Exclusive-Nvidia in talks to invest in Anthropic’s mega IPO, sources say

Anthropic is reportedly seeking to raise up to $100 billion in an IPO that could value the AI startup at about $2 trillion, with Nvidia considering an anchor investment of up to $10 billion. The offering, potentially the largest IPO ever, would test public-market appetite for frontier-AI valuations and capital requirements while strengthening Anthropic's strategic compute relationship with Nvidia. Anthropic's annualized revenue run rate reportedly exceeded $65 billion by end-July, up from about $9 billion at end-2025, while its 2028 revenue projections are roughly $190 billion-$200 billion.

Analysis

The key equity issue is not incremental AI demand but circularity of its financing. A strategic equity commitment by NVDA to a customer that simultaneously funds GPU-backed cloud capacity can support near-term accelerator orders while weakening the quality-of-demand narrative investors use to justify NVDA's premium multiple. Over the next 1-3 months, disclosed IPO terms, related-party commitments, lockups, and any use-of-proceeds language will matter more than a headline valuation; evidence that capital is effectively recycled into vendor purchases would raise the risk of a multiple reset across AI infrastructure.

AMZN, MSFT and GOOG have different exposures: their cloud economics benefit from sustained model-training and inference workloads, but their returns depend on utilization and pricing rather than simply booked capacity. AMZN has the most direct upside if committed AWS capacity converts into high utilization, while AVGO is the better 6-18 month second-order beneficiary if frontier labs increasingly use custom ASICs to reduce inference cost. That substitution risk is underappreciated in NVDA estimates because it emerges gradually through mix and unit economics, not an abrupt GPU demand collapse.

A successful public-market clearing price would create a new, highly liquid benchmark for frontier-model economics and could initially lift AI multiples. The contrarian risk is that public investors demand disclosure on compute costs, customer concentration, and cash burn that private financing has obscured; a weak aftermarket would transmit first to high-duration infrastructure names rather than diversified hyperscalers. Falsification of the circular-financing concern would be independently demonstrated third-party demand, durable cloud utilization, and stable NVDA data-center gross margin despite customer chip diversification.

Key Decisions for Investors

• Do not chase NVDA on preliminary financing headlines. Maintain core exposure only; add if IPO documentation shows arm's-length commitments and NVDA's next earnings sustains data-center gross margin/guidance. Reduce tactical exposure if related-party funding becomes material or management indicates customer-financing intensity is rising.
• Build a 6-18 month relative-value position: long AVGO versus short a beta-adjusted NVDA basket, sized small initially. The thesis is custom-ASIC share gains in inference; reassess if AVGO fails to show AI semiconductor backlog growth or if NVDA demonstrates that GPU total-cost-of-ownership remains decisively superior.
• Prefer AMZN over GOOG and MSFT for cloud-capacity upside only after quarterly AWS revenue acceleration is accompanied by stable or improving AWS operating margin. The risk/reward is unfavorable if capacity commitments inflate capex without utilization; exit on a material AWS margin step-down without corresponding revenue acceleration.
• Treat the eventual IPO as an event-risk alert rather than a pre-listing trade. Monitor the revenue multiple implied by the indicated range and post-listing trading: a sustained discount to profitable software peers would be a signal to trim high-multiple AI infrastructure exposure, while a strong aftermarket with transparent compute-unit economics would support renewed AI-beta longs.