Anthropic Just Revealed a $518 Billion AI Spending Plan. Here Are the Stocks That Could Win.

Anthropic plans to spend $518 billion on cloud computing and AI infrastructure over the next decade, with 80% of commitments non-cancelable or payable regardless of usage. Alphabet is set to receive at least $111.1 billion, Amazon at least $110 billion, and Microsoft $31.4 billion, creating substantial long-term revenue visibility for the largest cloud providers. The planned IPO seeks a $2 trillion valuation despite Anthropic posting a nearly $42 billion net loss in 2025, while Alphabet and Amazon have already recorded significant unrealized gains on their Anthropic stakes.

Analysis

The important signal is not nominal backlog but capacity underwriting: long-dated take-or-pay commitments convert AI infrastructure from speculative capex into contracted utilization, supporting higher cloud revenue visibility and reducing the risk that recent hyperscaler build-outs become stranded assets. GOOG has the cleanest incremental read-through because third-party adoption of TPUs validates its vertically integrated accelerator stack; this can improve Cloud gross-margin durability versus GPU-rental economics and create a credible alternative to NVDA-dependent deployments over the next 6-18 months.

AMZN gains comparable contracted demand, but its economics are more mixed. Project-specific chip commitments should lift Trainium scale and lower unit costs, yet AWS must fund power, networking and depreciation ahead of revenue recognition; execution risk is highest if customer demand ramps slower than the fixed capacity schedule. MSFT's smaller infrastructure exposure is strategically more valuable through distribution: Claude availability inside Copilot products may improve enterprise conversion, but it also weakens the exclusivity premium of MSFT's OpenAI relationship and could raise model-provider bargaining power.

Consensus is likely overstating the immediate EPS impact. A ten-year commitment is not equivalent to near-term revenue, and the key swing variable is the annual ramp profile, not headline contract value. The more consequential near-term catalyst is whether these contracts allow GOOG and AMZN to guide to accelerating Cloud growth while holding capex-to-revenue stable; failure to do so would revive free-cash-flow and depreciation concerns within 1-3 quarters. A stressed Anthropic financing outcome would also test the quality of supposedly non-cancelable obligations, as renegotiation risk rises sharply when a customer lacks access to equity or debt capital.

Key Decisions for Investors

• Initiate a 6-12 month long GOOG / short AMZN pair, sized market-neutral: TPU third-party validation offers greater multiple-expansion potential for GOOG, while AMZN carries more construction, power and depreciation exposure. Reassess if AWS growth reaccelerates by more than 300bps versus Google Cloud for two consecutive quarters or if GOOG Cloud operating margin contracts.
• Add GOOG on evidence that Cloud growth remains above 30% while consolidated capex intensity stabilizes over the next two earnings reports. Target 10-15% relative outperformance versus the Nasdaq; stop if management signals incremental capex without a corresponding improvement in cloud backlog, utilization, or operating-margin outlook.
• Do not chase NVDA lower solely on custom-chip adoption. Instead, monitor disclosed accelerator mix and TPU/Trainium performance benchmarks: a sustained shift of large-model training workloads away from GPUs is a 12-24 month margin and multiple risk, but current information is insufficient for a standalone short recommendation.
• Watch Anthropic IPO filings and financing disclosures before assigning full credit value to the contracted spend. A delayed offering, material reduction in expected proceeds, or widening AI-infrastructure financing spreads would be a trigger to reduce hyperscaler AI-capex exposure rather than a signal to add.