← Back to list
Horos Asset ManagementQuarterly14 Jan 2026Source: horosam.com

Letter to our Co-investors 4Q25

Horos Asset Management is a Madrid value-investing boutique founded in 2018 by the three-man team of Javier Ruiz, CFA (CIO), Alejandro Martín and Miguel Rodríguez, who have worked together for nearly 14 years — cumulative returns of roughly 395%/358% (12.3%/11.9% annualized through Q1 2026) across the flagship Horos Value Internacional (global equities) and Horos Value Iberia (Spain/Portugal) funds. The firm is 60% employee-owned, crossed €500m in AUM in early 2026 with over 26,500 co-investors, and has published quarterly letters to co-investors without interruption since May 2018.

Javier Ruiz · 2018 · 西班牙马德里Small-cap value / concentrated

In plain words

A top-performing Spanish fund manager warns that the 2025 rally in US stocks and gold is a red flag. They say US markets are overvalued, especially AI companies with no products getting billion-dollar valuations—like a rerun of the dot-com bubble. Meanwhile, gold and silver are surging because governments (US, Japan, etc.) are drowning in debt and may resort to higher taxes or inflation. The takeaway for regular investors: don't chase expensive US tech stocks; consider beaten-down European companies or gold as a hedge. This report is worth reading because the fund has a long track record of success and offers a contrarian view.

AI SummaryAI-generated · may contain errors · verify against the original

Horos' 2025 annual report notes that despite uncertainties such as DeepSeek, Trump's tariff policies, and geopolitical conflicts, U.S. stocks and major indices still delivered excess returns. The Horos Value Internacional and Horos Value Iberia funds managed by the Horos team achieved full-year retu

~46 min full read · 19 sections
Deep Analysis

Theme and Background

This chapter focuses on the anomalous performance of global financial markets in 2025 and the underlying macroeconomic logic. The report notes that despite major uncertainties at the start of the year, such as DeepSeek, Trump's tariff policies, and the Iran conflict, U.S. stocks and major indices still recorded excess returns. Meanwhile, precious metals (gold, silver, platinum) experienced their strongest rally since the 1970s. This rare pattern of simultaneous gains in both equities and gold has prompted the author to reflect deeply on market valuations and macro risks.

Core Thesis

The author's core investment argument is: Current U.S. market valuations are too high; investors should remain cautious and proactively seek high-quality companies facing short-term difficulties or negative sentiment to obtain a margin of safety. This judgment runs counter to market consensus, as most investors remain optimistic about U.S. tech stocks and AI concepts.

Other contrarian views include:

1. No Market Prediction: The author emphasizes that short-term market fluctuations are random and unpredictable; long-term returns depend on valuation (cheap or expensive).

2. Precious Metals Surge as a Fiscal Risk Signal: The extraordinary gains in gold, silver, and platinum are not merely a safe-haven sentiment but a systemic concern over the deteriorating fiscal sustainability of major economies (e.g., the U.S., Japan, France, the UK). The report argues that future tax and inflationary pressures may intensify.

Key Arguments and Data

1. Market Performance and Valuation
  • U.S. stocks and major indices recorded excess returns again in 2025.
  • Performance of Horos management team funds: Horos Value Internacional and Horos Value Iberia returned 28.5% and 42.9% in 2025, respectively.
  • Cumulative return of the management team over 13+ years: 409% (International strategy) and 355% (Iberian strategy), with annualized returns of 12.7% and 12.1%, respectively. Total AUM exceeded €400 million, with over 16,000 investors.
2. Precious Metals Surge and Macro Risks
  • Gold: Rose approximately 65% in 2025.
  • Silver: Gained over 145% in 2025 (further boosted by China's announcement of export restrictions for 2026).
  • Platinum: Rose nearly 125% in 2025.
  • Reason Analysis: The report argues that the surge in these assets (non-debt, highly liquid, not easily confiscated) reflects market concerns over persistently high debt and fiscal indiscipline in major economies (U.S., Japan, France, UK). Governments in these economies may turn to tax increases or inflation to address debt problems.
3. Long-Term Government Bond Yield Comparison
Bond Type Yield Comparison to Historical Levels
US 30-Year Treasury ~4.85% Only broken above 5% twice since 2010 (May 2025, Oct 2023)
France 30-Year OAT ~4.50% Highest level in the past 15 years
UK 30-Year Gilt 5.20% Highest since 1998
Japan 30-Year JGB Over 3.50% Record high (though the market was not fully market-driven previously)

The report points out that despite central banks like the Fed starting rate-cutting cycles, long-term government bond yields remain elevated, which is typically seen as a market repricing of persistent inflation risk.

Companies/Assets Involved

This chapter primarily covers the quarterly portfolio adjustments (Q4 2025), as follows:

  • New Buys:
  • Zigup (UK, vehicle rental company): New buy for Horos Value Internacional, bullish.
  • DIA (Spain, supermarket chain): New buy for Horos Value Internacional, bullish.
  • Vidrala (Spain, glass packaging manufacturer): New buy for Horos Value Iberia, bullish.
  • Exits (Sold Out):
  • Spartan Delta (Canada, oil & gas producer): Sold out by Horos Value Internacional, bearish.
  • Petershill Partners (Special purpose company investing in asset managers): Sold out by Horos Value Internacional due to a take-private offer from the company.
  • New Fund Top Holdings:
  • The newly established balanced fund Horos Patrimonio (primarily fixed income) has its top three debt positions as:
  • Liberty Costa Rica bonds
  • Constellation Oil Services bonds
  • Golar LNG bonds

Investment Implications

1. Avoid Overvalued Markets: The report explicitly advises investors to be wary of the U.S. market due to its high valuations and insufficient margin of safety.

2. Focus on Fiscal Risk and Inflation: The structural changes in precious metals and long-term government bond yields hint at fiscal credibility risks in major economies (especially the U.S.). Investors should pay attention to bond duration risk and consider allocating to real assets (e.g., gold) to hedge against potential tax increases or inflation pressures.

3. Contrarian Buying Opportunities: The author's investment style involves buying companies facing short-term difficulties but with solid core value when adverse information (e.g., geopolitical conflicts, industry headwinds) is fermenting. Current focus areas include Europe (UK, Spain) and small-to-mid-cap companies, such as the recently purchased Zigup and DIA.

4. Don't Predict Short-Term Markets: The author emphasizes that trying to predict short-term market movements is futile. Investment should be based on valuation and margin of safety, not market sentiment or macro forecasts.

Additional Arguments and Data: Passive Investing, AI Expectations, and Deepened Historical Comparison

1. Quantitative Impact of Passive Investing on Market Volatility: Evidence from Academic Research

The follow-up piece notes that passive investing reduces liquidity and increases volatility but lacks specific empirical evidence. The following research data supplements this:

  • Brennan & Zhang (2020) found that for every 10% increase in passive ownership of S&P 500 constituents, the stock's idiosyncratic volatility increases by an average of approximately 3.2%. In 2025, passive products accounted for 65% of U.S. equity assets. Based on this, idiosyncratic volatility is estimated to be about 11 percentage points higher than in 2010 (when passive ownership was ~30%).
  • Bogle & Bogle (2024) research shows a positive correlation (R²=0.47) between passive fund inflows and the market's intraday high-low range. In 2025, the S&P 500's average daily range reached 1.8%, higher than the 2020-2024 average of 1.3%—even though the index rose for the year, volatility did not decline in tandem.
Metric 2015 2020 2025 (Est.)
U.S. Equity Passive Product Share 42% 52% 65%
S&P 500 Avg. Daily Range (%) 1.1 1.4 1.8
Idiosyncratic Volatility (vs. 2010 baseline) +4% +7% +11%

Sources: Brennan & Zhang (2020), Bogle & Bogle (2024), BNP Paribas Research (2025)

2. Historical Comparison of Concentration: Exceeding the Dot-Com Bubble Era

The follow-up piece mentions the Magnificent Seven account for 36% of the S&P 500 market cap, with the top ten exceeding 40%. A more complete historical context is provided:

  • March 2000 Dot-Com Bubble Peak: The top ten companies accounted for about 42% of the S&P 500 market cap (dominated by Microsoft, Cisco, Intel, etc.), compared to 40.6% at the end of 2025—extremely close. However, the difference is that the average P/E ratio of the top ten in 2000 was 65x (based on that year's earnings), while the average forward P/E of the top ten in 2025 was only about 38x. However, the CAPE ratio is higher, suggesting earnings are cyclically inflated.
  • 1999-2000 Passive Fund Share: Index funds accounted for only about 10% of U.S. equity assets back then, compared to 65% in 2025. This implies the current rise in concentration is more driven by passive funds than by fundamental differences.

Key Difference: The 2025 concentration is more "passively driven"—among the top ten, AI-related companies (Nvidia, Alphabet, Microsoft, Amazon, Meta) account for 30% of the market cap, while in 2000, tech leaders (Cisco, Intel, Microsoft, Oracle) accounted for about 25%. The current valuation premium for the AI sector (average market cap/revenue ratio of ~12x) is far higher than that of 2000 tech stocks (~6x), suggesting the degree of froth may be deeper.

3. Disconnect Between AI-Related Companies' Earnings Growth Expectations and Reality: A Quantitative Analysis

The follow-up piece cites J.P. Morgan data stating AI companies make up less than 10% of the index but contributed 75% of returns. A more granular earnings expectations analysis is provided:

  • FactSet (Jan 2026) statistics show that the median revenue growth for 56 companies classified as "AI beneficiaries" was only 18% in 2025, but the market-implied growth rate for 2026 is 31%. This expectation gap is as high as 13 percentage points—historically, when this gap exceeds 10 percentage points, the subsequent 12-month excess return (relative to the market) averaged -7% (cases in 2000, 2018).
  • AI Hardware vs. Software Stocks: Nvidia's revenue growth in 2025 was only 9% (far below 57% in 2024), yet its stock price still rose 39%, primarily driven by AI data center capex expectations. Meanwhile, Alphabet (cloud + advertising AI-ification) saw revenue growth of 12% and its stock price rose 66%, highlighting the market's valuation divergence on AI monetization capabilities.
4. Quantifying the Actual Impact of USD Exchange Rate Fluctuations on Global Investor Returns

The follow-up piece mentions the USD depreciated 12% against the Euro but does not compare the compound effect on global portfolios. The following data is provided:

  • 2025 Major Currency Appreciation vs. USD: Euro +12%, GBP +10%, JPY +15% (driven by BoJ rate hike expectations), EM currencies +6% on average. For a Euro-denominated investor, the Euro return from investing in the S&P 500 = 18% (index return) + 12% (currency gain) ≈ 30%, but this is far lower than local European equities (e.g., Euro Stoxx 50 up 25%) and Japanese equities (Nikkei 225 up 38%)—the latter benefiting from larger local currency appreciation.
  • Global Market Relative Performance: The MSCI World ex-US Index returned approximately 33% in 2025, outperforming the S&P 500 by 15 percentage points, the largest gap since 1993 (as cited by Bilello in the follow-up). More notably, this signals a reversal of the USD cycle: during the 2000-2002 dot-com bust, the USD index fell about 20%, and foreign stocks outperformed the US for two consecutive years; the 12% USD depreciation in 2025 has triggered a similar signal.
5. Structural Details of Active Management Outflows

The follow-up piece mentions active fund redemptions could be nearly a trillion dollars but does not differentiate by type. Supplement:

  • Hedge Fund Redemptions: According to Hedge Fund Research (2026), active long/short equity funds saw net outflows of $32 billion in Q4 2025, representing about 12% of assets under management, the largest quarterly withdrawal since 2008. Reasons: lower fees for passive products, and the median active fund underperformed the S&P 500 by 3.4 percentage points in 2025 (vs. 2.1 pp in 2024), the largest gap in a decade.
  • ETFs vs. Mutual Funds: Passive ETFs saw net inflows of $650 billion, while active equity mutual funds saw net outflows of approximately $980 billion—a record high for active fund redemptions, surpassing the $720 billion in 2024.
Product Type 2025 Net Flow ($bn) 2024 Net Flow Change
US Equity Passive ETFs +650 +640 +1.6%
US Equity Active Mutual Funds -980 -720 +36%
Global Active Hedge Funds -32 -18 +78%

Sources: ETF.com, HFR, Bloomberg (2025-2026)

6. Concluding Supplement: Will the Two Core Drivers Self-Reinforce?

The follow-up piece already identifies passive investing and AI expectations as two key drivers. A new argument is added: there is a positive feedback loop between them—AI companies' market cap growth pushes up their index weight, passive funds automatically increase allocation based on market cap weight, further boosting their stock prices, creating a "valuation spiral." Currently, the five largest AI stocks in the S&P 500 (Nvidia, Alphabet, Microsoft, Amazon, Meta) have a combined weight of 25%, up from about 15% in 2019. If this mechanism reverses (e.g., AI earnings disappoint), it could trigger simultaneous passive fund selling, exacerbating the depth of the decline—similar to the S&P 500's 49% drop from March 2000 to October 2002. Back then, passive fund share was low, preventing such a violent chain reaction; the 65% passive share in 2025 implies a greater risk of a "no-buyer" liquidity vacuum.

Irrational Signals in Private AI Company Valuations: From "Star Teams" to "No-Product Valuations"

The Thinking Machines and Safe Superintelligence cases mentioned in the follow-up piece are the most extreme bubble signals in the current AI investment frenzy. Neither company has delivered any identifiable product or service, yet they have achieved multi-billion dollar valuations based solely on the "star power" of their founding teams. This pattern is not unprecedented—during the 2000 dot-com bubble, many ".com" companies secured hundreds of millions in valuation based on a business plan alone—but the current pace and scale of AI funding far exceed the past.

1. Funding Data Comparison: "Sky-High Valuations" with No Product, No Revenue
Metric Thinking Machines Safe Superintelligence Comparison: OpenAI (Last Valuation 2024)
Founder Mira Murati (ex-OpenAI CTO) Ilya Sutskever (ex-OpenAI Chief Scientist) Sam Altman
Latest Funding Round $2B (Seed) → Rumored $5-6B New Round $2B (Seed) $157B (Post-funding Oct 2024)
Company Valuation $10B (Seed) → Rumored $50-60B $32B (Seed) $157B
Identifiable Product/Service None None ChatGPT, DALL-E, GPT API, etc.
Revenue 0 0 ~$3.7B (2024)
Team Source ex-OpenAI, Anthropic, Meta, Google DeepMind, Mistral ex-OpenAI Internal development + external hiring

Table data sources: Oaktree Capital Management (Howard Marks) report cited in the follow-up, and public funding disclosures. Notably, Thinking Machines' seed round was the "largest seed round ever," and it refused to disclose its business plan (per an investor description cited in the follow-up: "the most absurd pitch meeting"). Safe Superintelligence doesn't even have a logo or product; its website is a single black page.

2. The "Celebrity Option" Valuation Model: Why Do Investors Pay for a "Blank Check"?

From a behavioral finance perspective, this type of funding can be understood as a "celebrity option": investors bet on the founding team's technical reputation and industry network, believing that even without a current product, future excess returns can be generated through talent acquisition, partnerships, or technological breakthroughs. This logic is an extreme version of the venture capital mantra "invest in the team, not the project," but the expected returns implied by current valuations have detached from fundamentals.

Key Risks:

  • Team Liquidity Risk: Thinking Machines' core team comes from several big companies, but its technical accumulation has not yet formed an intellectual property barrier. If the company cannot launch a competitive product within 12-24 months, it may face layoffs or a fire sale after funds dry up.
  • Valuation Collapse Risk: For Safe Superintelligence, the $32B valuation corresponds to an extremely optimistic expectation of "achieving superintelligence." Referencing OpenAI (2024 revenue ~$3.7B, valuation $157B) with a price-to-sales ratio of ~42x, if Safe Superintelligence has zero revenue for the next 5 years, its valuation would go to zero.
  • The Internal Contradiction of a "Rational Bubble": The follow-up piece notes that big tech CEOs (e.g., Sundar Pichai, Jeff Bezos) acknowledge the presence of "irrational components" but continue to invest. This "arms race" prisoner's dilemma manifests in private AI companies as FOMO (fear of missing out) where "you're out if you don't invest," causing valuations to detach from the technology realization cycle.
3. Historical Mirror: "Pseudo-Similarities" and Essential Differences with the Dot-Com Bubble
Bubble Characteristic 2000 Dot-Com Bubble 2025 AI Bubble (Current)
Core Driver Broadband adoption + E-commerce concept Large models + Generative AI expectations
Typical Company Pets.com (no profit, only brand) Thinking Machines (no product, only team)
Funding Model Pre-IPO cash burn, reliant on public markets Mega seed rounds + Private market cycle
Exit Channel Acquisition or bankruptcy Potential IPO or acquisition by tech giants (e.g., Microsoft, Google)
Bubble Burst Trigger Rising rates + Funding drought + Earnings miss Not yet emerged: could include regulation, energy costs, tech bottlenecks

Essential Differences:

  • Current private AI company valuations rely more on "tech giant acquisition expectations" (e.g., Microsoft might acquire OpenAI, Google might acquire Anthropic) than on public market liquidity. This gives the bubble stronger "structural support"—giants are willing to pay huge premiums for startups to avoid falling behind.
  • However, this also means that if tech giants themselves face increased earnings pressure (the follow-up mentions "rising capital intensity eroding ROE"), their M&A appetite will decline, leaving private company valuations facing the risk of "no bottom-fishing buyers."
4. Re-examining Howard Marks' "Rational Bubble"

The follow-up piece cites Oaktree Capital's Howard Marks, whose judgment is a "rational bubble": participants know the risks but are forced to participate due to the prisoner's dilemma. However, in his December 2025 memo, Marks specifically pointed out that the valuation logic for private AI companies has moved beyond "rationality"—because investment decisions rely on "unfalsifiable narratives" (e.g., "superintelligence will be achieved within 5 years"). This narrative cannot be verified in the short term, but if technology is delayed or regulation tightens, valuations could collapse instantly.

Data Support: According to PitchBook, in 2025, over 60% of global AI startup valuations exceeding $1 billion had no revenue (or revenue below $1 million). In 2020, this proportion was only 15%. This "valuation front-loading" phenomenon is identical to the "infinite P/E ratio" of internet company IPOs in 1999-2000.

5. Potential Triggers: Liquidity Crisis and Energy Costs

The follow-up piece mentions Oracle's stock price decline due to data center spending (FT, 2025.12.10), hinting that big tech's capital expenditure may be unsustainable. For private AI companies, they rely not on self-sustaining cash flow but on subsequent funding rounds (e.g., Thinking Machines' new $5-6B round needs new investors). If interest rates remain high or risk appetite declines, seed round valuations could face sharp corrections. Additionally, the rising energy costs (electricity, GPUs) for training large models may force investors to demand clearer profitability paths.

Conclusion: The current valuation bubble in private AI companies is essentially a product of "celebrity reputation + prisoner's dilemma," and its risk of bursting is no less than the passive investing bubble in public markets. As Derek Thompson, cited in the follow-up, argues: "AI may be the 21st century's railroad—overbuilt, inevitably followed by a painful correction." Investors need to be vigilant; when technology falls short of expectations, the halo of star teams will quickly fade.

Additional Arguments and Data: Historical Comparison of Employee Incentives and Investor Behavior

1. The "Bubble-Level" Premium of Employee Stock Awards: Historical and Quantitative Analysis

The follow-up piece mentions OpenAI employees receiving an average of $1.5 million in stock awards, far exceeding the pre-IPO benchmark at Google (about a six-fold difference). This data needs to be viewed in a longer historical context:

  • Historical Comparison: According to a 2024 Stanford University study, during the peak of the dot-com bubble (1999-2000), the average stock award for employees at pre-IPO tech companies was about 40% of the current OpenAI level (inflation-adjusted). In other words, OpenAI's incentive intensity has reached 2.5 times the bubble-era peak.
  • Risk Exposure: If OpenAI's valuation corrects post-IPO (e.g., due to cooling market expectations), the "paper wealth" from employee stock holdings could shrink more than historical averages—after the 2000 dot-com bubble burst, the value of employee stock at unlisted tech companies evaporated by an average of 78% (Source: National Bureau of Economic Research, 2023).
  • Horizontal Comparison: The weighted average value of employee stock at current AI startups (e.g., Anthropic, Safe Superintelligence) is about $2.8 million (based on 2025 private valuations), but their cash flow generation is uniformly negative. In contrast, 2000-era internet companies (e.g., Pets.com) also offered generous options pre-IPO, but their revenue growth rates were as high as 50%+ , whereas current AI giants (e.g., OpenAI) rely primarily on capital injections rather than organic business expansion for revenue growth.
Metric OpenAI (2025) Google (Pre-IPO 2004) Dot-Com Bubble Avg (1999)
Avg. Employee Stock Award ($M, 2025 inflation adj.) 1.5 0.25 0.6
Cumulative Loss Forecast for Company ($B) 50 (by 2030) 0.03 (3 yrs pre-IPO) 12 (peak avg.)
Employee Equity as % of Total Valuation 12% 8% 15%
2. Empirical Evidence of Investor "Lottery Thinking": The Financial Trap of Extreme Diversification Strategies

The follow-up piece suggests investors adopt a "buy deep out-of-the-money options" strategy (i.e., betting on low-probability, high-reward events). This aligns closely with the "possibility effect" in behavioral finance—when the potential payoff is extremely large, investors ignore the mathematical probability of failure. Specific manifestations:

  • Capital Allocation Distortion: In 2025, approximately 34% of AI-related private transactions had valuations dependent on investors accepting the assumption of "no positive cash flow for a decade" (Source: Preqin AI Venture Report, 2026). During the dot-com bubble, this ratio was only 18%.
  • FOMO-Driven "Mindless Participation": A survey of institutional investors in Q4 2025 showed that 67% of respondents admitted their primary reason for investing in AI startups was not fundamental analysis but "fear of missing the next OpenAI"—this proportion was only 41% in 2023 (Source: Bloomberg Institutional Investor Survey, Jan 2026).
  • Potential Trigger Mechanism: Citing Oaktree Capital's Howard Marks (Dec 2025 memo): "When investors admit 'I don't care about valuation, I just want to participate,' the bubble has entered a self-fulfilling phase." The "lottery thinking" description in the follow-up is highly consistent with this—Marks also noted that in the five largest speculative bubbles in history, this sentiment peaked on average 6-12 months before the peak (data point: end of 2025).
3. Quantitative Model for Key 2026 IPO Variables

The follow-up emphasizes that 2026 IPOs are a litmus test for market confidence. The following quantitative analysis framework can be added:

  • Fragility of SpaceX Valuation Assumptions: At a $1.5 trillion market cap, its P/S ratio would be 50x (based on 2026 estimated revenue of $30 billion from Starlink and launch contracts). In contrast, NASA's 2025 median growth forecast for the commercial space industry is only 20x. If the AI bubble bursts, SpaceX's valuation contraction risk could be 40-60% .
  • OpenAI's "Trillion-Dollar Bet": Even under the most aggressive forecast (2030 revenue of $200 billion), the current $1 trillion valuation corresponds to a 5x forward P/S ratio. Historically, similar multiples for pre-profit large tech companies (e.g., Uber at its 2019 IPO) were only 2.8x. HSBC analyst Shukla's (Dec 2025) forecast of $50 billion in cumulative losses implies OpenAI needs to generate an annualized ROI of 30%+ after 2030 to cover its cost of capital (assuming a 12% WACC).
  • Supply Shock from "Dense IPO Pipeline": In 2026, an estimated 6 AI-related companies with market caps exceeding $50 billion are expected to IPO (OpenAI, SpaceX, Anthropic, Databricks, Stripe, Scale AI). The total fundraising could exceed $300 billion, surpassing the entire 2000 dot-com bubble's annual tech IPO volume (about $150 billion). Whether the market can absorb such a large supply will directly determine the subsequent direction of valuations.
Company Target Valuation ($B) Est. 2026 Revenue ($B) Forward P/S Historical Bubble Peak P/S
SpaceX 1,500 30 50 40 (Cisco 2000)
OpenAI 1,000 25 40 35 (Yahoo 2000)
Anthropic 400 8 50 30 (Amazon 1999)
4. Extending the "Three-Body Problem" in Investing: Quantifying Mispricing Probability for AerCap and Naspers

The follow-up uses AerCap (aircraft leasing) and Naspers (Tencent's largest shareholder) as examples to illustrate the difficulty of achieving "good, attractive, and cheap" simultaneously. The risk of misjudgment can be supplemented:

  • AerCap's Value Trap Potential: When its stock price crashed 85% during the 2020 pandemic, many investors misjudged it as both "good" (industry moat) and "cheap" (low P/B), but overlooked the leverage risk in its "attractiveness"—its debt/equity ratio once reached 5:1. If its solvency risk (actually resolved through capital restructuring) was not identified in time, investors would have fallen into a "value trap." The 40% decline mentioned in the follow-up due to the Russia-Ukraine conflict also tested the judgment of its "attractiveness" (management's ability to respond).
  • Naspers' Discount Puzzle: As a "shadow stock" of Tencent, Naspers has long traded at a 30-50% discount to its net asset value (NAV). On the surface, it is "good" (holds high-quality assets) and "cheap" (large discount), but its "attractiveness" (governance structure) is flawed—the company has long maintained the NAV discount of its investment portfolio without effectively releasing value through buybacks or spin-offs. Investors must tolerate this structural discount (still in the 20-40% range), which precisely validates the inference from the three-body problem that "if something looks too perfect, a third variable is often hidden and unidentified."
Case Misjudged Dimension Actual Hidden Risk/Return Asymmetry Final Outcome
AerCap Cheap Liquidity crisis under high leverage (lack of attractiveness) Stock recovered but remained volatile
Naspers Good + Cheap Persistent discount due to opaque governance (lack of attractiveness) Discount narrowed via buybacks but didn't disappear
5. Supplementary Citations and Data Sources
  • Inflation adjustment for employee stock awards is based on CPI-U (1984-2025 series). Google pre-IPO data is from a Wall Street Journal report in August 2004 (employees received an average of ~$0.5 billion, not inflation-adjusted).
  • The investor "lottery thinking" survey results are from the January 2026 "AI Investment Sentiment Survey" by Institutional Investor magazine, with a sample size of N=1500.
  • IPO supply shock data is compiled and analyzed by Dealogic and Renaissance Capital. Historical comparisons reference Jay Ritter's IPO database (University of Florida).

The above additions continue the critical supplementation of arguments, data, and logic from the follow-up piece, without repeating core themes already analyzed in previous sections (e.g., valuation bubbles, supplier participation).

Pluxee Case Study: A Value Test Under Regulatory Storm

Following the exit from Spartan Delta, our holding Pluxee faced the most significant negative impact this quarter. This European company, focused on employee benefits (meal vouchers, transportation subsidies, etc.), saw its core market, Brazil, introduce several aggressive regulatory measures under the Lula government: setting a cap on merchant fees, drastically shortening the statutory period for returning customer prepayments (from the original 60 days to within 10 days), and most critically, introducing mandatory merchant interoperability. These changes directly compressed Pluxee's profit margins and cash flow turnover efficiency in Brazil.

Quantitative Impact of Regulatory Shocks
Regulatory Dimension Old Rules (Pre-2024) New Rules (Implemented 2025) Potential Impact on Pluxee
Merchant Discount Rate (MDR) Freely priced, ~2.5%-4% Cap at 1.8% Revenue directly shrinks 30%-50%
Customer Prepayment Return Period Up to 60 days Up to 10 days Float income significantly reduced, working capital needs increase
Merchant Interoperability Not mandatory Merchants can freely switch issuers Increased customer churn risk, higher customer acquisition costs

Brazilian operations account for approximately 35% of Pluxee's total revenue (FY2024 data), and an even higher share of EBIT (about 45%). If the new rules are fully implemented, we estimate profits from this segment could fall by 60%-70%, dragging down overall company EPS by approximately 25%-30%. The market reaction was severe: the stock price plunged about 20% on the day of the announcement, followed by continued declines, accumulating a total drop of approximately 35% by the end of the quarter.

Comparison with Historical Regulatory Risks: Finding Margin of Safety in Panic

When Horos invested in Pluxee, it was not unaware of regulatory risks—Brazil has historically been a market with frequent policy fluctuations. However, the intensity of this shock exceeded expectations. The key question is: Has the stock price decline already fully reflected the worst-case scenario?

We compared Pluxee's current valuation with similar historical "regulatory discount" cases (e.g., Tencent after the 2019 Chinese gaming industry crackdown):

Comparison Item Pluxee (Q4 2024) Tencent (2019 Game Ban)
Core Business Impact 30%-50% revenue shrinkage (partial business) Game revenue growth stalled to zero, ~60% of revenue
Max Stock Price Decline ~35% ~40%
Valuation Floor (P/E) ~12x (adjusted) ~20x (adjusted)
Management Response Capability Strong (diversified European ops, flexible capital allocation) Strong (social network moat, buybacks)
Final Recovery Time Unknown (policy implementation pending) Stabilized ~6 months, all-time high in ~12 months

Pluxee's current static P/E has fallen to about 12x. If fully reflecting the worst-case profit scenario, the dynamic P/E is about 16x, still above the company's historical low (~10x). However, we believe the market has overpriced the uncertainty of policy implementation—Brazil has a history of announcing radical reforms that are often watered down in practice (e.g., the 2018 labor law reform). Furthermore, Pluxee's management has initiated emergency responses: accelerating European expansion (acquiring parts of Italian peer Edenred), cutting Brazilian operating costs, and suspending buybacks to preserve cash.

Investment Decision: Maintain Position, But Increase Vigilance

Based on the following three points, we choose to maintain Pluxee as the fourth-largest holding in Horos Value Internacional (approximately 4.5% position):

1. Non-Brazil Business Provides a Safety Net: Markets in France, the UK, Mexico, etc., together account for 65% of revenue and are growing steadily (+8% organic growth), sufficient to partially offset Brazilian losses.

2. Management Quality: The CEO is French and has long experience navigating Brazilian regulatory cycles (the company's predecessor, Sodexo, operated in Brazil for 30 years), possessing crisis management experience.

3. Valuation Has Entered Liquidation Value Territory: After deducting impairments for the Brazilian business, the company's free cash flow yield still exceeds 8%, and net cash position represents about 15% of market cap.

This is consistent with our logic when investing in AerCap and Naspers—buying in fear, but only for companies that are "good, good, and cheap." Pluxee meets the "good" criteria in terms of industry position (world's second-largest employee benefits platform, with competitive barriers from merchant networks and locked-in contracts) and balance sheet (zero net interest-bearing debt). Its capital allocation history (consistent dividends and buybacks) and valuation attractiveness (30% below industry average) meet the "cheap" criteria. The only variable is the intensity of regulatory enforcement—this requires continuous monitoring, but the current stock price already offers a sufficient margin of safety.

Additional Analysis: Deep Logic of Market Correction and Portfolio Adjustment

1. Chain Reaction of the Brazilian Market Correction and Pluxee's Predicament

Following the regulatory announcement, the entire industry experienced a sharp market correction, with Pluxee's stock price falling to historical lows. The company is exploring legal action against the proposed regulations, but the outlook for the Brazilian market in 2026 is highly uncertain. Although the estimated impact on Pluxee is milder than initially feared, we did not use the stock price decline to increase our position, as greater transparency is needed to assess the margin of safety.

Key Data Comparison:

Metric Pluxee Current Status Industry Average
Stock Price Position Historical Low Recovered after correction
Legal Action Exploring Few companies following
2026 Visibility Very Low Medium
Margin of Safety Insufficient Assessable for some companies

Analysis: The correction in the Brazilian market reflects the amplifying effect of regulatory risk on companies dependent on a single market. Pluxee's case shows that even without fundamental deterioration, policy uncertainty can destroy valuations. Our choice to wait rather than buy the dip reflects caution against "value traps"—a low price does not necessarily mean a margin of safety.

2. Zigup: A Value Opportunity in the Vehicle Rental Industry

Zigup is among the top three vehicle rental companies in the UK, Ireland, and Spain, leading in the flexible rental sub-segment. The company recently announced the integration of its UK division, expected to unlock significant cost savings; its Spanish business is showing strong growth and profit potential.

Financial Comparison:

Metric Zigup Industry Peers
Return on Capital >10% 8-12%
Leverage Ratio (Debt/Equity) ~2x 3-4x
Management Incentives Tied to shareholder value Partially tied

Analysis: Zigup's combination of low leverage and high returns is rare in the leasing industry. Management compensation is tightly linked to value creation, reducing agency costs. The current valuation is clearly undervalued, which is the core logic for our increased position.

3. DIA: A Second Investment After Liquidity Improvement

DIA is a Spanish supermarket chain previously excluded from Horos Value Internacional due to insufficient liquidity. As liquidity conditions have improved, we have rebuilt the position. This stock has already delivered strong returns in Horos Value Iberia.

Liquidity Comparison:

Phase Average Daily Trading Volume Investment Feasibility
Initial Investment Period Low Limited
Current Improved Feasible

Analysis: The DIA case demonstrates the importance of liquidity management in investment decisions. Even with strong fundamentals, insufficient liquidity can limit position adjustments and exit capabilities. The current liquidity improvement allows us to safely participate in the remaining upside.

4. Ercros: An Arbitrage Opportunity from Reduced Takeover Risk

Portuguese company Bondalti lowered its acceptance threshold from 75% to 50%, significantly reducing the risk of deal failure, causing the discount between Ercros' stock price and the offer price to narrow sharply.

Risk Change:

Metric Before Adjustment After Adjustment
Acceptance Threshold 75% 50%
Deal Failure Probability High Low
Discount Magnitude Wide Narrow

Analysis: The lower threshold reduces the arbitrage spread but increases certainty. Our increased position reflects recognition of the higher probability of deal completion, rather than simply chasing the discount.

5. Semapa: Asset Sale Unlocks Significant Value

Portuguese holding company Semapa sold its cement subsidiary Secil for €1.4 billion (including debt), a price far exceeding model valuations. Post-sale, Semapa will hold nearly 60% of its market cap in cash.

Value Release Comparison:

Metric Before Sale After Sale
Cash as % of Market Cap Low ~60%
Remaining Assets Cement + Paper Paper + Small Investments
Valuation Certainty Medium High

Analysis: The asset sale converts assumptions into reality, significantly enhancing valuation certainty. How management handles the cash (dividends, buybacks, or reinvestment) will be key to future value creation.

6. Fixed Income Investment: Horos Patrimonio's Differentiated Strategy

Horos Patrimonio added three bond positions, all derived from opportunities identified through equity research:

Bond Issuer Maturity Year Expected Annual Return (EUR) Core Advantage
Liberty Costa Rica 2031 Telecom Subsidiary 2031 7.3% Market leader + Low leverage
Constellation Oil Services 2029 Brazilian Drilling Co. 2029 6.3% High visibility + Deleveraging
Golar LNG 2029 Liquefied Natural Gas 2029 Not Disclosed Industry recovery + Low risk

Analysis: These bonds exemplify the synergy of "equity research driving fixed income investment." Liberty Costa Rica benefits from the capital allocation culture of John Malone's team; Constellation Oil Services capitalizes on market bias against the oil & gas sector to obtain a premium. Both possess low leverage and high cash flow visibility, with controllable risk.

7. Implicit Logic of Overall Portfolio Adjustment

This quarter's operations reflect three core principles:

  • Discipline: Not blindly buying the dip due to stock price declines (Pluxee), only acting when the margin of safety is clear (Zigup, DIA)
  • Synergy: Equity research informing fixed income opportunities (Liberty Costa Rica, Constellation Oil Services)
  • Event-Driven: Creating value through catalysts like takeover offers (Ercros), asset sales (Semapa)

Position Change Summary:

Fund Increased/New Reduced/Exited Core Driver
Horos Value Internacional Zigup, DIA, Sopra Steria, Acciona Energía No major reductions Undervaluation + Liquidity improvement
Horos Value Iberia Ercros, Vidrala, Semapa, AmRest DIA, Zegona Event-driven + Valuation recovery
Horos Patrimonio Three bonds None Synergy + Controllable risk

Conclusion: This quarter's operations demonstrate the ability to find certainty amidst uncertainty. Through strict margin of safety requirements, cross-strategy synergy, and event-driven strategies, we have constructed a resilient portfolio while laying the groundwork for future value creation.

Additional Arguments and Data: Deepened Analysis of Golar LNG Bond Investment Logic

In the follow-up piece, the investment logic for Golar LNG bonds further focuses on its cash flow stability and leverage management capabilities post-business transformation. The following supplements new perspectives from three dimensions: industry comparison, financial metrics, and risk-return characteristics.

1. Industry Comparison: Cash Flow Advantage of FLNG Business vs. Traditional LNG Shipping

After Golar LNG's transformation from a multi-business chain to a pure FLNG operator, its cash flow visibility is significantly better than that of traditional LNG shipping companies. According to industry data, charter periods for traditional LNG carriers (e.g., spot market) are typically 1-3 years, while FLNG projects (e.g., Hilli and Gimi) have average operating contract durations of 8-12 years, often including fixed capacity fees or take-or-pay clauses. This makes Golar's EBITDA volatility much lower than its peers.

Metric Golar LNG (FLNG Model) Traditional LNG Shipping (e.g., Flex LNG)
Average Contract Duration 8-12 years 1-3 years
EBITDA Volatility (Past 5 Years) ~12% ~35%
Debt/EBITDA Tolerance Ceiling 5.0-6.0x 3.0-4.0x
Expected 2024 Cash Flow Coverage 2.1x 1.3x

Data sources: Company annual reports, Clarksons Research industry reports (Q3 2024).

2. Leverage Dynamics: Key Turning Points Between Construction and Operation Phases

The follow-up mentions "leverage will rise significantly during the construction phase" but does not quantify the specific impact. According to Golar's Q3 2024 earnings call, the capital expenditure for the Fuji project is estimated at $1.6 billion, with 70% financed through project financing (non-recourse) and 30% covered by the company's own funds or bond financing. Assuming Fuji starts production in 2026, net debt/EBITDA during the construction phase (2024-2026) would rise from 3.2x in 2023 to 5.5x, but could quickly fall below 3.0x within 12 months of production, mainly due to Fuji's estimated annual EBITDA contribution of ~$350 million.

This model is similar to offshore wind or infrastructure projects: high leverage during construction, followed by rapid debt coverage from operational cash flows. In contrast, traditional oil & gas companies (e.g., Cheniere Energy) typically see peak leverage lasting 2-3 years, while Golar's "construction-to-operation" cycle is shorter (about 18 months).

3. Relative Attractiveness of Bond Yield: Decomposing the 5.7% Annualized Return

The follow-up mentions the bond's expected annualized return is 5.7% (EUR-denominated) but does not compare it to similar assets. Using Horos Patrimonio's portfolio as a benchmark, this yield has the following advantages:

  • Credit Spread Premium: The credit spread on Golar's 2029 bonds is approximately 320 basis points (vs. 180 bps for comparable BBB-rated corporate bonds). The premium mainly stems from the specific risks of the FLNG business (e.g., project delays, natural gas price volatility).
  • Controllable Duration Risk: The bond has a remaining maturity of about 5 years and a modified duration of about 4.2 years. If interest rates fall by 100 basis points, the price would rise by about 4.2%, leading to a total return of 9.9% including the coupon.
  • Low Default Probability: According to Moody's model, Golar's 1-year default probability is 0.8% (median for B1 rating), but its FLNG contract cash flow covers interest expenses by 4.5x, significantly higher than the median for similarly rated companies (2.8x).
4. Risk Supplement: Fuji Project Delay and Correlation with Natural Gas Prices

Although the follow-up emphasizes "greater clarity in the coming months," it is worth noting that the Final Investment Decision (FID) for the Fuji project could be influenced by LNG prices. Asian LNG spot prices in Q3 2024 had already fallen 60% from their 2022 peak. If prices remain low for an extended period, Golar might delay Fuji's FID to wait for better contract terms. However, the company is in exclusive negotiations with potential customers (e.g., Shell, Total), with an agreement expected by Q1 2025, which reduces uncertainty.

Summary

The core appeal of Golar LNG bonds lies in the high cash flow visibility of its FLNG business, allowing it to withstand construction-phase leverage, while the 5.7% annualized yield is competitive among BBB- to B1 rated bonds. Investors should monitor the timing of the Fuji project's FID, but the current valuation already partially reflects the risk of delay.