Dear Investor,
During the Third Quarter, Third Point returned 3.2% net in the flagship Offshore Fund, bringing our year-to-date performance to 7.2% gross (5.4% net). This performance reflects a rebound after lagging the high-yield market in the first half of 2025, partially due to costs associated with hedges taken to protect capital during Liberation Day.
### AI Compute Demand and Technological Advances
TSMC’s breakthroughs in large language model (LLM) training and inference efficiency, including their use of automated reinforcement learning and mixture-of-experts model architectures, initially sparked fears that AI compute demand had peaked. However, reality has proven quite different. While efficiency gains for existing models have been substantial, new AI capabilities and architectures have more than offset those efficiencies, leading to a substantial acceleration in AI compute demand.
The introduction of reasoning-based models, popularized by OpenAI’s o1 and o3, has taken LLMs beyond simple pattern recognition to executing complex, multi-step tasks critical in mathematics, coding, and scientific analysis. Due to reasoning’s chain-of-thought process — where models process multiple tokens before producing a final output — reasoning-based LLMs are several orders of magnitude more compute-intensive than their predecessors.
In addition to reasoning, LLM training has evolved beyond the initial “pre-training” phase, where models studied vast amounts of data before release. A key element of progress today lies in “post-training,” where reinforcement learning techniques fine-tune models through a mix of human and automated feedback. Borrowing Jensen Huang’s framing, AI compute has moved from a single scaling law (pre-training) to three: pre-training, post-training, and reasoning.
This increase in compute intensity has coincided with an acceleration in user attention, as evidenced by public comments from Microsoft in the Third Quarter.
### Credit Market Overview
While high-yield spreads continued to grind tighter, now approaching the lows of 2021, the leveraged loan market paints a different picture. Defaults, including restructurings, are running above 5%, and recovery rates are more than 20% below levels seen a decade ago. This divergence continues to create attractive opportunities.
Our largest contributor for the quarter was Michaels (MIK), the craft retailer. As anticipated, Michaels delivered strong second-quarter results and issued favorable guidance for the remainder of the year. Although tariff exposure remains a challenge, the company has successfully raised prices to offset much of this impact. Additionally, the liquidation of competitors Party City (PRTY) and Jo-Ann Fabrics has opened significant whitespace for Michaels to expand product offerings in adjacent categories.
Our credit exposure to the Elon Musk empire also contributed positively. We hold significant positions in both X and X.AI Corp, Elon Musk’s artificial intelligence company. Under Mr. Musk’s leadership, X has undergone what we see as an extraordinary turnaround. While AI remains an early-stage, capital-intensive industry, we view these securities as extraordinarily cheap given Mr. Musk’s support, resources, and access to capital.
On November 6, Tesla (TSLA) shareholders will vote on formalizing and expanding Tesla’s investment ties with X.AI. We believe this measure is likely to pass, and Tesla’s investment-grade rating and $1.5 trillion market capitalization would provide a further catalyst for X.AI debt.
At month-end, Electronic Arts (EA) announced the largest leveraged buyout (LBO) in history, valued at $55 billion. Some have drawn parallels to the “skyscraper curse,” suggesting that record-setting deals foreshadow financial instability. While hubris often plants the seeds of its own undoing, we believe comparisons to pre-GFC mega-deals are misleading. The leverage backdrop today is not analogous to the housing bubble or systemic risks that triggered the 2008 crisis.
That said, we are closely monitoring the unprecedented scale of investment in AI infrastructure, particularly in power and data centers. The scale is staggering: the combined total high-yield and leveraged loan market is about $3 trillion, yet the capital being discussed for AI buildout runs multiples of that figure. How this ultimately unfolds remains uncertain.
It is highly unlikely that demand and supply will grow smoothly and in tandem. While linear (and exponential) extrapolations may forecast an end state, history shows that growth is rarely smooth, and significant speed bumps are probable. For example, during the 2000 internet boom, demand for fiber optic cable seemed infinite until wavelength division multiplexing advancements instantly multiplied capacity.
We do not yet know if an AI analogue will emerge. Possibilities include technological shifts like quantum computing rendering giant investments stranded, or chip technology advances accelerating obsolescence. Conversely, broad substitution of human labor by AI might displace jobs at a scale sufficient to risk sparking a deep recession. Either way, the implications—and opportunities—will be enormous.
### Structured Credit Update
Following the Federal Open Market Committee’s (FOMC) September rate cut and market expectations of further cuts before year-end, spreads in structured credit generally tightened over the quarter as bank and insurance capital continued to buy investment-grade risk. The rate rally—led by the 10-year Treasury hovering around 4% and tightening credit spreads—is constructive for our predominately fixed-rate asset portfolio.
We believe this creates a strong environment to buy whole loans backed by hard assets and access securitization financing. In our current mortgage portfolio, most loans are held at a discount, and we have been waiting for expansionary monetary policy to monetize these investments. Both performing and non-performing loans are expected to rally as rates drop, increasing refinancing capacity for borrowers with 40–55% equity in their homes, resulting in par paydowns.
With bank deregulation anticipated in 2026, bank financing spreads continue to trend lower, which should drive higher demand for first-lien, owner-occupied residential mortgages. Prices for delinquent mortgage loans have risen almost 10 points over the year and are higher than those on currently paying loans. Because non-performing loan investors can accelerate liquidation timelines on loans with high home equity percentages, these assets are more valuable and should further enhance returns on our residential mortgage portfolio in coming quarters.
In asset-backed securities (ABS), we have remained in senior tranches and opportunistically added to credits such as solar and triple net lease ABS. These business models were challenged by the steep rise in rates in 2022 and 2023, facing bankruptcy or extended maturity dates. While the underlying credits remain strong in our view, negative issuer headlines created a distressed trading environment where we could add risk at 15–20 points down.
### Business Updates
We are pleased to welcome Maarten Bauters and Lukas Schwarzmann to the equities team during the Third Quarter.
– **Maarten Bauters** was previously an associate in the Hybrid Value fund at Apollo Global Management and earlier worked in Industrials investment banking at Goldman Sachs. He holds a Master’s in Finance from Columbia University and an M.S. from KU Leuven.
– **Lukas Schwarzmann** was most recently a Senior Associate at Blackstone Private Equity and began his career in the restructuring group at PJT Partners. He holds a B.A. in Applied Mathematics from Harvard University.
Sincerely,
Daniel S. Loeb
CEO
—
*Editor’s Note: The summary bullets for this article were chosen by Seeking Alpha editors.*
*Editor’s Note: This article discusses one or more securities that do not trade on a major U.S. exchange. Please be aware of the risks associated with these stocks.*
https://seekingalpha.com/article/4837143-third-point-q3-2025-investor-letter?source=feed_all_articles