Amazon (AMZN) Q2 2026 Earnings Call Transcript

Amazon reported Q2 revenue of $200.6B (+20% YoY) and operating income of $27.5B (+43% YoY), supported by a $1.2B net operating tailwind from tariff refunds (~$600M) and energy-derivative accounting gains (~$600M). AWS revenue surged to $42.2B (+36.7% YoY) with operating income of $16.6B and a 39.4% operating margin; AWS also hit a $496B backlog and an $169B annualized revenue run-rate, while generative AI and chips run-rates each exceeded $25B. CapEx guidance for 2026 was raised to ~$220B (from ~$200B) despite near-term free cash flow headwinds and transport/memory supply-chain inflation, and Q3 net sales guidance is $197B-$202B (includes ~80 bps FX headwind).

Analysis

The market is likely to over-focus on the capex step-up and underwrite it as a drag, but the more important signal is that AWS demand is now arriving with enough duration and contractual visibility to justify pre-build. That shifts AMZN from a “growth at any cost” story toward a utility-like balance of contracted backlog plus optionality, which should support multiple expansion if management converts capacity into revenue through 2027-28. The hidden winner is still AMZN’s own infrastructure stack: higher AI usage should raise core CPU/storage/network consumption alongside inference, improving attach rates even where model pricing compresses.

Second-order beneficiaries are the picks-and-shovels names tied to AI deployment, but the mix matters. NVDA still benefits near term from the industry’s aggregate capex wave, yet AWS’s push toward Trainium/Graviton and even direct chip supply is a medium-term share-overhang for merchant silicon if large cloud buyers internalize more of the stack. On the software side, AWS is increasingly trying to own workflow, security, and productivity layers; that is a subtle threat to standalone point solutions such as SNOW, NOW, CRM, and ADBE if enterprise budgets consolidate around one platform with built-in agents and security controls.

Contrarian view: consensus will likely anchor on free-cash-flow pressure and conclude the spend is too big. I think that is the wrong frame unless capacity fails to monetize quickly; the real falsifier is not capex alone, but whether AWS margin stays structurally above the mid-30s ex one-offs while backlog converts into visible revenue. If 2H26 and early-2027 guidance starts to lag the buildout, then the stock becomes a capital-intensity story; until then, the setup argues for buying weakness rather than fighting the earnings power.

Key Decisions for Investors

• Buy AMZN on a post-earnings pullback or via a 3-6 month bullish risk reversal; thesis is that capex is being funded ahead of contracted demand, not ahead of hope. Falsifier: AWS growth or backlog conversion slows materially, or ex-one-off AWS margin falls toward the low-30s.
• Tactically long NVDA for 1-3 months into the broader AI capex cycle, but keep size modest. Near-term AWS spend still lifts the whole semi chain; medium-term, Trim exposure if AWS increasingly substitutes Trainium for merchant accelerators.
• Pair trade: long AMZN / short SNOW over 1-3 months. The platform consolidation narrative and AWS-native agent/security stack raise the risk that standalone data/AI workflow budgets are captured inside hyperscaler tools. Cover if SNOW shows renewed net retention acceleration tied to non-AWS workloads.
• Watch-list, not a trade: short CRM or NOW only on evidence of budget consolidation into Amazon Q/Bedrock-style workflows. Without proof of displacement, this is more of a medium-term competitive alert than an immediate catalyst.