Magnificent Seven, Markets, and Mailbag with CEO Tom Gardner

Alphabet, Amazon, Microsoft, and Meta all posted strong AI/cloud-related earnings, with cloud revenue growth ranging from 28% at AWS to 63% at Google Cloud, while the four companies collectively are expected to spend over $600B on capex this year. The discussion focused on rising memory and infrastructure costs, historically weak consumer sentiment, and whether hyperscaler spending supports downstream suppliers more than it threatens margins. Nvidia’s moat remains intact near term, but the panel flagged potential margin pressure if hyperscalers’ custom chips gain traction.

Analysis

The signal is not “AI is slowing”; it’s that the capex cycle is shifting from scarcity to bottleneck management. Public-cloud leaders with backlog visibility should keep compounding because higher component costs and longer procurement lead times make on-prem economics worse, not better, for most enterprises. That should widen the moat for GOOGL/MSFT/AMZN versus more advertising- or consumer-exposed names, while creating a second-order boost for the picks-and-shovels layer that sells power, cooling, networking, memory, and installation capacity.

The market is correctly discriminating between demand-backed spend and speculative spend. META’s risk is that incremental capex is still being funded against a more cyclical revenue base; if ad budgets soften while memory and build costs inflate, operating leverage can flip quickly over the next 2-3 quarters. By contrast, the cloud trio can pass through pricing and monetize AI workloads faster, so higher capex there may actually be a near-term negative to margins but a medium-term positive to backlog conversion and share gains.

The more interesting contrarian is on NVDA: custom silicon is not an existential threat, but it does cap the multiple if hyperscaler in-house chips keep taking repetitive workloads. The market still seems to be pricing “all compute runs through NVDA,” when the more realistic outcome is a two-tier stack: frontier training and high-value inference stay NVDA-heavy, while internal, low-margin workloads migrate to custom ASICs. That implies NVDA earnings can keep growing while the stock derates if gross margin/ROIC peak before revenue does.

Key Decisions for Investors

• Stay long GOOGL/MSFT/AMZN versus META for the next 2-3 quarters; the cleaner earnings quality and backlog support should outperform if enterprise AI spend stays hot. Prefer a relative-value long GOOGL / short META pair.
• Add to MU on pullbacks: memory inflation is not just a cost story, it is a demand-validation story for the AI stack. Use 3-6 month horizon; risk is a sudden digestion pause if hyperscaler capex gets delayed.
• Reduce or hedge NVDA exposure with a covered-call overwrite into the next 1-2 earnings cycles; upside remains intact, but margin compression from custom silicon is a credible multiple ceiling over 6-12 months.
• Long DELL on weakness versus short-duration consumer cyclicals: if on-prem AI economics deteriorate, enterprise buyers should continue shifting toward managed infrastructure. Best risk/reward is a pair against a hardware name with less AI content.
• Watch META as a tactical short only if ad-market commentary weakens further; otherwise avoid outright shorting because the trade is fragile until capex-to-revenue payback visibly stalls.