Earnings season takeaways: resilient, but for how long?

The setup is less about a clean growth scare than about a margin-compression regime shift. The market is starting to price the lag between strong top-line data and deteriorating second-half operating leverage: if input costs keep moving up while consumer volume slows, earnings revisions can stay negative even without an outright recession. That is especially toxic for high-multiple software, semis, and ad-exposed tech where valuation depends on durable gross margin expansion, not just current demand.

The more interesting second-order effect is that AI spend is becoming a capital-allocation sink that can support a narrow set of winners while starving the rest of the index. Hyperscalers can keep lifting capex, but the funding source increasingly comes from slowing share buybacks, tighter hiring elsewhere, and weaker incremental demand in non-AI categories. That creates a bifurcated tape: infrastructure beneficiaries can outperform even as the broader market de-rates on lower breadth and more selective earnings beats.

On the consumer side, the pressure is likely to show up first in discretionary subsectors with low pricing power and high gasoline sensitivity, then in staples as trade-down and private-label mix intensify. The consensus seems too comfortable extrapolating resilient spending into the back half; the risk is not a collapse in aggregate consumption, but a rotation from premium to value that destroys unit economics for certain retailers and brands. That argues for being long firms that own the infrastructure of AI and short names exposed to middle-income consumer elasticity and freight/energy pass-through.

For GS, the near-term read-through is mixed: trading and financing can benefit from volatility and capex activity, but investment banking and risk appetite may weaken if the market keeps repricing duration and equity issuance windows close. For NDAQ, the immediate pain is lower transaction activity and a weaker risk tone, but the bigger issue is that a prolonged factor rotation away from growth would pressure listing and market data multiples, especially if tech breadth keeps narrowing.