Anthropic Plans to Spend $518 Billion on Cloud and Computing Power

Anthropic’s draft IPO prospectus reportedly outlines more than $518 billion in planned cloud, computing and infrastructure spending over the next decade, with about 80% non-cancellable. Revenue grew 12-fold to nearly $4.6 billion in 2025 and its annualized run rate reportedly reached $65 billion by July 2026, but the company also recorded an $8 billion operating loss and nearly a quarter of revenue came from two customers. The article questions the growth assumptions behind a reported $2 trillion IPO valuation and recommends waiting post-IPO for evidence of more diversified, durable revenue.

Analysis

The key asymmetry is contractual: if the reported non-cancellable commitments are accurate, Anthropic could preserve demand for cloud and compute vendors even as its own growth slows—but it would absorb utilization and margin risk before those suppliers necessarily do. That makes the commitments a potential revenue floor for infrastructure providers, not evidence of attractive returns for Anthropic shareholders. The second-order risk is a capacity race: if AI demand disappoints, cloud providers may compete harder on price to fill capacity, weakening returns across the sector.

AMZN and GOOG have two-sided exposure. Their partnership and investment links may benefit from Anthropic’s infrastructure purchases and strategic value, but direct competition and any investment losses can offset that upside; neither is a clean proxy for Anthropic’s growth. AVGO may participate through processor supply, but the article does not quantify its Anthropic revenue exposure, and a slowdown could pressure supplier expectations before contracted infrastructure spend unwinds.

Near term, IPO pricing and positioning may dominate fundamentals; a headline valuation is not a basis for underwriting without dilution, share-sale, and valuation details. Over 1–3 months, monitor customer concentration, contract duration, and reported revenue conversion. Over 6–18 months, the critical test is whether recurring, diversified demand supports utilization and economics against committed spend. The contrarian risk to the cautious view: rapid revenue expansion could make today’s concentration temporary. Falsifiers include customer diversification and longer contracts; thesis confirmation would be slowing revenue or weaker cloud/AI infrastructure guidance. No valuation or exposure data here supports an outright directional trade.

Key Decisions for Investors

• Do not chase the IPO on growth claims alone; wait for prospectus detail on customer concentration, contract terms, dilution, and the composition and timing of committed spend.
• Treat AMZN and GOOG as diversified, two-sided strategic exposures—not substitutes for direct Anthropic ownership. Reassess only against their own cloud growth, investment disclosures, and guidance.
• Keep AVGO on an exposure watchlist rather than treating it as a direct Anthropic proxy; verify customer-level revenue disclosure and monitor semiconductor demand and guidance for evidence of order changes.
• Potential relative-value alert, not a recommendation: compare diversified cloud incumbents with AI infrastructure suppliers if AI demand weakens; avoid expressing this as a trade until valuation, exposure, and post-IPO supply data are available.
• Falsify the cautious thesis if Anthropic demonstrates sustained customer diversification and longer-duration contracts alongside revenue growth; strengthen it if growth decelerates or cloud and chip suppliers cut AI-related outlooks.