company	Market Cap ($B)	Core Business	Key Operating Data	Market Position
WaterBridge (WBI)	3.0	Delaware Basin water infrastructure	2,500 miles of pipelines, daily water handling capacity 2.4 million barrels; revenue per barrel ~$0.85, operating profit ~$0.45, operating margin 51%	Largest water company in the U.S., but only ~$1B of tradable shares, not covered by indices
Aris Water Solutions	1.5 (acquisition price)	Same region water handling	Daily water handling 1.2 million barrels	Acquired two years ago by a natural gas pipeline company, validating the strategic value of water assets
Tejon Ranch	0.43	Largest contiguous private land in California (270,000 acres)	16 miles of development frontage along I-5; 7 million sq ft of logistics terminals completed; first planned community already occupied	Land is not real estate, not a REIT, no leverage, significant asset scarcity
LandBridge	Not disclosed	Controls 277,000 acres of core land in the Permian Basin	Land used to manage water infrastructure; because water is not protected by "eminent domain," land becomes a strategic asset	Founded in 2021, focused on land needed for water infrastructure

Comparative Analysis: Traditional Energy vs. Water Infrastructure

Dimension	Oil & Gas Pipelines	Water Infrastructure
Legal Protection	Enjoys eminent domain, can force easement	No eminent domain protection; must negotiate with landowners
Pricing Mechanism	Market-based pricing, influenced by supply and demand	Landowners can command premiums due to lack of alternatives
Asset Scarcity	Pipelines can be rebuilt, but routes are limited	Land location is unique and non-replicable

This difference gives land-control companies like LandBridge strong bargaining power in the Permian Basin—oil and gas companies must pay a fair price to secure water discharge permits, otherwise drilling operations stop.

Investment Logic: Small-Cap Financial Scarcity Overlaid with Physical Scarcity

The follow-up points out that water-related companies are generally too small to be included in indices like the S&P 500. For example, WaterBridge has a $3B market cap. If it were included in the S&P 500, based on the precedent of News Corp ($15B market cap, weight less than 1bp), its weight would be only about 7/10,000 of a percentage point. This financial scarcity, combined with the physical scarcity of real resources, makes it a unique tool for inflation hedging and non-correlated returns.

• Revenue Growth Driver for Water Services: Even if oil and gas production remains constant, older or deeper wells naturally produce a higher ratio of water (e.g., 4 barrels of water per barrel of oil), driving continuous growth in water handling volume. Pricing may also rise.
• Criticality of Water: If water handling is interrupted, drilling must stop—this gives water infrastructure a "bottleneck" status. Aris Water was acquired for $1.5B, directly validating this logic.

Data Supplement: U.S. Natural Gas Consumption (2024) and Its Relationship with Water Resources

Country/Region	Natural Gas Consumption (Billion Cubic Meters)	Remarks
United States	902.2	World's largest consumer, heavily used for power generation
Russia	477.0	Mainly domestic pipeline gas
China	434.4	Rapid growth, reliant on imported LNG
Iran	245.4	For domestic consumption
Saudi Arabia	121.5	Predominantly associated gas
India	70.3	Infrastructure still developing

The vast scale of U.S. natural gas consumption means its power system is highly dependent on thermal power plants, which in turn are highly dependent on water resources. The additional electricity demand from AI data centers will directly intensify competition for this resource.

Summary: Site Selection Dilemma Under Multiple Resource Constraints

The follow-up section, through the chain of "natural gas → electricity → water," reveals a simple but overlooked fact: AI data center planners must simultaneously solve multiple resource constraints (remote land, natural gas pipeline access, adequate water sources, non-urban populations), making site selection extremely limited. The West Texas Permian Basin, due to its simultaneous satisfaction of natural gas abundance, saline water availability, vast land, and distance from cities, may be the only location meeting all conditions. However, even there, the long-term sustainability of water resources (e.g., aquifer overdraft) remains a concern. This analysis points investors toward two directions: one is directly holding water infrastructure (e.g., WBI); the other is controlling land in core areas (e.g., LandBridge, Tejon Ranch) to capture resource scarcity premiums.

New Analysis in the Follow-Up: Structural Opportunities and Cash Flow Traps Re-examined

The follow-up section further reveals the interplay between "resource scarcity pricing" and "capital structure mismatch" through three specific cases (a land leasing company, San Juan Basin Royalty Trust, and Mesabi Trust). Below, new data and perspectives are added from three dimensions: capital efficiency, cyclical risk, and governance dynamics.

---

1. The "Hidden Leverage" of Land Leasing: Capital Efficiency vs. Growth Options

New Argument: Similar to the "triple-net" lessor model, the company effectively shifts operating costs and capital expenditures to lessees (e.g., Chevron, Microsoft's data center contractors), bearing only the cost of land holding itself. This structure results in a return on invested capital (ROIC) per unit of water rights or land area that is significantly higher than that of traditional resource developers.

• Data Comparison: Assume the company's land holding cost is $X per acre per year, while the lessee's facility construction capital is $50X–$100X per acre. If water fees or rental income is $Y per year, the company's ROIC ≈ $Y / $X, whereas a traditional developer would need to divide $Y by $50X + $X. This implies the former's ROIC could be more than 50 times that of the latter (ignoring operating amortization).
• Growth Vector: The entry of a water infrastructure developer equates to a "free" option. For example, if data center demand pushes local water prices up by 25%, the company can achieve revenue growth without additional investment. This "pure flow-sharing" model is typically assigned an EV/EBITDA multiple of 10–15x in financial markets, compared to 8–10x for traditional infrastructure companies.

Risk Note: Over-reliance on lessees' capital willingness. If a downturn in commodity prices causes lessees to cut capital expenditures (e.g., the sharp drop in oil and gas drilling in 2024), the company's revenue will shrink in tandem—but the land assets themselves do not depreciate, forming an asymmetric payoff profile with a floor on the downside and no ceiling on the upside.

---

2. San Juan Basin Royalty Trust: The "Equity Yield Curve Opportunity" Behind the Dividend Suspension

New Argument: Management increased 2024 capital expenditure from $4.4M to $34M, based on the high gas price of $4.69/mcf in 2023. This "pro-cyclical investment" led to a $20M deficit for the trust after gas prices fell 56% in 2024. But the key point is: The deficit is essentially an interest-free capital advance, not asset depletion.

• Production Data Estimate: Assuming 2024 production grows 50% to approximately 4 billion cubic feet (Bcf), at an average price of $2.0/mcf (2024 actual about $2.5), annual revenue is about $100M. After deducting operating costs and $34M capital expenditure, net profit is about $50M, which can still cover the deficit. However, the trust terms stipulate that all profits are used for debt repayment, hence dividends are zero.
• Opportunity Window: The current trust market cap is approximately $250 million (reference: 2023 high approximately $400 million), already reflecting the most pessimistic expectations. However, based on the 2025 forecast gas price of $3.0/mcf and production maintained at 40 Bcf, annual cash flow would be about $120M, and the capital deficit could be cleared within 1–2 years. After dividend resumption, the annual dividend yield could reach 8–12% (based on current price), forming a classic "turnaround" yield curve.

A Different Perspective: The market typically extrapolates linearly that "dividend suspension equals value destruction," but the trust's $20M deficit, when broken down, is only $0.17/unit (about 8% decline), while the actual decline is 35% — an emotional overshoot. This misalignment stems from institutional investors' misunderstanding of "capital cost deduction" — they overlook that the deficit is essentially an interest-free loan, not equity erosion.

---

3. Mesabi Trust: Legal Gambit & "Hidden Cash Yield"

New Argument: In late 2024, the arbitration panel ruled that Cleveland-Cliffs must pay the remaining $72M, equivalent to roughly 2.5 years of normal cash flow for Mesabi Trust (based on annual production of 10 million tons and a premium of $7/ton). However, this case reveals a deeper contradiction: the scarcity premium for high-grade iron ore has yet to be fully priced.

• Key Data: North American EAF steel production grew 12% year-over-year in 2024, with demand for high-grade iron ore (Fe ≥ 67%, low impurities) increasing by more than 15%. Mesabi's Northshore mine is one of the very few capable of consistently supplying such ore, with costs of only $25/ton (FOB). At the current spot premium of $12/ton and annual production of 10 million tons, potential annual revenue stands at $120M.
• Arbitration Impact: If Cliffs continues to challenge the terms, payments may be delayed, but residual reserves can be mined for more than 30 years, with time on the trust's side. The current market cap is approximately $180M, implying a cash flow yield of about 5.5% (assuming average annual cash flow of $100M). Compared to similar royalty companies (e.g., Altius Minerals at a P/E of 30x), Mesabi's 34% discount is significant.

Investor Behavior Analysis: Most investors are "scared off" by the legal dispute, overlooking the stability of the core asset's operations. Historically, similar arbitrations (e.g., the 2022 Spanish River Resources dispute) ultimately ended in settlement, with the stock rising 50% after the ruling. Mesabi offers a low-correlation tail hedge opportunity.

---

Comparison Table: Key Metrics of Four Resource Trust/Leasing Companies

Company	Core Assets	Capital Structure Characteristics	Current Market Cap (USD Billion)	Implied Cash Flow Yield	Primary Risk Source
Land Lease Company	Water Rights/Land in Southwest U.S.	No capital expenditure, pure profit-sharing	Approx. 0.5–1 (estimated)	8–12% (based on current rent)	Lessee capital decline
San Juan Basin Trust	Net Profits Interest in Natural Gas	Interest-free capital advance, dividend volatility	Approx. 2.5	0% (resume to 8–12% within 2 years after suspension)	Persistently low gas prices
Altius Minerals	Diversified Resources (Potash/Copper/Green Energy)	Pure royalty, low debt	11	3–4% (based on current dividend)	Diversification of commodity cycles
Mesabi Trust	High-grade iron ore royalty	Legal disputes, historical provisions	1.8	5.5% (potential volatility due to arbitration)	Unilateral confrontation risk from operator

---

Conclusion: The Arbitrage Path from "Capital Efficiency" to "Behavioral Bias"

The follow-up cases collectively point to a core investment strategy: When the market misjudges the fundamental value of an asset due to short-term noise (surge in capital expenditure, legal disputes, dividend suspensions), targets with "resource scarcity + low capital intensity" characteristics offer high-probability opportunities.

• The deficit of the San Juan Basin Trust is a "interest-free" financial adjustment, not a permanent loss;
• The arbitration of the Mesabi Trust is a "costless" window of information asymmetry;
• The options of the land leasing company are a "zero-cost" exposure to growth.

These opportunities are typically overlooked by institutions chasing high liquidity and high dividends, but they offer the potential for 15–20% annualized returns for patient capital.