Amazon Stock Just Keeps Rising. Here's the Latest Catalyst Driving Shares Higher -- and a Close Look at the Stock.

Amazon shares are up more than 25% over the last 30 days as Meta signed a multiyear deal to use AWS Graviton5 chips for AI workloads, while Anthropic committed over $100 billion to AWS technologies over the next decade. Amazon said its AWS AI revenue run rate was above $15 billion in Q1 2026 and its chip business exceeded a $20 billion annual run rate, both growing at triple-digit rates. The article is positive for Amazon's long-term AI and custom silicon strategy, though the stock's 37x P/E and roughly $200 billion 2026 capex plan temper the near-term upside.

Analysis

The market is beginning to re-rate Amazon from a “cloud retailer” to an AI infrastructure landlord with a differentiated cost curve. The second-order implication is that custom silicon success does not just improve margins; it increases switching costs for model builders who optimize their inference stacks around AWS-native hardware, making compute a sticky annuity rather than a commodity spend. That matters more for inference and agentic workloads than headline model training, so the earnings power can compound even if GPU pricing remains intense.

Meta’s adoption is the cleaner signal because it validates AWS outside of Amazon’s own captive ecosystem. If a hyper-scale buyer with real bargaining power is willing to allocate meaningful production workloads to Graviton, that lowers the perceived “build-it-and-they-will-come” risk around Trainium and future iterations. Anthropic’s long-dated commitment is even more important for capacity planning: it effectively de-risks utilization of Amazon’s capex pipeline over several years, but it also turns AWS into a quasi-strategic utility exposed to one customer’s model roadmap and funding cadence.

The stock move looks directionally correct, but the easy multiple expansion may already be behind us. At current levels, the market is pricing faster monetization of AI spend, while the real test shifts to whether capex converts into incremental free cash flow faster than depreciation, power, and networking costs rise. If conversion lags for even two quarters, the narrative can de-rate quickly because investors will start treating the spend as defensive infrastructure build-out rather than high-ROIC growth.

The contrarian read is that the biggest beneficiaries may not be AMZN or META, but rather the adjacent picks-and-shovels that bottleneck AWS deployment: power equipment, networking, and datacenter cooling. If AWS custom silicon displaces more general-purpose GPU demand in certain workloads, it could pressure the premium economics of exposed AI semiconductor names over time, while improving utilization and pricing discipline for AWS. The key monitor is whether the next disclosed AWS AI deal comes from another top-tier external customer; if not, the market may conclude this is still a narrow validation story rather than a broad demand wave.

Key Decisions for Investors

• Stay long AMZN, but hedge with a near-dated collar or partial call overwrite into the next earnings window; upside remains, but valuation now embeds continued capex-to-cash-flow conversion over the next 2-3 quarters.
• Add META on pullbacks versus AMZN: the deal reduces AWS platform risk for Meta while preserving Meta’s own AI spend leverage; use a 3-6 month horizon and target relative outperformance if inference economics matter more than frontier-training capex.
• Initiate a basket long in AI infrastructure enablers not called out in the article — power, cooling, and networking suppliers — against AMZN on a 6-12 month view; the pair captures the capex spillover without paying the full multiple for AWS execution risk.
• Avoid chasing NVDA here; if custom silicon adoption broadens, it can subtly shift some inference mix away from merchant accelerators. Use the move as a signal to trim into strength rather than add.
• If AMZN gaps further on any additional partnership headlines, consider a tactical short-dated put spread to express mean reversion risk; the trade works if the market is already extrapolating the next 2-3 years of AI economics into the next few weeks.