As part of the Natixis CIB Research 2026 Outlook, Eric Benoist – Tech and Data Expert, Christopher Hodge – Chief US Economist, Alain Durré – Chief EMEA Economist, Gary Ng – Sectoral Research Asia, and Florent Pochon – Head of Cross Asset Research, took a deep dive into Artificial Intelligence, one of the most powerful forces shaping markets, corporate strategy, and global economic dynamics today.
Eric
Benoist
Christopher Hodge
Alain
Durré
Gary
Ng
Florent
Pochon
Their combined insights highlight both the scale of the AI transition and the complexities that lie ahead.
AI Adoption: Between Enthusiasm and Fatigue
Despite extraordinary momentum, evidence of “AI fatigue” has begun to surface. Recent studies show that a large majority (~95% according to MIT) of companies experimenting with AI have yet to extract meaningful value from their investments. According to data from the US Census Bureau, adoption rates, particularly among larger firms, have softened after reaching a peak earlier this year.
At the same time, some of the latest model deployments (notably ChatGPT 5.0) have not met market expectations. Incremental improvements have left unresolved challenges such as hallucinations, logic inconsistencies and persistent bias. A growing body of research also questions the assumption that scaling models indefinitely leads to increased accuracy, indicating potential structural limits to current generative architectures.
AI is moving from hype to a more realistic phase. Many companies are experimenting, but only a small share have started to extract real value. This doesn’t diminish the long term potential of AI though; it simply highlights where we are on the adoption curve.
Eric Benoist
In market terms, AI leaders continue to command elevated valuations, but levels remain far below previous technology bubbles. Recent performance has been supported by exceptionally strong earnings growth over the past decade, particularly among the largest global technology platforms.
An Intensifying Race for Infrastructure
One defining feature of the current cycle is the massive global investment in data infrastructure. Across regions, hyperscalers are deploying unprecedented levels of capital to expand data-center capacity, cloud architecture, and computing power.
In the U.S., data-center development remains dominant, with close to 5,000 facilities and further acceleration expected. Europe hosts roughly 1,000 facilities, while China now represents about one quarter of global data-center capacity. China’s cloud and data-center industry has also expanded rapidly, contributing around 1% of national GDP in 2024.
Looking ahead to 2025 and 2026, the major global players are planning combined capital expenditures exceeding half a trillion dollars. This intense infrastructure build-out is reshaping energy markets, labor demand, and corporate balance sheets, with rising leverage emerging as a point to monitor.
Productivity: The Central Economic Question
The long-term value of AI will ultimately depend on its impact on productivity. AI is increasingly becoming embedded into workplace tools (e.g. AI assistants) improving workflow efficiency, reducing repetitive work and allowing teams to focus on higher-value tasks.
While it is still too early for definitive measurement of productivity, two analytical approaches offer perspective.
Historical comparison with previous technological waves suggests potential annual gains of 0.3 to 0.6 percentage points in advanced economies, based on past digitalization cycles.
Theoretical modelling, however, yields a wide range of estimates. Depending on assumptions regarding task exposure, profitability of task replacement, and efficiency gains, long-term effects could be negligible – or could reach up to 0.7 percentage points per year.
The divergence illustrates the high uncertainty still surrounding AI’s macroeconomic contribution.
Alain Durré
The productivity impact of AI is still uncertain as we lack sufficient data. History shows us that digital innovations can generate meaningful productivity gains, but the range of possible outcomes for AI remains very wide. The next decade will depend on how broadly tasks can be automated and how effectively companies can implement change.
The United States: AI as a Growth Engine
The US economy has demonstrated ongoing resilience, supported in part by AI-related capital expenditure.
Leading tech companies are allocating close to 60% of operating cash flow to capex, and new fiscal measures taking effect at the start of next year (part of the One Big Beautiful Bill Act – which comes into effect on Jan 1, 2026), are expected to further encourage investment.
AI-related investment has become a key pillar of US economic strength. Capital expenditure by major tech firms is unprecedented, and new tax incentives will serve to accelerate this trend. The labor market effects are complex, but overall, it remains a powerful driver of growth.
Christopher Hodge
In the labor market, early signs are mixed. While automation expectations have weighed on hiring for certain graduate profiles, the surge in data-center construction has created strong demand in other sectors, such as construction employment – offsetting some of the impact. Short-term effects appear inflationary, but the longer-term employment dynamics remain unclear.
China: Seeking Scale Through Efficiency and Application
China’s role in the global AI landscape continues to expand. Despite constraints in advanced semiconductor supply, domestic efforts are intensifying across three fronts:
- Increasing computing capacity
- Developing more cost-effective models
- Deploying rich, application-driven use cases across consumer and industrial ecosystems
China benefits from one of the world’s deepest datasets for large-scale digital applications, offering fertile ground for AI agents and platform-based growth. The primary challenge remains access to cutting-edge chips, in contrast to the U.S., where electricity supply has become a limiting factor.
Gary Ng
China remains one of the world’s strongest competitors in the AI race. Despite semiconductor constraints, the country is rapidly expanding computing capacity and focusing on efficient models and large-scale applications. The real advantage relies on China’s ability to deploy AI at scale across its digital ecosystem.
Although it is not possible yet to conclude that China has overtaken the US in the global AI race, it is clearly among the leading contenders to do so. For global organizations, China’s progress serves both as a roadmap and a reminder of the importance of speed and ecosystem collaboration.
Cross-Asset Perspective: High Valuations, Rising Concentration
Market valuations across leading AI platforms remain elevated but broadly consistent with underlying earnings performance. Current price-to-earnings levels are well below those seen during the dot-com era or the Japanese equity bubble of the 1980s.
The more significant structural risk lies in concentration. The largest U.S. technology firms now account for roughly 35% of the S&P 500, creating heightened sensitivity to any slowdown in earnings, free cash flow generation, or economic growth.
What matters most is not so much an AI bubble, but the extreme concentration of market performance in a handful of tech companies. Valuations are elevated but remain far below historical bubble levels. The real risk is that any slowdown in earnings or cash flow among the largest players could have significant consequences for global markets.
Florent Pochon
Given this backdrop, balanced portfolio construction becomes essential. Diversification beyond U.S. technology – including Europe, Japan, and sector-specific opportunities – offers a way to mitigate concentration risk. In multi-asset portfolios, maintaining high liquidity and combining trend-following strategies with alternatives and low-volatility equity exposures can help stabilize returns in a more uncertain environment.
The Next Chapter: 2026 and Beyond
AI remains a transformative force and will continue to play a central role in markets and across economic activity through 2026. The technology remains a core theme for both productivity and corporate investment. However, the risk landscape is increasingly shaped by concentration in technology and elevated capex requirements.
As with previous technological revolutions, the current environment resembles a moment where the technology is widely visible, but not yet fully reflected in productivity statistics. Historically, such phases have often proceeded meaningful improvements. Whether this holds true for AI, will be a critical question for the years to come.
