If I Were in My 20s, I'd Buy These 2 Trillion-Dollar Stocks and Hold Them Forever

The article recommends Amazon and Microsoft as long-term AI-driven investments, citing AWS revenue growth of 37% and Azure growth of 43% in their latest reported quarters. Amazon's AWS has a $496 billion capacity-related customer backlog and Microsoft Azure has a $678 billion backlog, underpinning substantial future cloud and AI demand. Both stocks are characterized as fairly valued, with forward P/E ratios of 24.4 for Amazon and 25.0 for Microsoft, broadly in line with the Nasdaq-100's 24.2 multiple.

Analysis

The investable issue is not AI demand but conversion of reserved capacity into recognized cloud revenue without a disproportionate step-up in depreciation, power costs, and lease obligations. Both AMZN and MSFT have unusually visible demand signals, but reported backlog/RPO is not equivalent to near-term revenue: delivery schedules, customer cancellation rights, and the mix of non-AI commitments determine the conversion rate. Over the next 1-3 quarters, quarterly capex guidance, depreciation growth, and cloud growth ex-capacity constraints matter more than promotional AI-user metrics.

AMZN has the cleaner operating-leverage setup if AWS growth persists while North American retail margins remain stable: incremental cloud revenue has a greater potential to lift consolidated earnings than consensus typically credits when retail fulfillment costs are controlled. MSFT has the stronger distribution advantage in enterprise software, but Copilot monetization faces a harder proof point—paid-seat growth must translate into sustained revenue per seat rather than bundled discounts or customer pilots. The second-order beneficiary is power and data-center infrastructure (VRT, ETN, CEG); the underappreciated loser is conventional software vendors with weak proprietary data or limited workflow embedment, where AI pricing pressure can precede actual customer churn.

The contrarian risk is that hyperscaler capacity additions normalize the scarcity premium. If supply catches demand in 2027, AI compute pricing could fall faster than inference volumes expand, compressing cloud gross margins even as revenue grows. A separate near-term risk is that internally designed accelerators shift a portion of spend away from NVDA; this is a share risk for NVDA, not necessarily a demand-collapse thesis, since custom silicon deployment still requires a broad networking, memory, and packaging ecosystem.

We would avoid treating this as a fresh broad-AI beta signal after a bullish retail-oriented publication. The actionable edge is to own the companies with credible monetization and hedge the capex/power bottleneck, while requiring evidence of margin preservation at the next two earnings prints.

Key Decisions for Investors

• Initiate a 6-12 month long AMZN / short IGV pair, sized beta-neutral. Thesis: AWS capacity conversion and retail-margin resilience can drive estimate revisions, while the software basket carries greater AI-disruption and valuation risk. Exit if AWS growth decelerates for two consecutive quarters or consolidated operating-margin guidance falls despite cloud growth.
• Maintain MSFT as a core long but add only on post-earnings weakness tied to capex concerns, not on Copilot seat headlines. Require evidence that paid Copilot attach and ARPU are rising; reduce if Intelligent Cloud growth remains capacity constrained while depreciation growth materially outpaces revenue growth over the next two quarters.
• Buy a 6-9 month infrastructure basket of VRT, ETN, and CEG against a partial short in SMH, rather than adding outright NVDA exposure. This captures the physical bottleneck in power, cooling, and electrical equipment while limiting exposure to accelerator share shifts; invalidate if hyperscaler capex plans are cut or data-center project lead times normalize materially.
• Set an earnings-monitor alert for AMZN and MSFT: cloud growth below consensus accompanied by rising capex/depreciation is the bearish combination. In that outcome, close cloud longs and consider a tactical short in QQQ for the 1-3 month multiple-compression risk, as mega-cap AI spending has become an index-level earnings-duration assumption.