The week’s reports about AI revenue and financing made the customer side harder to ignore. The portfolio question is not just whether a supplier can make a desirable product. It is whether the buyers can keep paying for the spending schedule on which the supplier’s growth depends.
The published holdings span semiconductors, cloud infrastructure and enabling equipment. I can describe AMD, AVGO, MU, VRT, CRWV and NBIS as separate businesses and still find several of them exposed to a common funding cycle.
I do not want the sector labels to create a false sense of independence. A delay in a major customer’s expansion can travel backward into hardware orders and forward into utilization assumptions. The exposure may appear in different quarters and under different accounting lines, which makes it easier to miss if the review stops at the current revenue mix.
I would make the dependency map explicit before claiming that the book is diversified against a spending slowdown. That is a research task, not a hypothetical trade recommendation. The written timing and holding rules still govern any actual decision.
No historical daily account return is reconstructed in this letter, and no undocumented trade is asserted. The current holdings file can identify exposures to investigate; it cannot tell me every position size that existed during the gap. The value of this retrospective note is the question it preserves: how many of my apparently distinct growth assumptions require the same buyer to keep finding money?
I would start the review at the customer, not the sector heading in the holdings sheet. If a cloud operator finances new capacity, an equipment supplier delivers into that capacity and a platform expects to use it, the three businesses can expose me to one underlying expansion plan at different points in time.
That does not make the businesses interchangeable. A supplier can collect before an operator has demonstrated utilization. The operator can be paid under a contract while the end user is still testing whether the workload produces an acceptable return. The order in which each business receives cash can determine which part of the chain remains resilient if expansion slows.
For AMD, AVGO and MU, I would ask how much of the growth case depends on customers maintaining their investment plans. For VRT, I would connect equipment demand to the projects requiring it. For CRWV and NBIS, I would keep the funding and utilization questions beside the revenue forecast rather than relegating them to a risk paragraph at the end.
I would then write down what would actually contradict the concern. A customer showing durable collections, useful capacity and a manageable financing schedule would be different evidence from another announcement of ambitious spending. I do not want a dependency map that only becomes more alarming regardless of what companies report. It should be capable of becoming more reassuring, too.
The policy rules still matter here. A new research concern is not permission to rewrite the Blackout Rule or Holding Period policy around an uncomfortable headline. My job in this letter is to make the common assumption visible before it becomes a surprise. The holdings list tells me the names I own; the dependency map tells me which parts of their stories need to be true together.
Aydin Ali
AEA Capital Research