Amazon Is Spending $220 Billion on Data Centers. Here's Why That Makes The Stock an Incredible Buy

Amazon is projected to spend roughly $220 billion on data-center capital expenditures, consuming its available cash flow and requiring use of cash reserves and debt, but the investment is supporting accelerating AWS growth. AWS Q2 revenue rose 37% year over year versus 28% in Q1, while operating income increased 64%; its backlog reached $496 billion and grew at a triple-digit rate. CEO Andy Jassy said AWS lacks sufficient capacity to meet demand through 2026 and likely 2027, supporting the view that AI infrastructure spending could produce durable cloud-market-share gains.

Analysis

The investable issue is not whether AI demand exists, but whether AWS can convert capacity additions into durable return on invested capital before the depreciation, power, and financing burden resets Amazon's consolidated valuation. A backlog is not revenue and may contain multi-year commitments with cancellation, repricing, or deployment contingencies; the key next-quarter proof points are AWS growth, operating-margin resilience, and disclosed capacity utilization rather than headline demand commentary.

AMZN's elevated build rate creates a two-sided competitive dynamic. If capacity remains scarce through 2027, AWS can lock in workloads and raise switching costs, while customers unable to obtain AWS capacity shift incremental spend toward MSFT Azure, GOOGL Cloud, Oracle (ORCL), CoreWeave (CRWV), and specialized GPU clouds; supply scarcity is therefore supportive of the entire compute ecosystem, including NVDA, but weakens the claim that AWS captures all incremental demand.

Near term, the stock can work if earnings demonstrate that incremental AWS revenue scales faster than depreciation and interest expense. Over 6-18 months, the larger risk is simultaneous hyperscaler capacity completion: excess GPU/compute supply would pressure utilization and pricing just as capex is highest, producing FCF disappointment and a multiple reset despite healthy reported revenue growth. The contrarian view is that consensus may be underestimating this capital-cycle risk, particularly if enterprise AI pilots do not translate into recurring production workloads at the pace implied by supplier reservations.

Key Decisions for Investors

• Maintain a tactical long AMZN only through the next AWS earnings print, sized against a 6-8% downside stop: add if AWS revenue growth and operating income both accelerate while management does not raise the capex run-rate. Target a 10-15% relative move versus the Nasdaq over 1-3 months; exit if AWS margin declines despite stronger revenue growth.
• Express the capacity-scarcity thesis as long AMZN / short an equal beta-weighted basket of legacy enterprise software ETFs or names with AI-displacement exposure, rather than outright long NVDA. AWS monetization is the cleaner catalyst over the next two quarters, while the short leg hedges a broad AI risk-on move; reassess if enterprise software bookings reaccelerate.
• Use a relative-value watch: long AMZN versus short ORCL only if AWS's next reported growth rate remains above 30% and AWS margin is stable or higher. This tests whether AWS is retaining the highest-quality workloads; do not initiate absent those data because Oracle may be the marginal beneficiary of AWS supply constraints.
• Set a risk alert for any 2027 capacity-supply commentary from AMZN, MSFT, GOOGL, or NVDA indicating easing lead times or lower utilization. That would undermine the scarcity premium and favors reducing hyperscaler exposure before the 6-18 month capex payback debate intensifies.