Notes from my inbox

CYCLES TURN

In the January edition of this note, we observed that gold had become the single most crowded trade among global fund managers, and that the market was pricing the world’s gold stock as more valuable than the seven largest US technology companies, the London Stock Exchange and all US farmland combined. We added an uncomfortable historical observation: after every previous gold rally, investors in the miners suffered significant losses when prices declined.

It took six months for the cycle to reassert itself. From a peak around US$5,600 an ounce in January, the gold price has fallen back by roughly a quarter towards US$4,000 in late July 2026. The mining shares, as ever, amplified the move: several of the large precious metals counters on the JSE lost between a third and nearly half of their value this year. The index-level concentration that resulted from gold mania means that the overall local market is down 10% since the gold price peaked, for a sobering sequel to 2025’s exceptional returns. Beneath the headline, however, the picture was far healthier: domestically focused shares, largely ignored during the gold rush, were, in aggregate, also unaffected by the correction.

We would caution against reading this outcome as evidence of any ability to time the turn in sentiment – we don’t think we have any special ability to do so, nor did we try. The point is simpler: commodity prices are cyclical, crowded trades eventually disappoint, and the price you pay for an asset remains the most reliable predictor of the return you will earn from it.

THE NEW OBJECT OF AFFECTION

In today’s narrow, narrative-driven markets, reflexive valuation overshoots are more regular occurrences. Leveraged retail capital seems less likely to retire to cash; rather, it rotates to the next theme. As the gold trade began to unwind, the crowd found a new home. The second quarter of 2026 was the strongest for global equities since 2020, with the world index up almost 15% and emerging markets up 24%, powered overwhelmingly by a single theme: the build-out of artificial intelligence infrastructure. The world’s hyperscalers[1] raised their combined capital expenditure guidance for 2026 towards US$700 billion. The main semiconductor index[2] rose nearly 70% in two months – its strongest quarter on record – with the world’s chipmakers collectively adding some US$3 trillion in market value over a few weeks. Korean equities, home to two of the largest memory-chip makers, produced their best quarter since 1998.

These are extraordinary numbers, and they have been accompanied by the familiar chorus. Strategists debate whether valuations now resemble those of early 2000; one widely followed bubble indicator sits a whisker below its trigger level; and the first sharp wobble arrived in early July, when chip shares sold off on little more than the fear that spending growth might merely slow. When markets price perfection, disappointment does not require bad news – slightly-less-good news will do.

REAL TECHNOLOGIES, UNREAL PRICES

None of this means artificial intelligence is a mirage. We agree that the technology is transformative and will lead to profound changes in the structure of the economy and financial markets. Like every great speculative episode in history (railways, electricity, the internet), investor excitement is clearly based on something real. In most cases, the technology went on to change the world exactly as promised, yet still destroyed the capital of those who overpaid for participation. The investors who bought the world’s dominant networking company at the peak of the dot-com boom owned a business that kept growing for decades, yet they waited the better part of twenty years just to recover their starting capital. The question that matters is not whether the technology is important. It is what price you are paying, what expectations that price embeds, and how much margin for error remains if the future arrives slightly later, or slightly differently, than consensus expects.

AI-related earnings are, for now, real and growing rapidly − this is not 1999 redux in every respect. But the gap between the winners and the losers will be very wide, while the recent indiscriminate rally lifted both businesses with durable competitive advantages and businesses with none at all. We believe that this is the environment in which valuation discipline earns its keep.

PATIENCE, THE PRICE OF SUPERIOR LONG-TERM OUTCOMES

Regular readers will recognise the pattern that connects 2025’s gold mania to 2026’s silicon one. Markets overshoot in both directions because human beings extrapolate: we are prone to mistake the recent past for the permanent future. The cycle has not been repealed; it merely marches to a different drum. Our job is not to forecast when sentiment turns − we don’t think anyone can do that reliably − but to ensure that the portfolios we manage on your behalf never depend on the crowd remaining enthusiastic. We aim to own businesses we understand, bought at prices below our assessment of their long-term worth, held in portfolios diversified across enough independent ideas that no single narrative can dictate your long-term outcome. At times this discipline will look out of step for a quarter or two. As the first half of this year again demonstrated, that is the price of admission for compounding capital safely across decades.

For detailed insight into how your funds are positioned in this environment, I encourage you to read this quarter’s portfolio manager commentaries.

Thank you, as always, for the trust you place in us to manage your long-term capital.

[1]The very large cloud-computing companies (Amazon, Microsoft, Alphabet, Meta and their peers) that are building the data centres on which artificial intelligence models run.

[2]The Philadelphia Semiconductor Index, the most widely followed benchmark of the world's major chipmakers.

Insights Disclaimer