Anthropic Has Committed More Than $100 Billion to AWS, and Its Prospectus Could Reveal More Details About This Contract

Anthropic said it will spend more than $100B with Amazon Web Services (AWS) over the next 10 years, a commitment equal to ~1/5 of AWS’s $496B backlog reported for June (up from ~$195B mid-2025). AWS also disclosed rapidly increasing profitability and AI demand alongside the contract, with Q2 2026 AWS revenue up 37% YoY to $42.2B and segment operating income up ~63% YoY to $16.6B. Anthropic’s IPO prospectus is expected after Labor Day, potentially in late September/October, which may improve visibility on Anthropic’s audited financials and purchase commitments that underpin AWS backlog.

Analysis

The near-term market setup is less about the IPO itself and more about whether the prospectus converts a large, reported customer commitment into auditable evidence of demand durability. For AMZN, that matters because AWS has already won the economics of the narrative: if third-party funding and customer cash generation are as strong as reported, the market can start capitalizing a longer-duration AI backlog with less discount for cancellation risk. The bigger second-order winner may be Amazon’s own silicon stack; if the workload truly stays on proprietary chips, every incremental AI dollar is more defensible margin than merchant-GPU spend.

The risk is that the prospectus does the opposite and exposes a financing loop: aggressive compute orders supported by private-market valuation rather than operating cash flow. That would not just pressure the customer story; it would also make AWS backlog look more like contingent demand than revenue visibility, which could compress the multiple on cloud growth quality even if top-line remains strong. In that scenario, NVDA may also catch a relative bid underperformance if investors conclude a larger share of AI training spend is being internalized by hyperscalers rather than flowing to merchant accelerators.

The catalyst path is staged. In the next days, AMZN can trade on headline extrapolation; over 1-3 months, the prospectus should reveal whether the customer mix is self-funding or still heavily equity-subsidized; over 6-18 months, the key question is whether AWS’s AI attach rate converts backlog into sustained operating leverage or merely deferred capex. The thesis is falsified if AWS growth decelerates sharply after the filing, or if the prospectus shows materially weaker cash generation / higher burn than private estimates imply.

Key Decisions for Investors

• Long AMZN into the prospectus window; use the filing as the catalyst. Risk/reward is best if the market is underestimating backlog quality, but trim aggressively if the document shows weak customer liquidity or slower-than-expected AI monetization.
• Relative value: long AMZN / short NVDA for 1-3 months if the prospectus emphasizes Amazon-controlled silicon and non-merchant AI infrastructure. This is a cleaner way to express hyperscaler share capture than a naked AMZN long.
• Watchlist, not a trade yet: if the prospectus shows Anthropic operating losses still widening faster than revenue and funding support is equity-dependent, fade any post-filing AMZN pop. That would imply backlog quality is being overcapitalized.
• Set a tactical alert on AMZN if the stock gaps >5% on filing hype; fade part of the move unless the prospectus also shows audited unit economics and strong cash conversion. The edge here is verification, not rumor.
• If NVDA underperforms on the filing, consider adding a short-term hedge via NVDA puts rather than outright short exposure; the falsifier is evidence that AI capex is still expanding broadly across merchant silicon, which would blunt the relative thesis.