Why Investors Must Shift Focus from AI Infrastructure to Tangible Value Creation
Every summer, an unsettling pattern appears in the data. As temperatures rise, so do drowning deaths. The relationship is consistent, measurable and, on the surface, compelling enough to warrant action.
Does this mean hot weather causes people to drown?
No. Both are driven by a third variable hiding in plain sight: when it is hot, more people go swimming. More swimmers means more drowning. The statistical relationship is real. The causal inference is wrong. This distinction between correlation and causation might seem like something best left for a statistics course. But in business and investing, mistaking the two can cost dearly – something that investors currently enamoured by artificial intelligence need to keep front-of-mind.
In 2011, the board of U.S. department store chain JC Penney made what looked like a brilliant CEO hire. Ron Johnson had just overseen Apple’s retail transformation into a chain generating more revenue per square foot than virtually any retailer on earth. The Apple Store was a cultural phenomenon, and Mr. Johnson was its architect.
The pattern seemed obvious: Apple Retail delivered spectacular results; Mr. Johnson ran Apple Retail; therefore, hire Mr. Johnson to deliver similarly outstanding results. The logic held together if you ignored the third variable.
Mr. Johnson replicated his Apple playbook at JC Penney. He eliminated hundreds of promotional events, introduced boutique shop-in-shop concepts and redesigned the store experience to feel more like a curated mall than a discount chain.
What he failed to grasp was that the Apple Store succeeded not merely because of its brand or minimalist design, but because customers genuinely coveted its products. Remove that variable – Apple’s desire-driven product ecosystem – and the formula collapses. JC Penney’s customers wanted deals, not a curated browsing experience.
In less than two years, the company lost more than US$4-billion in revenue, and Mr. Johnson was fired 17 months after his celebrated arrival. The correlation was real; the implied causation – that the format alone drove success – was not.
Nassim Taleb calls this the narrative fallacy – the deep human tendency to construct causal stories from correlations and then mistake those stories for insight. In investing, this pattern repeats with remarkable consistency.
Fitness equipment maker Peloton is one of the clearest recent examples. Its pandemic-era surge was powered by a story that lockdown demand reflected a lasting reset in fitness habits, but much of the buying was driven by constrained circumstances rather than permanent behavioural change. When those constraints disappeared, demand faded and the stock collapsed by more than 97 per cent from its peak.
Beyond Meat followed a similar arc. Its 2019 IPO was driven by a compelling story about a permanent move away from animal protein, and investors priced the company as if that shift were already under way. In reality, early sales were driven by novelty and curiosity, not durable conversion. Once the novelty wore off, demand softened and the stock crashed by more than 90 per cent from its highs.
Now consider artificial intelligence. The technology is real. The capabilities are truly remarkable. What is worth examining is whether the story driving AI valuations reflects a proven causal relationship between the technology and future economic returns, or whether it more closely resembles Peloton and Beyond Meat.
The financial data gives pause. OpenAI’s revenue is estimated to have reached an annualized run rate of over US$20-billion in 2025. Despite this rapid growth in revenue, its costs are growing even faster, meaning losses (estimated to be US$14-billion for 2026) are continuing to increase rather than narrow. Cumulative losses are projected at roughly US$44-billion from 2023 through 2028. The four largest cloud providers are on track to spend around US$635-billion to US$665-billion on AI-related capex in 2026 alone.
Microsoft’s Copilot, designed to monetize AI across more than 450 million commercial Microsoft 365 licences, had converted only about 3.3 per cent of that base to paid subscriptions by early 2026. A 2025 S&P Global survey found that 42 per cent of organizations had scrapped most of their AI initiatives, with AI proof-of-concepts frequently killed before going live. Analyses from RAND and others put enterprise AI implementation failure rates in the 70- to 85-per-cent range, far above typical IT projects. Goldman Sachs still finds no meaningful relationship between AI adoption and economy-wide productivity, despite sizable gains in targeted use cases. Deutsche Bank points out that the market is valuing AI companies at levels rarely seen in history despite their large losses and uncertain path to profitability. The spending is real. The associated returns, at scale, have not yet arrived.
So how can investors prevent the third variable from messing up their portfolio? The first way is to identify the third variable before acting on the pattern. For AI stocks, that answer includes institutional momentum, narrative quality and FOMO on a perceived generational opportunity. These are real forces. They are not the same as demonstrated paths to profitability.
The second is to actively seek disconfirming evidence. Every compelling narrative has a version that ends differently. The AI version requires asking whether the gap between infrastructure spending and enterprise returns is a timing problem or a structural one – and being honest that current data does not yet answer that question.
The third is to distinguish durable causation from market sentiment. As Nobel economics laureate Robert Shiller has argued, contagious narratives can drive market prices to levels divorced from the underlying fundamentals, especially when the story itself becomes the main driver of valuation.
Renowned investor Benjamin Graham observed that in the short run, the market is a voting machine, and in the long run, it is a weighing machine. Narratives drive the votes. Fundamentals determine the weight. The investors most affected when the weighing begins are those who mistook correlation for causation.
Sam Sivarajan is a speaker, independent consultant and author of three books on investing and decision-making. He writes two free Substack publications on decision-making and the good life: The Uncertainty E.D.G.E. and The Good Human Practice.
This article was first reported by The Globe and Mail
