SpaceX Briefly Dropped Out of the $2 Trillion Club This Week. Here's My Top Buy of the 6 Stocks That Are Secure Members.

Amazon plans to spend $200B on AI infrastructure in 2026, aiming to sustain AWS’s ~28% cloud market share while monetizing investments. AWS net sales rose to $37.58B in Q1 (+28% YoY), and Amazon’s custom chip business (Graviton/Trainium/Nitro) is said to run at an annual revenue run rate above $20B. Despite investor concerns about capex cost and potential GPU obsolescence, management (Andy Jassy) cites customer commitments and expects compelling operating margins and ROIC, supporting a modestly positive outlook for AMZN.

Analysis

The market is still treating AI capex as a near-term free-cash-flow drag, but the more important mechanism is control of scarce inference capacity. For AMZN, that means higher near-term depreciation and working-capital intensity, but also a stronger moat because enterprise workloads tend to stick once they are integrated into procurement, security, and data pipelines. The key question for the next 1-3 months is not spend size; it is whether utilization and backlog convert that spend into pricing power without forcing AWS discounting.

Second-order winners are the picks-and-shovels names tied to deployed compute, but the mix matters. NVDA benefits immediately from the build-out, yet AMZN’s custom silicon trajectory is a medium-term substitution risk: every incremental internal chip dollar is a dollar of margin capture that does not accrue to external GPU vendors. That makes AMZN a better long-duration AI infrastructure owner than a pure hardware beta trade, while CBRS-like names are more fragile because their upside depends on sustained customer concentration and continued funding access.

The contrarian miss is that the spend may actually be a signal of confidence rather than desperation: hyperscalers with the largest installed base can monetize AI faster because they already own distribution. If AWS keeps comping at high-20s growth, the bear case needs a rising evidence bar—slowing backlog conversion, weaker attach rates, or capex guidance cut. SPCX looks like a different trade entirely: post-IPO volatility and index/lockup flows can dominate fundamentals for months, so earnings disclosure is the real catalyst, not headline market-cap swings.

Key Decisions for Investors

• Long AMZN on pullbacks over the next 2-4 weeks; treat the current capex headline as a multi-year moat investment rather than an earnings miss. Risk/reward is favorable if AWS growth stays >25%, but falsify on any deceleration toward the high teens or a cut to capex commitment language.
• Pair trade: long AMZN / short NVDA for a 6-12 month horizon. Thesis: NVDA captures the first wave of spend, but AMZN’s custom silicon and vertical integration should increasingly internalize margin, making the external GPU wallet share vulnerable if Trainium/Graviton adoption keeps comping triple digits.
• Maintain a tactical long NVDA into the next 1-2 quarterly print windows, but size smaller than AMZN. This is a flow trade on hyperscaler spend, not a structural call; exit if hyperscaler capex guide flattens or gross margins stop expanding despite strong bookings.
• Do not chase SPCX strength for now; wait for first quarterly earnings and lockup expiry dynamics to settle. The stock is more likely to trade on supply overhang and index rebalancing than on business fundamentals in the near term.
• Watch LUV as a secondary beneficiary only if AWS commitments translate into measurable enterprise usage and cost savings; otherwise it remains an anecdotal customer-win, not an investable signal.