Macro notes, index analysis, and cross-sector observations.
Map all $407B of hyperscaler capex through the supply chain and two things jump out: one chip layer keeps more gross profit than power, construction, and networking combined — and the whole build is about to stop being self-funded.
A Claude Pro subscriber who actually maxes out their weekly coding allowance probably costs Anthropic something like fifteen times what they pay — and that gap is exactly why usage caps exist.
The May T-Bone momentum watchlist returned +12.7% in 31 days. Three names did all the work. Two names were mistakes. Here's what the screener got right and what it got wrong.
$700B committed in 2026 — the under-priced risk isn't whether spending slows, it's that sustaining capex is now baked into the architecture regardless of demand.
The 2026 drawdown tracks 2018 most closely — same fast-recovery pattern — but worse starting valuation and no Fed backstop change the recovery math.
