Technology — the AI capex cycle is the new earnings story.

Technology is the sector where capital allocation matters more than top-line growth. The hyperscalers — Microsoft, Alphabet, Amazon, Meta — are running combined annualized capex in excess of $300B for AI infrastructure, and the names that compound from here are the ones whose returns on that spend show up in the operating margin within 24 months.

Inside that frame, the desk pays attention to three sub-pockets. First, the megacap platforms — these are the customers, and their capex commentary on every print sets the demand curve for everyone downstream. Second, the semiconductor stack — NVIDIA at the top of the value chain, AMD and Broadcom adjacent, with TSMC, ASML, and Applied Materials providing the picks-and-shovels. Third, the AI-native software companies — Palantir, ServiceNow, Snowflake, and a long tail of vertical incumbents whose pricing power is being tested by GenAI.

Software businesses with sticky distribution are revaluing higher because the marginal cost of every new AI feature is the inference bill, not engineering hours. That puts a premium on companies with deep customer datasets they can fine-tune against — and a discount on those whose moats were UI polish.

The risks the desk watches: (a) AI capex digestion — at some point the hyperscalers slow the buildout and the entire semis chain takes a multiple compression hit, (b) regulatory, particularly export controls on advanced semis to China and the antitrust posture toward the megacap platforms, (c) the scarcity of next-generation power supply, which has gone from 'cheap' to 'gating'. Anyone underwriting the sector without a power-availability view is mispricing it.

Editorial coverage cadence: weekly market-structure piece, monthly capex tracker, earnings-week deep dive on each megacap. Subscribe to the markets newsletter for the morning brief.