What a Thesis Break Actually Looks Like

Research desks are reluctant to write about their reversals. The incentive is to let a retired view fade quietly and keep publishing new ones. We think that habit removes the most instructive material a research archive contains.

What the original thesis claimed

The position rested on three claims: that pricing power in the segment was durable, that the capital intensity of expansion would fall as the network matured, and that the competitive response would be slow because incumbents were structurally constrained.

Each claim came with a stated invalidation. That is what made this identifiable as a break rather than a period of underperformance.

Which claim broke

The second. Capital intensity did not fall as the network matured — it held roughly flat, then rose. Two quarters of that could be timing. Four quarters, with management commentary shifting from "temporary" to "the new base", is a thesis break.

The first and third claims held up. That is worth stating plainly, because a partial break is more common than a total one and considerably harder to act on.

What we got wrong in the process

The original model assumed a maintenance capex ratio derived from a comparable business in an adjacent market. That comparable turned out to operate under different regulatory requirements which materially changed its capital cycle. We should have caught that at the modelling stage; it was in the filings.

What changes as a result

• The screening checklist now flags cross-market comparables for regulatory differences explicitly.
• Capital intensity assumptions require two independent sources rather than one comparable.
• The original note stays online, with the reversal linked from it.

We publish this in the same format as an initiation, with the same prominence. A research archive that only contains views that worked is a marketing document.