Clyde Rossouw, Head of Quality, Ninety One 
On the other side is the less glamorous reality of the underlying economy. Aggregate GDP figures and index-level performance appear supportive, but beneath the surface, they mask a weakening consumer environment. Resilient spending by the wealthiest households and heavy investment by AI hyperscalers contrast with the growing financial strain on middle- and lower-income consumers. This imbalance leaves markets resting on a fragile equilibrium.
Traditional defensive areas, such as consumer staples and healthcare, are no longer seen as being able to provide the ballast typically expected during periods of economic stress or elevated valuation dispersion. Their weaker pricing power and challenged end-markets have pushed them out of favour with investors. Meanwhile, the very companies whose fortunes depend on large, synchronised capex cycles are viewed as ‘safe bets’ simply because their near-term growth outlook appears unstoppable.
Cyclicals as the new defensive
The current narrative is based on a fundamental contradiction. Companies powering the AI boom, such as semiconductor producers, networking suppliers, and power-infrastructure providers, trade at valuations that imply utility-like stability. The belief is that generative AI represents such a profound secular shift that it overrides normal cyclical dynamics. The demand for advanced computing remains structurally above supply, and hyperscalers must invest aggressively to maintain their leadership, regardless of macroeconomic conditions.
But how durable is this narrative? History provides perspective.
In the late 1990s and early 2000s, telecom carriers invested vast sums in fibre-optic networks, expecting unending bandwidth demand. Equipment manufacturers saw explosive earnings and soaring share prices until overcapacity, falling valuations and a pullback in spending triggered a rapid and severe correction. Order books that once looked unshakeable vanished almost instantly.
While AI is unquestionably transformative, arguably more so than early internet technologies, the scale and speed of current spending may have brought forward multiple years of future demand. Periods of digestion, inventory adjustments and pauses in deployment are inevitable. Companies dependent on concentrated capex from a handful of large customers are, by definition, cyclical. Unlike fibre-optic cables, which last decades, GPUs have a short operational life, amplifying the impact of overinvestment. Yet markets continue to treat the most momentum-driven parts of the AI ecosystem as defensive.
Investors appear to be seeking safety in growth, convinced, somehow, that AI can overpower normal business cycles.
The economic shadow of a two-tiered consumer
A second factor behind the market’s current imbalance is the increasingly stressed consumer landscape. Headline retail figures mask a widening divide. The highest-income households, who hold the bulk of asset wealth, have benefitted from strong housing and equity markets, with spending on luxury goods, travel and premium services remaining robust. In many cases, the wealth effect created by AI-driven equity gains is reinforcing this strength.
In contrast, middle- and lower-income households are facing intensifying pressure. Inflation in essential categories, such as housing, food, and energy, combined with the depletion of pandemic-era savings and support measures, has left budgets strained.
Rising delinquencies in credit cards and auto loans, along with widespread trading down at retailers, confirm the strain. Although consumer staples should theoretically fare well in such periods, their weaker growth and contracting multiples leave them out of favour in a momentum-driven market that prizes AI-related narratives.
AI losers as defence in disguise?
Ironically, the most compelling defensive opportunities may sit among the market’s perceived ‘AI losers.’ Many high-quality businesses in software, IT services and information services have been indiscriminately de-rated. Investors worry that AI will automate their workforces, replace their products or erode their data advantages. Just a few years ago, these same companies were expected to be the biggest beneficiaries of AI adoption.
Many now trade at or below market valuations despite robust recurring revenues, entrenched customer relationships and highly scalable business models. Consider professional information services providers. They own proprietary datasets, protected by regulation, contracts or high verification costs, that are essential for clients. The idea that open AI models can simply replicate this data overlooks the licensing, legal and maintenance requirements that underpin these businesses. In reality, data owners may become toll-collectors as AI usage grows.
Similarly, specialised vertical software providers have seen valuation compression despite offering mission-critical tools embedded deeply within customer workflows. In a downturn, CFOs are far more likely to pause optional AI projects than to remove core systems such as payroll, compliance or industry-specific software. Many of these businesses exhibit the very characteristics, high switching costs, strong retention, recurring cash flows, that historically define defensive investing.
Yet current sentiment prices them as though they are in structural decline. While some assets will unquestionably be disrupted by AI, the market’s broad-brush approach appears overly corrective. These companies may present some of the most attractive opportunities when the current cycle turns.
Conclusion: Navigating the market’s cognitive dissonance
Markets are wrestling with a profound cognitive dissonance. Investors recognise signs of a consumer recession, rising delinquencies, trading down, and growing financial stress, yet continue to award defensive valuations to the most cyclical parts of the AI supply chain. At the same time, they shy away from companies with the most predictable, recurring revenue profiles in corporate history.
The greatest risk lies in the crowded assumption that AI-driven capex will continue uninterrupted. When the inevitable slowdown arrives, markets will quickly reassess the cyclicality of today’s perceived ‘safe bets’, and elevated multiples may offer little protection.
True defensiveness does not come from chasing momentum into cyclical extremes, but from identifying high-quality, recession-resistant businesses whose valuations have been reduced simply because they lack an AI tailwind.
History suggests that when cyclicals are priced as defensives and defensives as cyclicals, a major rotation is rarely far away. For patient investors, the most resilient opportunities may lie where the market currently sees only risk. DM
Author: Clyde Rossouw, Head of Quality, Ninety One