Amazon and 2 Other AI Stocks to Buy With $5,000

The article recommends a $5,000 AI-infrastructure basket allocated 50% to Amazon, 35% to Taiwan Semiconductor, and 15% to Applied Digital. AWS revenue grew 37% year over year to $42.2 billion in Q2, generating $16.6 billion in operating income, while TSMC's August revenue reached a record NT$514.8 billion ($16 billion), up 53%. Applied Digital offers the highest-risk upside, with roughly $36 billion of contracted 15-year rent against fiscal 2026 revenue of $611.3 million and a $249.2 million net loss, as most leased capacity remains under construction.

Analysis

The apparent diversification is lower than advertised: AMZN, TSM, and APLD are all levered to the same hyperscaler capex cycle, but with materially different cash-conversion profiles. AMZN can absorb a utilization shortfall through retail and advertising, while TSM’s leading-edge utilization and APLD’s project returns are far more exposed to a pause in orders from NVDA-linked cloud deployments. In a 1-3 month risk-off move, the highest-beta data-center landlord should underperform both the cloud platform and foundry despite its contractual backlog.

APLD’s contracted-rent framing should be discounted heavily until delivered MW, energization dates, customer credit concentration, project-level debt terms, and required equity funding are independently confirmed. The key equity risk is not merely construction delay; it is a financing loop in which delayed service commencement pushes cash receipts out while interest expense and dilution arrive on schedule. A lease portfolio with take-or-pay provisions can still create poor equity returns if incremental capital is raised at a depressed valuation or if tenant concentration weakens lenders’ willingness to fund expansion.

TSM offers cleaner AI volume exposure than most chip designers, but its valuation discount is partly an embedded geopolitical and customer-concentration option, not simply a mispricing. The more non-obvious beneficiary of sustained leading-edge demand is ASML, whose bottleneck position converts node migration into backlog and service revenue with less Taiwan-specific operational exposure. Conversely, a broad AI capex deceleration would likely hit semiconductor equipment orders after foundry utilization rolls over, creating a 6-18 month lag rather than an immediate collapse.

Consensus is likely underestimating the possibility that AWS growth comes with structurally lower incremental returns as cloud providers compete on AI capacity availability and pricing. The near-term equity response can remain positive on revenue acceleration, but the more important catalyst is whether capex intensity peaks before depreciation and power costs compress segment margins. AMZN is therefore a quality holding, not a clean AI-margin expansion trade.

Key Decisions for Investors

• Prefer a 3-6 month pair trade: long AMZN / short APLD in equal dollar exposure. This isolates execution and financing risk at APLD against a better-capitalized AI infrastructure beneficiary; reassess if APLD demonstrates on-time energization and project funding without incremental equity issuance.
• Accumulate TSM on 8-12% pullbacks rather than chase momentum, with a 6-18 month horizon. Falsify the thesis on sustained leading-edge utilization deterioration, a material reduction in hyperscaler capex guidance, or a Taiwan-risk escalation that widens the valuation discount.
• Use ASML as a higher-quality complementary long to TSM for investors seeking foundry-capex exposure with less single-island production risk. Size modestly because equipment order weakness is typically delayed; reduce if TSM and Samsung signal node-capex deferrals.
• Avoid treating APLD backlog as current enterprise value support until disclosed live capacity, construction milestones, debt maturities, and customer concentration support a project-level cash-flow model. If initiating, cap exposure at venture-style sizing and use a 12-month horizon; the downside case includes dilution and delayed rent commencement, not just a lower multiple.
• For AMZN, watch quarterly AWS operating-margin progression relative to capex and depreciation. Maintain exposure only while growth converts to stable or improving incremental margins; a margin reset alongside elevated infrastructure spending would challenge the multiple within 1-3 quarters.