THE VIRTUES OF DIVERSIFICATION Dominique Marchese, 2026-03-03

Key words: Volatility, Technologies, Artificial intelligence, Diversification, War.

Version PDF

The first two months of the year confirmed the validity of our recommendations: only optimal portfolio diversification, and more specifically caution regarding highly concentrated major stock market indices, allows for performance, while large technology stocks, particularly the Mag7 ("Magnificent Seven"), are suffering from asset managers' shifts towards cyclical stocks, European and Asian markets, and small-cap stocks. Fears related to the impact of artificial intelligence (AI) in many sectors, especially the software industry, have exacerbated the dispersion of sector returns, creating a more favorable environment for active investors. However, sectors benefiting from the AI investment cycle continue to perform well (semiconductors, electrical equipment, etc.). The legality of US tariffs is being challenged by the US Supreme Court, but this has not deterred the Trump Administration. The return of war in the Middle East increases the risks of a new energy shock, and weakens the economic prospects of oil and natural gas importers (Europe, Asia).

The Great Rotation: The Return of Active Management

“There’s no AI bubble” is the most widely held view within the financial community . The ever-increasing demand for computing power and the strong growth in use cases do indeed point to a bright future. So why have the Mag7 index and many other tech stocks seen their share prices suffer since the beginning of the year, despite more than satisfactory quarterly results, which haven't managed to reassure investors about the disruptive risks of AI? The global software publishers index, for example, has contracted by more than 20% since January 1st . The index of major tech stocks on the New York Stock Exchange is posting negative performance . In reality, AI has burst a bubble that we highlighted in 2021: that of SaaS ( software as a service or online software on the cloud), whose valuations, which had become stratospheric after the pandemic, had already been battered by the rise in real interest rates in 2022. This time, it is not a simple actuarial effect on the discounting of future cash flows, but rather fundamental questions about the sustainability of business models .

To summarize the eventful news of the past few weeks, we highlight the presentations by AI giants, such as Anthropic (creator of the powerful Claude language model), of AI agents in vertical markets like finance, law, sales, and coding. These presentations have exacerbated fears of massive shocks in many industries that have so far enjoyed high gross margins and strong competitive positions. These concerns have had collateral effects on credit markets (especially private), as over $500 billion in debt is tied to issuers in the software industry, some of which has been structured as Collateralized Loan Obligations (CLOs ), a practice that brings back painful memories for investors who experienced the 2008-2009 financial crisis. So far, financial markets do not seem able to categorize companies as the future winners or losers of AI. However, in the software and digital services industry, companies that handle proprietary data, possess in-depth expertise in business processes and production workflows, and have developed long-term relationships with their clients within critical infrastructure seem better equipped to withstand new entrants. The cybersecurity concerns appear exaggerated. Software vendors that combine their expertise with AI-powered innovations should fare better. The market will eventually separate the wheat from the chaff. The bursting of the SaaS bubble will undoubtedly create investment opportunities for active investors once the dust settles.

AI has become an unavoidable reality that will revolutionize entire sectors of the economy and boost productivity (around +0.5% per year over the next decade, according to the latest forecasting models). But this revolution is only just beginning. That's why we don't believe the risks lie in the short-term investment cycle, but rather in the robustness of business models , and ultimately in the longer-term return on investment. The risks of cannibalization between technology leaders are very real. For example, in digital advertising, Amazon.com's ambitions threaten the oligopolistic positions of Alphabet and Meta Platforms. In turn, agentic AI threatens marketplaces in digital commerce (routing flows outside of the Amazon.com platform). The risks of downward price pressure in IT services and software should not be underestimated. The more specifically physical constraints (availability of dedicated computing and memory chips, access to the electrical grid, etc.) should not be overlooked, particularly regarding the delayed effects on cash flows, which risk worsening the financial situation of new entrants. These key issues are excellently illustrated by OpenAI (the creator of ChatGPT ), on which a significant portion of the circular economy around AI, worth several hundred billion dollars (partnerships with Nvidia , AMD, Oracle, Coreweave , etc.), is based.

The technology sector is undoubtedly essential for investors in the context of the AI revolution. However, it no longer plays the role of safe haven (or "risk-free" asset) it did during the pandemic, due to increased capital intensity and pressure on cash flow. The technological and financial risks associated with AI cannot be ignored and require the rebuilding of an appropriate risk premium. This is precisely the message that financial markets are sending us at the start of this year. Investors will need to focus more on the strengths and weaknesses of the technologies developed (for example, dynamic or "world" models versus Transformers models), on the sustainability of business models and their " scalability " (the marginal cost of a new user), and on the economic relevance of use cases (value creation).

A high level of uncertainty

War has made its grand return. Incidentally, it is diverting media attention from the illegality of the US tariffs, recently declared by the US Supreme Court. Regarding Iran, at this stage we can only follow events day by day. The most pressing issue is the security of the Strait of Hormuz, through which 20% of the world's oil consumption passes daily (of which only about a quarter could be quickly rerouted) and about a quarter of the liquefied natural gas (LNG) trade. If the conflict lasts only a few weeks, the consequences will remain relatively insignificant on a global scale, and economic forecasts will remain valid for the rest of the year. A longer war and profound destabilization of the entire region would have considerable repercussions. A new energy shock cannot be ruled out if the Strait is not sufficiently secured – but this scenario remains unlikely given the history since 1979. Incidentally, we commend the foresight of the Europeans, whose strategic natural gas reserves are at their lowest levels as winter ends. Europe continues to refuse to sign long-term supply contracts for a primary energy source it considers doomed within the framework of the energy transition – and it is virtually alone in this stance. In the event of a conflagration in the Middle East and widespread destruction of energy infrastructure, Europe may well have to compete with Asia to acquire LNG tankers (particularly from the United States) at spot prices far exceeding forward prices.

As the midterm elections approach, we nevertheless dare to hope that the White House will exercise restraint and base its strategy on a scenario of a short-term conflict, without any certainty, however, of achieving the fall of the current regime in favor of a poorly structured and still unarmed opposition. We point out that in 2025, 95% of the 1.5 million barrels of oil exported daily by Iran were purchased by China, which also sources its oil from other Gulf states. We therefore see little strategic (or economic) benefit in a prolonged blockade of the Strait by Tehran. While the rise in oil prices remains limited for the moment (+10% for North Sea Brent since the start of Operation Epic Fury ), we bear in mind that a sustained 25% increase in crude oil prices would reduce global volume growth (currently on a trajectory of +3% per year) by about 0.5% over a two-year period (source: the International Monetary Fund).

Regarding tariffs, we can only consider the likely scenario of tariffs remaining at 10 to 15% in the coming months. Given the illegality of the application of the 1977 law (IEEPA, for International Emergency Economic Powers Act ), the Trump Administration has already invoked Section 122 of the 1974 Trade Act. Other legal avenues are possible should further unfavorable court rulings be issued (for example, Section 232 of the 1962 Trade Act, and Sections 201 and 301 of the 1974 Trade Act). All of these articles would nevertheless be legally questionable, particularly due to the deliberate confusion between the very real trade deficit and the balance of payments deficit, which is clearly not the United States' primary concern. In reality, Donald Trump is primarily betting on the slowness of legal processes: after multiple appeals, the use of the IEEPA was only invalidated by the Supreme Court ten months after the tariffs came into effect! Ultimately, the critical question that will arise is that of the potential reimbursement of duties already collected (currently up to $175 billion), which promises a formidable mess, and more seriously concerns the trajectory of the US federal deficit and debt.

Conclusion

We anticipated unsettling geopolitical news, increased volatility, and sudden market reversals, which would have the advantage of offering new opportunities at bargain prices. And here we are! Geographic and sector diversification will remain key to performance in 2026. We believe that passive portfolio management strategies, indexed to major indices and indifferent to valuation issues, are the most vulnerable in the current environment of high uncertainty . For balanced portfolios between equities and fixed income, we reiterate our advice to exercise caution in the bond allocation, as risk premiums for lower-rated issuers remain at historically very low levels.