Markets are being carried by momentum, and nowhere more than in the growthier corners tied to artificial intelligence. That momentum is not baseless, but it now sits alongside very high expectations, and the combination is worth thinking through carefully. ​Quite a few managers have put to us equally convincing narratives around why this both is and isn’t a typical chip cycle or a bubble. At this stage this is unknowable but at some stage it will matter to investors who have either been left behind or suffered a setback and we think it merits discussion and scenario analysis.

In this week’s What We Are Working On video we discuss this with Qiao Ma of Munro Partners, Munro's case for the AI complex is a first-principles one and it is deliberately not the usual "chips cycle" story. For decades, global compute demand grew at roughly 10–15% a year while semiconductors improved performance faster, producing the familiar boom-and-bust of over- and under-supply. In 2026, the manager argues, total compute demand is growing closer to 100% a year. An industry built for 15–20% growth is now short of almost everything — processors, memory, power, and the unglamorous components in between — and Munro's holdings are the picks and shovels selling into that shortage.

Two points support the view that this is more than a sentiment trade. First, the AI theme has fractured into a dozen or so sub-themes, high-performance computing, connectivity, power equipment, specialised contractors, that no longer move as one; through 2026 they have traded on their own earnings cadence rather than in lockstep. Second, and more reassuringly, the rally has been driven by earnings, not re-rating. Nvidia trades on a lower multiple today than in 2023 and several peers are flat. The share prices have followed the profits.

That is a coherent bull case. The scenario question is what happens if it is wrong, or merely early. A great deal of consensus growth is already embedded in current prices. In a mean-reverting world, or a recession, high expectations are precisely what gets punished. Multiples that look reasonable against 20%-plus growth compress quickly once that growth is questioned, and sentiment in the largest names can turn well before the fundamentals do. The risk is not that the companies are bad; it is that the price already assumes the good news arrives on schedule.

This is where the character of the manager matters more than the thesis. The most useful thing about Munro in this environment is not that they are bullish, plenty are, but that they hold high conviction without becoming wedded to it. Positions that fall 20% are reviewed in front of the whole team and kept only by unanimous agreement; winners are trimmed or sold at price targets; whole areas, digital enterprise among them, have been exited when the facts changed. The diversification across uncorrelated themes is deliberate, not accidental.

In this discussion Qiao frames their own edge as imagination rather than information, seeing where the dots connect next, the part AI cannot yet do. Interestingly, this exactly mirrors some very recent articles by software engineers who see opportunities in what they might think to ask the most advanced Large Language Models to do on their behalf (given their deep knowledge and broad experience), whereas until very recently their workplace edge had been very specific and technical. For investors, the near-term practical translation is about discipline: a willingness to act when the story changes rather than defend a past decision.

None of this resolves the scenario question, and it is not meant to. With a wide range of outcomes ahead, the combination worth owning is conviction paired with the discipline to change one's mind, a fair description of how Munro invests. A useful near-term tell, the manager suggests, will be hyperscaler cloud margins this earnings season: if the returns on close to a trillion dollars of capex show up in margins, the foundation is sound.

‍

This article is general commentary only and does not constitute personal financial advice. Portfolio holdings are referenced for illustration and are subject to change.