Q3 2025 - Market Review
Six months have passed since the chaos that Trump unleashed on Liberation Day[1] and, against all odds, the US economy did not suffer as much as many pundits feared. On the contrary, after having receded by 0.6% during the first quarter, US real gross domestic product (“GDP”) improved 3.8% during the second quarter[2] – the highest level in two years. What is curious is that this is happening at the same time as labor market indicators are weakening. This is perplexing because in the US, spending by households has historically been the most significant driver of economic expansion. Not this time. The most significant step change in GDP growth this time has been capital spending by corporations and, at the root of this, lies the accelerating speed at which firms are building the Artificial Intelligence (“AI”) Infrastructure. It is hard to estimate both direct and indirect impacts on GDP, but some estimate that it corresponds to 1/6th of GDP growth in the past two years and close to 40% when taking into account the necessary upgrades to the power grid and the intellectual property associated with AI itself[3]. On the other hand, some argue that it is curtailing growth elsewhere by keeping power prices high. Nonetheless, equity markets are not lost on this phenomenon and posted strong gains with relatively muted volatility during the 3rd quarter. The S&P 500 Net Total Return Index advanced 8.02%, bringing year-to-date gains to nearly 15% despite a difficult first quarter. The broader MSCI All Countries World Index recorded a 7.98% gain while the S&P TSX Composite Index moved 12.50% higher, not because of its exposure to the AI theme but rather due to gold. More on this further…
From a stylistic viewpoint, growth outperformed value, momentum, and quality globally by a small margin. From a geographic viewpoint, one of the only equity markets that did better than the Canadian equity market, in local currency terms, was the MSCI China A Onshore CNY Index which jumped by more than 21% on rumors of yet another massive stimulus round and the excitement surrounding the 25th Shanghai Cooperation Organisation (“SCO”)[4] convention in Tianjin, which coincided with the celebration of the People’s Republic of China’s victory over imperial Japan 80 years ago.
We have been writing about the increasing concentration risk in the S&P 500 for at least 5 years[5], but the risk only seems to have gotten more extreme since. Not only do the top 10 names in the S&P 500 today represent a higher percentage of the overall index than at any other point in history, but their destiny is also more intrinsically linked to each other than ever before, thereby compounding the risk in the index itself. To this point, in late September, Nvidia announced that it will be investing $100 billion in OpenAI[6], and OpenAI announced that it would be buying $300 billion worth of computing power from Oracle, over 2/3rd of Oracle’s backlog.[7] When a restaurant, a dry cleaner, or a florist chain goes out of business, it does not create a systemic risk. We are not sure that the same happens when an AI domino falls because a lot of the deals that have buoyed the equity markets recently had, at the center of it, Open AI, a private company which is reportedly worth more than $500 billion but which is losing more than $1 billion per month… Perhaps another telltale sign that things are coming to a close is that Fermi Inc, a company co-founded by former Energy Secretary Rick Perry that is looking to develop Project Matador[8], filed for an initial public offering (“IPO”) in late September. What is interesting is that Fermi is structured as a real estate investment trust but has no day one revenue and it will likely take many years to build the infrastructure that will deliver revenues. Nonetheless, at the time of writing this, a little over 2 weeks after the IPO, Fermi’s public equity was worth more than $15 billion.
Considering all this, we remain comfortable with our recommended global equity managers being underweight US equities, on aggregate. We do not know when and how the AI theme will have run its course. It could be when Nvidia starts to feel gross margin pressures, when one or more hyper-scalers announce a reduction in their capital spending program - because they don’t derive enough incremental revenues from it or because their end clients are not able to cut their labor force as much as anticipated with it -, or when a data center operator announces that it cannot find tenants for their more recent buildout. One thing that is clear, though, is that the concept of a capitalization weighted strategy to index US equities has not felt this risky in quite a while.
For the first time in 2025, the Federal Reserve reduced the target FED Funds rate by 0.25%, from 4.50 %to 4.25 % on September 17. While some governors, including the recently appointed, Trump aligned Stephen Miran, were hoping for a more dramatic move, it is worth noting that the situation is somewhat different than at the same time last year when the Federal Reserve proceeded with a 0.50% cut. To this point, at the same time last year, inflation was quickly decelerating, whereas this year it is ticking up, and there is still considerable uncertainty surrounding the percentage of imported goods and services that are going to be absorbed by the end consumer after all the low-cost inventory purchased prior to Liberation Day has been depleted. Nevertheless, Chairman Powell hinted that additional rate cuts are likely before year end. Against this backdrop, fixed income indices posted positive returns during the quarter. For instance, the ICE BofA Canada Broad Market Index, the ICE Bank of America Global Government Bond Index, and the ICE Bank of America Global Corporate & High Yield Index gained 1.42%, 0.41%, and 2.11%, respectively. We highlight that government bonds underperformed as short term rates generally declined, while rates for longer maturities, especially for issued bonds from France, Japan, and the United Kingdom increased noticeably as these countries’ structural challenges flared during the quarter. Separately, we note that credit markets were calm during the quarter until the sudden and quasi simultaneous bankruptcies of auto-parts dealers First Brands and sub-prime auto lender Tricolor hit the news, causing high-yield spreads to widen sharply around the end of the quarter. Some doomsdayers argue that this could be the canary in the coal mine[9] that puts an end to the chase for the yield-at-any-cost theme. At the time of writing this, however, even though creditors, regulators, and rating agencies are still combing through the web of off-balance sheet liabilities owed by First Brands, the damage appears to be contained. In fact, in a preliminary report published on October 8[10], Morningstar DBRS noted that, while portions of First Brands’ roughly $10 billion debt had been broadly distributed across regional banks, public credit funds, business development companies (“BDC”), and collateralized loan obligations (“CLO”), very few entities had more than 1% of the net asset value exposed to the issuer[11]. It is worth mentioning that no insurance companies appear on the list of the 30 most significant trade finance creditors. For the moment, our views align with Morningstar, but we’re old enough to remember that Bear Stearns was also supposed to be contained. Then came New Century, American Home Mortgage, and IndyMac Bancorp in a rapid sequence before the Lehman Brothers debacle. In other words, issuer defaults tend to be correlated over time.
There is a good chance that we write again on this topic when we have more visibility about how this debacle unfolds. As it stands, our families’ portfolios have no exposure to either First Brands or Tricolor and, to the extent that there are traces of high-yield exposure elsewhere in their portfolios, it generally comes in the form of unitranche debt within a simple capital structure that is not excessively leveraged and where terms are designed to be equally beneficial for the issuer and the creditor.
PREVIEWING TRUMP-STYLE CAPITALISM
Quoting Ian Bremmer and Cliff Kupchan of Eurasia Group, we warned in our 4th quarter letter of last year[12] of the implication on the US equity risk premium that could emerge if the US President was to start rewarding:
“…politically aligned business figures with preferential treatment on regulatory, legal, and contracting matters…This will amplify crony capitalism in the world’s largest economy, with risks for firms that must spend more time and money cultivating transactional relationships with Trump’s political apparatus than creating economic value… It will depress economic innovation and productivity as the US government rewards the most politically connected firms (and implicitly or explicitly punishes the rest) rather than the most competitive ones. Corruption will flourish. In the long term, the US would become a less attractive business and investment environment, and Americans would see their living standards stagnate”.
During the second quarter, we saw the first glimpse of what a Trump-style capitalist system might look like. It has taken three different forms: acquiring equity stakes in publicly listed companies, granting preferential treatment to certain publicly listed companies in exchange for payments to the government, and pressure tactics.
With regard to acquiring equity stakes in publicly listed companies, this is not unusual in countries like China but quite unusual in the United States. In fact, while subsidies or tax breaks are prevalent in certain industries, the US government purchase of a passive 9.9% equity stake in Intel (INTC) as part of the CHIPS & Science Act (6.3% or $5.7 billion) and a Department of Defense program (3.6% or $3.2 billion) is extremely rare. The last time it happened was in 2008 when the US government injected capital into multiple financial institutions. The striking difference here is that while Intel had been continuously losing market share, unlike the financial institutions that the US government rescued in 2008, the company wasn’t in immediate danger of failing. Separately, the US government also announced equity participations in rare minerals and metals processors MP Materials (MP) and Lithium Americas (LAC) and mentioned that it was considering plans to acquire a participation in Toronto-based Trilogy Metals (TMQ)[13]. While all these stocks significantly appreciated following the announcement of the US government involvement, we are left wondering how they are going to deal with the political backlash in a downturn if they let go employees or how they are going to fare when bidding for contracts overseas. Importantly, Commerce Secretary Howard Lutnick hinted, without confirming anything, that more deals could potentially be announced, notably in the defense sector.
With respect to the granting of preferential treatments, the most striking example happened when the US government agreed to allow Nvidia and AMD to export certain Artificial Intelligence (“AI”) chips to China in exchange for a voluntary royalty of 15% for such sales. Apparently, that 15% royalty was the price to pay to circumvent the AI chips export ban announced by Donald Trump earlier in the year. The thing is, much like the automotive industry, the semiconductor industry is capital intensive and, to compensate for that, large volumes are required to achieve profitability. We would argue that Nvidia and AMD, because they are the AI chip industry incumbents, are probably the only ones that could afford to make this “voluntary” payment to the US government. We strongly favor government policies that promote free competition. The pay-to-play scheme in which the US government forced Nvidia and AMD into does the opposite.
Finally, the most flagrant example of the government employing pressure tactics took place in August when U.S. Immigrations and Customs Enforcement (“ICE”) conducted a large-scale operation at a Hyundai-LG Energy Solutions battery plant that is in construction in Ellabell, Georgia, leading to the detention of a large number of South Korean nationals and to a diplomatic incident between the United States and South Korea. The raid, which was described as the largest single-site enforcement operation in the history of Homeland Security Investigations, led to the suspension of the facility construction and caused a lot of distress in the small Georgia community. We are not quite sure why ICE conducted itself with such blunt force, but it sends the wrong signal to foreign corporations at a time when the US government is attempting to lure them in with lower corporate tax and regulatory burdens.
The markets have praised the US government’s display of creativity in those deals but, as we wrote before, we doubt that these represent a positive development in the long term as it promotes the wrong set of incentives.
THE YELLOW METAL’S GOLDILOCK SCENARIO
One of our biggest missed opportunities in the past few years has been the rally in gold and the broader precious metals complex. At the time of writing this, the yellow metal broke through $4,000 US/oz for the first time, more than doubling over the past two years. To the great displeasure of the majority of active Canadian equity managers, gold mining equities have done even better because, as it stands, the S&P TSX Materials sector, which is heavily comprised of gold miners, contributed over 1/3rd of the index’s total return year-to-date through the end of September, despite representing just 10% of it at the beginning of the year[14]. We estimate that the structural underweighting of the materials sector by active equity managers explains between 6% and 8% of their underperformance through the third quarter. Historically, not being exposed to gold miners has been beneficial[15]. To this point, the Philadelphia Gold & Silver Index, which is a capitalization weighted index of 30 gold and silver mining companies, has achieved an annualized return of less than 2% since inception in 1979[16] and dramatically underperformed gold itself over the same period. The challenge is that gold miners face cost inflation that often exceeds the increase in the price of gold in addition to resource depletion and financial constraints that often force them to engage in the forward sale of at least a portion of their production at today’s price with a discount. For all these reasons, we believe active managers are right to underweight those stocks. This year, though, this position hurts.
Gold is not part of our portfolio construction framework. The reality is that we don’t understand its behavior very well. We think that few do. We also think that many who think they do, really don’t. Some argue it is a hedge against inflation. Others argue it is a hedge against deflation. Some even think it is a hedge against both or that it is a play on the volatility of inflation! What we can say, however, is that in the past 30 years, the relationship between the price of gold and inflation has been unstable. For instance, from mid 2021 to early 2023, when inflation averaged 5% in the Western Hemisphere, its highest level in 40 years, gold barely moved. We believe that the historical role as a hedge against inflation that gold used to play in most portfolios has been substituted by infrastructure and real estate, with superior results. Some argue that gold is a hedge against systemic or monetary debasement risks, but we would also argue against that because gold’s decline during the Lehman Brothers bankruptcy proved otherwise.
One idea we lent some credence to at one point was that the price of gold might be inversely related to the real yields on other asset classes. Known as the Summers-Barsky hypothesis[17], it suggests that since gold does not provide any income and costs something to store, it only becomes desirable when the real yield on other asset classes declines or becomes negative. Conversely, it becomes less desirable when real yields increase. For that reason, we thought it made sense to own a small portion of gold in late 2020 when large swaths of the fixed income market yielded negative real return and that equity risk premia were getting floored. Yet, gold stagnated again, indicating there was something else in play, and we were not sure what, so we exited. That said, gold is the largest tactical long position for one of the funds included in the Patrimonica Absolute Return Hedge Fund Portfolio, LP. and one of the top contributing positions in the entire portfolio in the past six months.
Gold’s almost uninterrupted ascent, which started in early 2023, seems to have coincided with the shifting central bank policies with respect to currency reserve allocations, notably in China and India, after sanctions were imposed on Russia following its invasion of Ukraine. We believe that China and India became worried that foreign governments could impose sanctions on them for their Russian connections and freeze their fiat currency accounts held in international financial institutions accounts and cause their domestic currency to devalue. As such, China and India have been strategically diversifying their fiat currency reserves by converting them into gold at unprecedented speed, and no doubt Russia would do the same if the country’s war chest wasn’t already plundered by its war efforts. In retrospect, we underestimated both the magnitude and the duration of China and India’s appetite for gold, but now that speculators have fully embraced the story[18], we feel it is getting late to consider it as the proportion of central banks’ foreign currency reserves that is comprised of gold has nearly doubled in the past 3 years[19]. It can hardly get much higher.
To summarize, because gold has historically produced lower returns than other asset classes with comparable volatility over the long term and that its correlative properties are unpredictable on the upside and on the downside, our optimization system has consistently found other asset classes preferable to gold for portfolio inclusion. It is still the case today.
PRIVATE CREDIT COMES OF AGE AND WANTS ITS AI BITE TOO
According to a report published earlier this year[20], the private credit market now exceeds the broadly syndicated loan market and the public high-yield market. Some estimate that its size is over 2 trillion US$[21]. Thanks partly to the banking regulation drive that ensued the Great Financial Crisis and prolonged accommodative monetary policies that pushed investors off the risk curve, private credit has become mainstream.
So much mainstream that private credit is starting to compete with the traditional syndicated loan market for deals, as witnessed in late August when Meta revealed that it had picked Pacific Investment Management Co. (“PIMCO”) and Blue Owl - not traditional investment banks - to finance the construction of the $29 billion[22] Hyperion data center facility that Meta will be leasing for 20 years. It was not the first time that direct lenders undercut traditional lenders[23] but, to our knowledge, it never happened for a transaction of that magnitude even if it appears unlikely that PIMCO will retain the entire debt piece for itself.
The Meta deal was not the only large AI-related deal that grabbed attention during the quarter. In late September, the videogame development company Electronic Arts (“EA”) agreed to be taken private by a consortium comprised of Saudi Arabia’s Public Investment Fund (“PIF”), Silverlake Technology Management, and Affinity Partners in a deal valuing the group that owns the rights to FIFA, Madden NFL, Need for Speed, and The Sims titles, among others at $55 billion. The deal, which is expected to close in 2027, will include $20 billion in debt financing from JP Morgan’s leverage finance arm, making it one of the largest leveraged buyout (“LBO”) transaction on record in current dollar terms[24]. To put things in perspective, EA’s EBITDA has averaged $2 billion per annum in the past 5 years[25], which means that the debt, initially, will be around 10 x EBITDA. For the record, we typically start to get nervous when the debt-to-EBITDA ratio exceeds 5 x. As such, 10 x is considerable leverage. Some LBOs with that degree of leverage have worked well, such as Hospital Corporation of America (“HCA”) which successfully went public just 4 years after Bain had taken it private in 2006. Others, not so much, as Texas Utilities, later renamed Energy Future Holdings, went bankrupt 7 years after being taken private by KKR & TPG as the shale boom caused the debt-to-EBITDA multiple to jump to over 13 x when gas prices collapsed.
Since EA is in the consumer discretionary sector, why do we say that the Electronic Arts LBO is an AI-related deal? Because there is no justification for putting 10 x EBITDA worth of debt for a company that needs to continually invest to renew its intellectual property catalog unless there is an implicit bet that an all-out AI strategy is going to considerably reduce the cost of developing future games. This rationale is the same as that employed by Skydance Media for its acquisition of Paramount Global when the deal closed earlier in the quarter, and it is the same that explains why the combined Skydance-Paramount is now rumoured to be looking to absorb Warner Brothers Discovery. This way, if the next Joker movie is a box office flop like the last one[26], the view is that the extensive use of AI will at least make it a less costly flop.
ChatGPT will soon hit its 3-year anniversary. Since it was unveiled on November 30th, 2022, a basket of AI stocks tracked by Goldman Sachs[27] has appreciated by 776%, more than 10 times the S&P 500’s appreciation over the same period. But the stocks in the basket are all providers of AI solutions or infrastructure providers that are making the AI solutions possible. For the next step, perhaps it is time to start putting together a basket of stocks that will enjoy the biggest operating cost reduction.
Patrimonica’s Investment Team,
[1] April 2, 2025, when trade tariffs were unilaterally imposed on all countries.
[2] Bureau of Economic Analysis, September 25, 2025
[3] The Economist, August 18th, 2025.
[4] Eurasian political, economic, and international security organisation of ten member states (including China, Russia, and India) plus a few observer states. We note that the United States petitioned to obtain an observer status in 2005, but the request was declined.
[5] The S&P 500’s “Metamorphosis”, April 28, 2020
[7] Oracle, OpenAI Sign $300 Billion Cloud Deal, The Wall Street Journal, September 10, 2025
[8] A planned 11GW energy and data center campus in Amarillo, Texas
[9] Popular expression that originates from the historical practice of miners bringing caged canaries into mines; the birds would die before humans if toxic gases like carbon monoxide were present, alerting the miners to evacuate. As such, the death of the canary is an imminent sign of danger.
[10] “First Brands: Limited Fallout for Financial Institutions, Auto Sector, and Private Middle-Market Lenders”, Morningstar DBRS.
[11] Only Franklin Floating Rate Mutual Fund (FAFRX) on the public side and Prospect Capital (PSEC) on the BDC side had more than 1% of their NAV tied to First Brands at the end of the 2nd quarter.
[12] Q4 2024 - Market Review, January 27, 2025
[13] As a side note, the largest equity holder in Trilogy Metals is former Trump campaign financial backer and economic adviser John Paulson.
[14] Source: Factset.
[15] The Philadelphia Gold & Silver Index which is a capitalization weighted index of 30 gold & silver mining companies has achieved an annualized return of less than 2% since its inception in 1979.
[16] Source: Bloomberg
[17] Gibson’s paradox and the gold standard, Barsky, R., Summers, L. Journal of Political Economy, Vol. 96, No. 3 (June 1988), pp. 528-550 (23 pages).
[18] Net long positions in Gold futures by Non-Commercial agent is approaching an all-time high according to the most recent CFTC’s Weekly Commitment of Traders report.
[19] Haver Analytics.
[20] LSTA, Member Firm Survey, April 2025
[21] Global Risk Institute
[22] According to Bloomberg, PIMCO will provide $26 billion in investment grade while Blue Owl will contribute $3 billion and own the facility. Interestingly, the bonds will reportedly have a 24-year term, including 4 years for construction before lease payments begin. Meta reportedly also agreed to a residual value sweetener in case it terminate the lease early or chooses not to renew and that technological developments cause the value of the facility to decline in the interim (such as if, as Jeff Bezos suggests, it becomes cheaper to put data centers in orbit…).
[23] According to Pitchbook, last year’s $16.5 billion Novo Holdings acquisition of Catalent included nearly $5 billion in private debt and a good portion of Sycamore’s $18.2 billion debt for as part of the Walgreen Boots Alliance take private transaction was private debt.
[24] Only the RJR Nabisco ($30.1 billion in 1989) and Texas Utilities ($48.4 billion in 2007) LBO were larger in inflation adjusted terms.
[25] Electronic Arts, 2025 Form 10-K
[26] ‘Joker: Folie à Deux’ to Lose $150 Million to $200 Million in Theatrical Run After Bombing at Box Office, Variety, Octobre 14, 2024
[27] Goldman Sachs TMT AI Leaders basket (Bloomberg ticker: GSTMTAIL Index) which consists of Nvidia and TSMC (Semiconductors), Palantir, Oracle, Microsoft, and Vertiv (software and data centers) and GE Vernova, Constellation Energy and Vistra (power generation).