How Andy Jassy Just Convinced Me That Amazon Stock Is a Buy

Amazon says its AWS AI business has reached a $15 billion run rate just three years in, versus $58 million for AWS at a comparable stage, supporting the case for its planned $200 billion of AI-related capex in 2026. Andy Jassy argued the investment is justified, with Amazon adding 3.9 GW of power in 2025 and monetizing that capacity quickly, including two customers requesting all 2026 Graviton chip capacity. The article is constructive on Amazon's long-term AI opportunity, though it is more commentary than a near-term catalyst.

Analysis

The market is still pricing Amazon’s AI spend like a cost overhang, but the better read is that this is a capacity pre-emption trade: whoever secures power, chips, and inference stack early will control the next pricing cycle. The second-order effect is that AWS is not just chasing demand; it is trying to lock in enterprise workflows before rivals can normalize comparable performance/latency, which raises switching costs and makes the capex look more like infrastructure toll-building than speculative AI spend.

The most important signal is the pace of monetization versus buildout. If demand is genuinely outrunning installed capacity, the risk is not that Amazon overbuilds in the near term, but that competitors underbuild and lose share in the 12–24 month window. That argues for relative winners in the AI supply chain where Amazon’s spending becomes downstream demand: power equipment, networking, memory, and foundry-adjacent names should see the cleaner earnings torque than the headline hyperscaler itself.

The contrarian risk is that investors may be underestimating execution complexity at scale, especially if inference economics worsen before utilization normalizes. If Bedrock or adjacent AI services require repeated architectural resets, margin expansion could lag capex growth for several quarters, creating a valuation air pocket even if revenue is strong. The key catalyst to watch is whether AWS AI revenue growth keeps outrunning capex intensity through the next two quarters; if not, the market will reprice this as a delayed-return story rather than a durable compounding asset.

Net: the setup favors buying the infrastructure picks and treating AMZN as a longer-duration core rather than chasing it after the capex headline. The sentiment read is mildly positive, but the real alpha is in dispersion: AWS spend is a tailwind for selected suppliers and a challenge to cloud peers that cannot match deployment speed.

Key Decisions for Investors

• Long AMZN on 3-6 month horizon on any post-capex weakness; use a staggered entry because the core thesis is duration, not immediate multiple expansion. Risk/reward improves if AWS AI monetization stays above capex growth for two consecutive quarters.
• Buy NVDA versus the broader Mag-7 basket as a relative winner from sustained hyperscaler spend; the cleaner trade is spread capture rather than outright beta. Target 6-12 months, with risk if Amazon shifts more spend toward internally optimized silicon.
• Pair long AMZN / short a lower-quality cloud beneficiary or AI-exposed software name with weaker monetization visibility over the next 1-2 quarters. The idea is to own the spend leader while fading names where AI narrative still exceeds revenue conversion.
• Initiate a basket long in power/grid and data-center infrastructure suppliers over the next 1-3 months; Amazon’s buildout increases structural demand for power delivery and thermal management. This is the highest second-order convexity trade if capex remains elevated.
• Avoid shorting AMZN outright into this print unless you have a catalyst for utilization disappointment; the asymmetric risk is that capacity scarcity leads to faster-than-expected revenue capture, forcing shorts to cover into strength.