AI Expectations Hit a Fever Pitch

Hyperscaler debt issuance is not a solvency tell; it is a duration-arbitrage tell. AMZN, GOOGL, and META can lock in long-dated capital below their project hurdle rates, which should keep AI capex elevated into 2026 and support the compute stack near term. The second-order effect is that balance-sheet strength becomes the real moat: large issuers can outspend smaller AI entrants and delay any normalization in spend, but the whole trade becomes vulnerable if spreads widen or capex growth decelerates.

The more important signal in semis is that the market is shifting from scarcity pricing to supply-response skepticism over the next 1-3 quarters. MU and SKHYV are the most exposed because incremental supply and fresh capital can cap margin expansion even if demand stays healthy; NVDA is less directly exposed, but sentiment can still de-rate if memory exuberance cools. Watch next earnings for HBM lead times and 2H pricing commentary; if those stabilize, the selloff is likely overdone.

In defense, the structural winner is not the cheapest drone vendor but the primes that can absorb sensor, sonar, and electronic-warfare capabilities into existing procurement channels. That favors LMT and LHX over standalone disruptors like AVAV and KTOS, whose unit growth may not translate into durable margins. Contrarian view: the market may be overpricing drone disruption and underestimating how quickly the primes can buy the relevant tech at modest multiples and repackage it for Pentagon/NATO demand.