Compass 2nd Quarter 2026
AI disruption between market conditions and geopolitical risks
AI disruption between favorable market conditions and geopolitical risks
Since OpenAI introduced ChatGPT in November 2022, artificial intelligence (AI) has become the dominant topic in financial markets. Its significance lies less in immediate macroeconomic disruptions and more in its potential to boost productivity and corporate profitability and gradually support long-term growth. Some data points already suggest significant productivity gains for various office jobs. AI has moved away from experimental use toward broad economic application, although the degree of technology adoption remains highly heterogeneous.
At the same time, macroeconomic conditions are favorable. Growth in the leading economies is moderate, inflationary pressures have generally eased over the past three years, while credit markets have remained stable so far and overall market liquidity appears robust. Should AI adoption continue, the resulting higher productivity and accompanying disinflationary effect could lead to a “Goldilocks” scenario featuring solid growth and stable inflation.
For financial markets, however, the implication is most likely not a uniform trend across all sectors and asset classes. The benefits are likely to be unevenly distributed due to differences in the pace of adoption, capital intensity, and monetization of the technology, which will widen the gap between winners and losers. Companies with strong market positioning, pricing power, and scalable, resilient business models should be better positioned than firms facing rising costs without a clear path to productivity gains.
The main risks to a positive scenario are geopolitical tensions and persistent fiscal deficits. An escalation of geopolitical uncertainty could trigger an inflation shock and worsen conditions in financial markets. In addition, fiscal deficits increase the supply of government bonds, thereby putting upward pressure on long-term interest rates. The stable macroeconomic environment is therefore tempered by elevated valuations and persistent geopolitical risks. This reinforces the need for diversification, selectivity, and disciplined portfolio construction.
Allocation: Unchanged across asset classes, adjustments to equity sectors
Our asset allocation remains unchanged. Macroeconomic conditions remain favorable. However, the market often underestimates risks and overestimates potential gains and opportunities. Consequently, geopolitical uncertainty, high fiscal deficits, and elevated valuations across several asset classes lead to a balanced overall positioning.
We are adjusting our sector allocation to benefit from AI infrastructure expansion. We see no compelling arguments for following short-term trends, as elevated valuations and expected sector and stock dispersion make it difficult to draw conclusions about future advantageous positions. We are overweighting the industrial sector to benefit from AI infrastructure expansion and other infrastructure investments. Consumer staples have the lowest earnings growth of all sectors and remain highly valued, which is why we are underweighting the sector.
The underweight position in bonds remains unchanged. Within the fixed-income segment, we remain underweight and favor investment-grade bonds, as the increased supply of government bonds and high fiscal deficits are dampening their yield potential.
We view gold and Swiss real estate as stabilizing elements in portfolios. Gold provides diversification, while Swiss real estate offers attractive CHF exposure and represents a compelling alternative to cash and bonds. Real estate is particularly attractive because structural scarcity will ensure stable returns over the long term. Therefore, by overweighting gold and real estate, we are focusing on stability and diversification.
Productivity: the impact of AI and the next technology cycle
Artificial intelligence (AI) has evolved from the experimental phase toward broad economic application and is on its way to becoming a key driver of productivity. Much of its potential lies in more efficient processes and the scaling of knowledge-intensive work. McKinsey estimates that generative AI could create up to $2.5–4.4 trillion in value added annually across industries. This supports the view that AI is a significant medium-term driver of productivity, not just a short-term market trend. From a macroeconomic perspective, this is relevant because higher productivity can support growth and boost corporate profitability through efficiency gains. For financial markets, this means that AI should be understood as a structural driver capable of transforming the supply side of the economy.
History shows that technological breakthroughs only boost growth after a prolonged phase of adoption and diffusion. Electrification, the personal computer, and new communication technologies all followed a similar pattern: an experimental phase was followed by widespread investment, adoption of the technology, and only then broader economic benefits. Electricity is often cited as one of the most striking examples: While its discovery dates back to the late 17th century, widespread commercial use did not occur until the late 19th century. The same logic applies to the development of the personal computer: The first devices appeared as early as the 1940s, while the economic benefits only became apparent after companies restructured their work methods, organizational structures, and business models accordingly. Between 1995 and 2005, information and communication technology contributed an estimated 0.3 to 0.8 percentage points to per capita GDP growth in OECD countries. AI will most likely follow a similar pattern, meaning that the technology can be economically relevant even if the positive effect on productivity becomes visible only gradually. However, the time span between introduction and productive use appears to be shortening: Since the launch of ChatGPT in late 2022, the technology has already been integrated into daily work routines at many companies.
Early indicators suggest that AI adoption is already yielding measurable productivity gains. In a randomized study of consultants, AI users completed 12.2% more tasks, finished them 25.1% faster, and produced over 40% more output. AI impacts both cognitively demanding tasks and routine work. For sectors such as consulting, software development, and healthcare, this holds the promise of significant efficiency gains, improved scalability, and lower marginal costs for knowledge production. With successful, widespread adoption, AI could increase the profitability of efficient users and support medium-term growth.
These gains are likely to be significant but unevenly distributed. Technological innovations typically generate adjustment costs before broad productivity gains become apparent. Labor markets, cost structures, and competitive positions also need time to adapt. This does not undermine the productivity thesis; rather, it suggests that the initial effects are more likely to manifest in relative performance across sectors and firms than in aggregated macroeconomic data. According to the results of an OECD survey, the productivity advantage of companies that use AI over those that do not ranges from 7.7% in France to 31% in Belgium, with the largest gains occurring among large companies. This difference is significant for investors. While AI may support aggregate growth over time, the initial effect is likely to vary across companies and sectors rather than manifesting as a uniform increase across all investments.
AI, capital intensity and market dispersion
For financial markets, the uneven distribution of Artificial Intelligence (AI) profits is likely to result in greater variation across different sectors and companies. Market leadership has been unusually concentrated recently, and the same applies to the expected value creation from AI. According to S&P Global’s AI Monitor, AI-related revenues are expected to rise to around $1.5 trillion by the end of 2027, up from $472 billion at the end of 2023. Growth is highly concentrated: the majority is accounted for by the ten largest companies in the AI Monitor, with Nvidia alone accounting for around 27% of the total volume. This supports the view that selectivity is more important than broad thematic exposure.
In equity markets, elevated valuations already leave little room for broad-based multiple expansion. The MSCI World forward P/E stands at 20.4x, above the 5- and 10-year averages. AI tends to reinforce a one-sided market structure rather than providing an equal boost to all companies: Leading technology stocks contributed 53% to the S&P 500’s return in 2025, while only 31% of small-cap AI stocks outperformed the Small-Cap Index. At the same time, expansion is extremely capital-intensive. AI therefore sharpens the distinction between technological leadership and investment attractiveness, as some companies benefit from demand for infrastructure or productivity gains, while others are primarily confronted with higher costs, stiffer competition, and delayed monetization.
Capital intensity is likely to be a defining feature of the early phase of the AI cycle. Every major technological wave goes through a phase in which spending rises before returns become apparent. During the internet boom, global investment in telecommunications networks increased fivefold between 1995 and 2000, peaking at $231 billion. A similar pattern is emerging with AI: its expansion is becoming increasingly physical and capital-intensive. Data centers, semiconductors, and the power requirements for model training all demand significant capital investment. Microsoft alone invested around $80 billion in AI-capable data centers in fiscal year 2025, while Alphabet invested $91.4 billion, a large portion of which went toward servers, data centers, and network infrastructure. The early gains from the AI hype were highly concentrated, as financial markets often reward the most visible pioneers in the early stages of such cycles. NVIDIA reported data center revenue of $62.3 billion for the quarter ending January 25, 2026, representing a 75% increase year-over-year and underscoring how strongly the current phase continues to benefit infrastructure providers. Over time, however, this dynamic may shift, as the early gains enjoyed by infrastructure providers are followed by later gains for companies that translate the technology into sustainable productivity, pricing power, and recurring revenue. This transition is one reason why market leadership can and will shift over the course of the cycle.
In summary, this implies for investment strategy that AI requires a stronger focus on quality and selectivity rather than broad exposure. When it comes to stocks, the key question is not simply which companies are associated with AI, but which companies can translate AI-related investments into sustainable economic returns. Companies with strong market positions, scalable business models, and solid balance sheets are likely to be better positioned than those primarily trying to chase the AI trend. This logic applies across industries, as AI is not only a technological development but also a catalyst for profitability and efficiency in manufacturing, healthcare, finance, and professional services. AI is therefore likely, like most technological innovations, to usher in an era of increasing selectivity. The most significant market effect here is not that companies will be displaced, but that the gap between winners and losers will widen.
AI amid geopolitical crises and macroeconomic pressure
The Artificial Intelligence (AI) buildout is taking place in a world where geopolitical conflicts and global cooperation continue to play a significant role. Since Trump’s first campaign, the global order has been shifting away from globalization and toward stronger national protectionism. The U.S.’s distancing from Europe is likely to lead to a multipolar world order in which international agreements are less important and the law of the jungle often prevails. If this leads to smoldering tensions in current or potential new conflicts—namely Iran, Ukraine, and Taiwan—it will result in macroeconomic shocks and global repercussions. Such shocks would not disprove the structural AI thesis, but they could make the development path more unpredictable, as the interplay between states and production locations in a globalized world is often underestimated.
The current conflict in the Middle East is relevant because the deployment of AI depends on stable energy, infrastructure, and financing conditions. Widespread use of the technology requires data centers, semiconductors, a reliable power supply, and cooling capacity. The International Energy Agency expects global electricity consumption by data centers to double to around 945 terawatt-hours by 2030, while that of AI-optimized data centers will more than quadruple. In the US, data centers are expected to account for nearly half of the additional electricity demand during the same period. AI is therefore not just a software technology, but also a matter of physical infrastructure, energy availability, and financing. This makes AI more vulnerable to geopolitical shocks. Another oil shock would be significant because it could raise long-term interest rates and the cost of capital for the AI transformation through increased inflationary pressure. Furthermore, wholesale electricity prices in areas near large data centers have recently been more than three times higher than five years ago, which can increase cost pressures for consumers.
Fiscal policy exacerbates this risk by increasing pressure on long-term interest rates. Persistent deficits increase the supply of government bonds and force markets to bear greater duration risk. According to estimates by the Congressional Budget Office, U.S. gross debt could rise to nearly 200% of GDP by mid-century. As central bank support wanes in the form of reduced bond purchases and less expansionary monetary policy, this could exert additional upward pressure on long-term interest rates. When long-term interest rates rise, capital-intensive investments become more costly, leading to a more selective choice of projects. This is particularly relevant for AI, as the strong narratives in the current early stage are associated with significant upfront investments.
AI remains the most significant structural growth opportunity, while balance, diversification, and resilience continue to be central components of our portfolio construction. The path from innovation to societal benefit is unlikely to be smooth: geopolitical shocks could impact inflation and energy prices, higher interest rates could push valuations lower, and uneven adoption will widen the gap between companies that benefit from it and those that merely bear the costs. For portfolios, however, this argues against treating AI as a momentum trade. We favor a diversified approach based on quality, resilience, and long-term prospects.
Authors:
Lars Fluri
Tobias Wagner
Samuel Nibal
Disclaimer
The prices used in our analysis are end-of-period prices. The figures used for our valuation model are estimates referring to dates and therefore carry a risk. These are liable to change without notice. The usage of valuation models does not rule out the risk that fair valuations over a specific investment period cannot be attained. A complex multitude of factors influences price developments. Unforeseeable changes could, for instance, arise from technological innovations, general economic activities, exchange-rate fluctuations or changes in social values. This discussion of valuation methods makes no claim to be complete. Dreyfus Sons & Co Ltd, Banquiers publishes Compass four times a year since June 2008. The publication is aimed at clients of the bank and interested parties. It describes some of the instruments and methods the bank uses to monitor everything to do with the financial markets. A description of the investment process can be obtained from your client advisers or our website. Compass provides guidance but cannot take the circumstances of an individual portfolio into account. It is for information and marketing purposes.