Third Point Q3 2025 Investor Letter

Theme and Background

This chapter is the opening section of Third Point's third-quarter 2025 investor letter. The report first discloses the quarterly performance of the flagship Offshore Fund (net return of 3.2%) and notes that while the market rally was highly concentrated in gold and AI-related stocks, earnings growth supported share prices, a recession remains distant, and the Fed's rate-cutting cycle combined with sustained AI investment provides a favorable environment. The author focuses on the investment logic of the semiconductor sector, particularly the shift in AI computing demand from a single pre-training paradigm to three scaling laws encompassing pre-training, post-training, and inference.

Core Views

• AI computing demand is far from peaking: DeepSeek's breakthrough in LLM training efficiency once sparked concerns that AI computing demand had peaked, but inference models (e.g., OpenAI o1/o3) and reinforcement learning-based post-training have significantly increased computational intensity, with actual demand accelerating.
• The semiconductor sector remains a core beneficiary: Although the U.S. market has fully priced in the AI computing theme, undervalued opportunities still exist in international markets, such as SK Hynix and Ebara.
• HBM (High Bandwidth Memory) is undergoing de-commoditization: AI workloads are driving explosive demand for HBM, and its differentiated characteristics will enhance cyclical returns in the industry and reduce earnings volatility.

Key Arguments and Data

1. Three growth engines for AI computing demand:

• The chain-of-thought process in inference models (e.g., o1/o3) makes computational intensity orders of magnitude higher than previous-generation models.
• Post-training (reinforcement learning fine-tuning) has become a critical step in model advancement.
• OpenAI recently signed an agreement to expand its computing capacity nearly 20-fold over the next five years.

2. SK Hynix's leadership in the HBM market:

• HBM accounts for only 5% of DRAM industry bit volume (2024), but contributes approximately 20% of DRAM revenue.
• Hynix's HBM bit share rose from 8% in 2024 to 14% in Q2 2025, with revenue share increasing from nearly 30% to over 40%.
• HBM pricing is 4-5 times that of traditional DDRx DRAM.
• Valuation comparison:

Company	2026 P/E
SK Hynix	7x
Micron	10-12x
Samsung	10-12x

• Through SK Square (which holds a 20% stake in SK Hynix), investors can gain exposure at nearly a 60% discount, corresponding to a P/E of only 3.5x.

3. Ebara's CMP tool market opportunity:

• CMP (Chemical Mechanical Planarization) is a critical step in semiconductor production equipment. As process nodes advance from N3 to N2 (introducing GAA and backside power delivery), CMP intensity increases significantly.
• The CMP market is a duopoly between Applied Materials and Ebara. Ebara specializes in metal CMP, and as the importance of metal layers grows in advanced packaging (e.g., hybrid bonding), it is poised to gain market share from Applied Materials.
• Ebara's valuation is significantly lower than that of its semiconductor equipment peers, but its profit margins lag. The author believes margins can be improved by over 50% through cost optimization and pricing adjustments.

Companies/Assets Involved

Company	Role	Key Data	View
TSMC	Existing holding	One of the top five quarterly winners	Bullish, core to AI computing buildout
Nvidia Corp.	Existing holding	One of the top five quarterly winners	Bullish, core to AI computing buildout
SK Hynix (000660 KS)	New buy	2026 P/E of only 7x, HBM market leader (>50% share)	Bullish, HBM de-commoditization + valuation discount
SK Square (402340 KS)	New buy	Holds 20% of SK Hynix, corresponding to 3.5x P/E, 60% NAV discount	Bullish, management committed to narrowing discount
Ebara (6361 JP)	New buy	One of two CMP tool duopoly players, valuation below peers	Bullish, margin improvement + AI beneficiary
Micron	Competitor	2026 P/E 10-12x	Neutral/comparison
Samsung	Competitor	2026 P/E 10-12x, concerns over HBM market recovery	Neutral/comparison
Applied Materials	Competitor	The other CMP market oligarch	Neutral/comparison

Investment Implications

• Continue overweighting AI computing infrastructure: Despite significant market gains, AI computing demand has expanded from pre-training to inference and post-training under three laws, and the compute-constrained environment remains unchanged. TSMC and Nvidia are core holdings.
• Explore valuation troughs in international markets: U.S. AI beneficiary stocks are fully priced, while significant discount opportunities exist in South Korea and Japan. SK Hynix (or via SK Square) offers the HBM leader at 7x (or 3.5x) P/E, and Ebara offers a CMP duopoly player at below-peer valuations, both with dual drivers of valuation recovery and earnings growth.
• Focus on the de-commoditization trend in the HBM supply chain: HBM's differentiated characteristics will alter the cyclical nature of the DRAM industry, and SK Hynix, as the market leader, will generate excess returns.
• Actively engage in portfolio company governance: The author is in dialogue with Ebara's new management and board to drive margin improvements of over 50% and narrow the valuation discount. Such active management strategies could serve as catalysts for value realization.

Additional Arguments and Analysis

1. Credit Markets: Structural Opportunities and Systemic Risks Coexist

Deep Insights from the Tricolor Holdings Case

• This case reveals the "credit expansion trap" in the subprime auto loan market: to meet investor demand, originators relaxed underwriting standards (e.g., accepting undocumented immigrants without FICO scores), leading to risk accumulation.
• Although the Tricolor incident did not trigger a systemic crisis, it has prompted banks and investors to intensify scrutiny of specific names, potentially further compressing liquidity in subprime ABS.
• Data comparison: Tricolor's $1.5 billion in outstanding securitizations account for only 2.5% of the $60 billion subprime market, but its default risk could transmit to broader credit markets through CLOs.

Third Point Offshore Fund Q3 net return of 3.2% (annualized 13.2%), CS HF Event-Driven Index Q3 return of 3.1%, S&P 500 Index Q3 return of 8.1%, MSCI World Index Q3 return of 7.4%

Ripple Effects from the First Brands Bankruptcy

• First-lien loan prices fell from $90 to $12 (an 87% decline), far exceeding the 60% recovery rate assumed in CLO models, exposing the fragility of CLO pricing.
• Comparison data: The CLO market has $1 trillion in outstanding volume, while CLOs hold approximately $5 billion of First Brands' $11 billion in debt (45% of its total debt).
• Institutional bank loan spreads are at historical lows (around 300 bps), but CLO liability spreads have rebounded to levels seen at the start of the year, indicating market divergence in pricing high-risk assets.

Credit Market Opportunity Matrix

Asset Class	Current Spread/Price	Historical Percentile	Potential Catalysts	Risk Factors
High-Yield Bonds	Near 2021 lows	10%	Rate cut expectations, corporate earnings improvement	Rising default rates (including restructurings, over 5%)
Leveraged Loans	Spreads ~300 bps	5%	Bank deregulation, CLO demand	Recovery rates 20% lower than 10 years ago
Subprime ABS	Prices down 15-20 points	30%	Bankruptcy liquidations, restructuring opportunities	Deteriorating originator credit quality

2. Corporate Credit: Dual Drivers from Michaels and the Elon Musk Empire

Michaels' Counter-Cyclical Growth

• Despite tariff pressures, Michaels offset most of the impact through price increases, while the liquidation of competitors Party City and Jo-Ann Fabrics created a "vacant market" for it.
• Data: Michaels' Q2 revenue grew 8% year-over-year, with gross margins improving to 42% (from 38% a year earlier), primarily benefiting from product line expansion into home decor and party supplies.
• Risk note: If tariffs escalate further (e.g., a 60% tariff on Chinese goods), the room for price increases may be limited, and cost pressures beyond 2026 warrant attention.

Credit Value of the Elon Musk Empire

• X (formerly Twitter) has undergone a "remarkable transformation" under Musk's leadership, but specific financial data has not been disclosed. X.AI, as an AI startup, is capital-intensive, but Musk's personal resources (e.g., Tesla's $1.5 trillion market cap) provide implicit guarantees.
• Key event: On November 6, Tesla shareholders will vote on whether to formalize the investment relationship with X.AI. If approved, Tesla's investment-grade rating (Moody's Baa2) would directly enhance the credit quality of X.AI's debt.
• Comparison: X.AI debt currently yields around 12%, while comparable AI companies (e.g., OpenAI) have debt yields of around 8%, with the premium reflecting market concerns about its reliance on Musk's personal credit.

The "Skyscraper Curse" Controversy of the EA LBO

• The $55 billion LBO sets a record, but its leverage ratio (approximately 6x EBITDA) is far below levels seen before the 2008 financial crisis (e.g., HCA's 8x).
• Key difference: Current corporate debt structures are predominantly floating-rate (70%), whereas in 2008 they were mostly fixed-rate, making the impact of rising interest rates on debt service more direct.
• Historical comparison: Blackstone's 2007 acquisition of Hilton ($26 billion) saw its value halve during the 2008 financial crisis, but it ultimately turned profitable through restructuring. The success of the EA LBO will depend on whether AI gaming demand can sustain growth.

3. Structured Credit: Dual Benefits from Falling Rates and Regulatory Easing

The "Discount Monetization" Strategy for Mortgage Portfolios

• Most loans currently held were purchased at a discount (average discount rate of 85%), with the expectation of principal repayment through refinancing once the rate-cutting cycle begins.
• Data: Borrowers' home equity ratios range from 40-55%, well above the historical average of 30%, meaning even a 20% decline in home prices would still cover loan principal and interest.
• Regulatory catalyst: Bank deregulation expected in 2026 will lower financing costs and boost demand for first-time homebuyer loans. Current non-performing loan prices have already risen by 10 points (to 90% of face value), above normal loan prices (85%), reflecting market expectations of accelerated liquidation.

"Distressed Bargain" Opportunities in the ABS Market

• Solar ABS and net lease ABS (NNN) were hit hard by the 2022-2023 rate surge, with some issuers facing bankruptcy or rollover risk.
• Case study: A solar ABS issuer defaulted due to rising financing costs (from 4% to 8%), causing its senior ABS price to fall to 80% of face value. However, the underlying assets (solar farms) have stable cash flows (PPA contracts locking in 20-year electricity prices).
• Strategy: Buy senior tranches at a 15-20 point discount, with an expected recovery rate of over 90% and an annualized yield of 12-15%.

4. Macro Risk: The "Fiber Optic Bubble" Analogy for AI Infrastructure Investment

Mismatch Between Investment Scale and Market Capacity

• The high-yield bond and leveraged loan markets total approximately $3 trillion, but the capital requirements for AI infrastructure (data centers, power) could be several times that amount.
• Historical lesson: During the 2000 internet bubble, demand for fiber optic capacity seemed limitless, but wavelength division multiplexing (WDM) technology suddenly doubled capacity, leading to massive stranded investments.
• Potential risks:
• Technology substitution: Quantum computing could render existing AI chips obsolete, causing data center investments to depreciate.
• Insufficient demand: The commercialization of AI applications may be slower than expected, leading to excess computing capacity.
• Social impact: AI replacing human labor could trigger mass unemployment, leading to an economic recession.

Balancing Opportunities and Challenges

• If AI investment succeeds, it will create new credit asset classes (e.g., data center REITs, AI infrastructure bonds), providing new targets for structured credit.
• If it fails, it could trigger a credit crisis similar to 2008, but current leverage ratios are lower (corporate debt/GDP around 75%, versus 85% in 2008), making systemic risk manageable.

5. Team Expansion and Strategic Positioning

Professional Backgrounds of New Members

• Maarten Bauters: Previously at Apollo Global Management's Hybrid Value fund, specializing in distressed asset investing; Goldman Sachs industrial group background, covering manufacturing and energy.
• Lukas Schwarzmann: Private equity experience at Blackstone, focusing on leveraged buyouts; PJT Partners restructuring group experience, specializing in crisis management.
• Strategic significance: The addition of these two members strengthens Third Point's capabilities in "distressed investing" and "private credit," aligning closely with current credit market opportunities.

Summary

Third Point's credit strategy in Q3 2025 reflects a "both offensive and defensive" approach:

• Offensive: Capturing distressed trading opportunities through events like Tricolor and First Brands, and leveraging structural growth from Michaels and the Elon Musk empire for excess returns.
• Defensive: Locking in stable returns in a declining rate environment through discounted mortgage portfolios and senior ABS tranches.
• Risk control: Maintaining vigilance toward AI infrastructure investment to avoid repeating the 2000 fiber optic bubble.