The AI earnings bonanza is splitting investors

Amazon, Google, Microsoft, and Meta all framed earnings around AI-driven growth, with Amazon AWS revenue reaching $37.6B versus $36.7B expected and Google Cloud growing 63% year over year to $20B. Amazon's AWS backlog rose $120B quarter over quarter to $364B, while Microsoft’s cloud backlog increased 99% to $627B and the company said supply constraints will persist through 2026. Investors rewarded Amazon and Google but sold Meta and Microsoft, making the reaction mixed despite broadly strong AI and cloud results.

Analysis

The market is rewarding the AI platforms where demand is proving monetizable now, and punishing the names where the near-term earnings math is being overtaken by capex anxiety. The second-order winner is not just the cloud operators but the suppliers that can relieve the bottleneck: custom silicon, networking, power, and datacenter infrastructure should keep taking share as hyperscalers race to secure capacity rather than optimize near-term ROIC. That dynamic is especially supportive for AMD and the broader non-Nvidia AI ecosystem, because every additional in-house accelerator deployment by cloud vendors increases buyer diversity and weakens a single-vendor pricing regime.

The key contrast is backlog quality. A backlog supported by external enterprise demand is more durable than one dominated by a single strategic customer or pre-committed internal buildout, so the market is implicitly assigning lower incremental value to demand visibility that is already tied to a supply-constrained capex cycle. That matters for the next 3-6 months: if supply remains tight into 2026, earnings estimates will likely rise slower than headline spend, creating a window where capex-heavy names can de-rate even as fundamentals stay healthy. In other words, this is less a demand shock than a duration shock.

Contrarianly, the move against Meta and Microsoft may be too blunt if investors are extrapolating one-quarter capex pressure into a permanent margin reset. If AI usage keeps expanding, the payoff can show up later in higher attach rates, better ad pricing, and stronger retention, which means the downside is more about timing than thesis failure. The real risk to the bullish cloud view is that customers eventually push back on price-per-token and infra costs, which would cap backlog conversion and compress returns on the incremental investment over the next 12-24 months.

Key Decisions for Investors

• Go long GOOGL vs short NVDA for a 1-3 month horizon: Google is moving from rented capacity to owned silicon distribution, which is structurally better for margin capture and could pressure NVIDIA pricing over time; keep the short sized modestly because NVIDIA remains the cleaner earnings momentum name.
• Buy AMZN on weakness for a 3-6 month trade: AWS backlog plus accelerating chip revenue suggests an underappreciated mix shift toward higher-value AI infrastructure; target a 10-15% rebound with a tight stop if cloud growth decelerates next print.
• Use META selloffs to sell puts or buy call spreads 6-9 months out: the market is over-discounting capex while underweighting the operating leverage from AI-driven ad pricing and impression growth; risk/reward improves if implied volatility stays elevated.
• Fade MSFT on rallies versus ORCL/AMD-style beneficiaries for the next 2 quarters: the stock is likely to stay capped until supply constraints visibly ease, making it a better relative short than an outright directional short given OpenAI backlog support.
• Long basket trade in AI infrastructure suppliers for 6-12 months: favor AMT, EQIX, ANET, and power/grid names over the mega-cap software cohort, because the real bottleneck is physical deployment capacity rather than software demand.