Duane Cable, Head of SA Quality, Ninety One. A story that should sound familiar
In March 2000, Julian Robertson, founder of Tiger Management and one of the most celebrated investors of his generation, wrote a letter to his limited partners and returned their capital. He had spent two years underperforming badly. His style of buying the best businesses and shorting the worst had stopped working in a market where earnings and fundamentals had been replaced by mouse clicks and momentum.
“In an irrational market,” he wrote, “such logic does not count for much.”
The Nasdaq peaked ten days later and lost 80% of its value over the following two years. Robertson was right. The clients who redeemed at the bottom never saw the reversal they had waited years for.
This pattern has surfaced before. Every significant market dislocation in modern financial history has been anchored to a compelling story. Railroads genuinely transformed economies. Japanese industry was indeed dominant. The internet changed the world. None of that prevented those markets from becoming deeply overvalued. Euphoria does not attach itself to bad ideas. It attaches itself to very good ones, taken too far.
The AI story is real. The question is the pricing.
The AI story today is similarly compelling. We believe the technology is indisputably transformative. But Jeremy Grantham, one of the great students of market cycles, makes a point worth considering: every significant technological breakthrough tends to follow the same pattern. Genuine innovation, massive capital inflows, overcapacity, and then a repricing. The technology survives. Investors who paid for perfection often do not.
Look at the data on market concentration across history. Railroads at 63% of US market cap. Japan at 44% of global equity. The dot-com TMT trade at 41%. Today, the AI Big 10 sits at 40% of US market cap and climbing. The numbers do not repeat exactly. But they rhyme with a consistency that is worth paying attention to.
Then consider the IPO market. SpaceX debuted on Nasdaq on 12 June at a valuation of roughly $1.8 trillion. Within three trading days it had surged nearly 50%, briefly touching $2.8 trillion and adding close to $1 trillion in market value, one of the most remarkable stock market debuts in history. For context, Amazon carries a market cap of around $2.2 trillion on $750 billion in annual revenue. SpaceX reported $18.7 billion in revenue in 2025 and a net loss of $4.9 billion. OpenAI and Anthropic are lining up behind it.
A glimpse of the exit
In early June, Broadcom reported strong earnings, with AI revenue up 143% year on year to $10.8 billion, yet the stock fell 15%. Not because the business disappointed, but because it merely reiterated rather than raised its full-year guidance. Good was not good enough.
The reaction offered a useful illustration of how sentiment can shift. Shares in AMD, one of Broadcom’s closest rivals in the AI chip space, fell nearly 11% on the same day despite reporting no news of its own. The Nasdaq had its worst day since April 2025. Over $1.3 trillion in market value was erased in a single session across the chip sector. A strong US payrolls number added to the pressure, raising the prospect of rates staying higher for longer, a particular concern for businesses carrying significant debt.
Markets have since partially recovered. But for a moment, investors got a clear sense of how crowded these trades have become and how quickly the door narrows when sentiment shifts.
The AI infrastructure being built is real and the demand underpinning it is genuine. But the behavioural patterns surrounding it are ones history has punished before: markets narrowing to a handful of names, valuations that price in perfection, and insiders rushing to sell while conditions allow.
Meanwhile, the market has broadly written off software, IT services, and information businesses as AI losers, assuming generative AI will erode their value. We think this misreads both the technology and the businesses we hold.
These companies own proprietary data that cannot be replicated by an open model. They have entrenched customer relationships and mission-critical software embedded so deeply in client operations that a CFO under budget pressure will cut discretionary AI projects long before touching core systems. Their recurring revenues are strong, their balance sheets clean, and their starting valuations, after years of de-rating, are as attractive as we have seen them in some time.
Our Global Franchise strategy has an active share of approximately 90%. In practical terms, it means that the majority of our portfolio, by position and by weight, looks nothing like the index. When sentiment eventually rotates, corrections tend not to fall evenly. They fall hardest on the most crowded trades.
The hyperscalers we hold are exceptional businesses. But even they are showing signs of pressure. Companies that once generated surplus free cash flow are now funding AI ambitions partly through debt issuance. That shift in capital allocation is meaningful, and it is one we watch carefully.
Nobody knows exactly when the music stops. What history does suggest is that concentration at these levels, and the behaviours that accompany it, has rarely persisted indefinitely. The opportunity set for quality investing is better today than it has been in years, in part because so much attention and capital is focused elsewhere.
We have been here before. We know how it tends to end. We intend to still be standing when it does. DM
Author: Duane Cable, Head of SA Quality, Ninety One