Anthropic disclosed detailed investment agreements with Amazon and Google this week, cementing cloud provider stakes in one of AI's most closely watched startups ahead of its initial public offering.
Amazon will hold 14.7 percent of Anthropic after the IPO, stemming from the conversion of promissory notes into non-voting shares that carry dividend rights and can convert to voting Class A shares upon sale. The cloud giant's original notes, issued in 2023 and 2024, carried conversion prices of approximately $2.09 or $5.05 per share. Had Amazon converted all outstanding notes at IPO time, its stake would have reached 23.8 percent—a decision to hold back billions in notes suggests confidence in future upside or deliberate position management.
Google will own 13.7 percent of Anthropic through a mix of note conversions and private share purchases. Beyond equity, Amazon locked in a decade-long commitment from Anthropic to spend over $100 billion on Amazon Web Services—a revenue floor that ties the AI startup's computational infrastructure directly to AWS margins.
A $26 billion private placement of non-voting shares drew participation from AMD, Broadcom, Google, Micron, Microsoft, Nvidia, and TSMC Partners. Within that syndicate, Google and Nvidia each committed $10 billion. The structure—non-voting shares for strategic investors—protects Anthropic's founder control while guaranteeing capital and locking in cloud provider dependency.
Amazon's early-stage bets on Anthropic, made when the startup was overshadowed by OpenAI, now yield both equity upside and a foundational cloud contract. The capital allocation reveals how cloud giants are using venture checks not for portfolio returns but as moat-building tools: securing preferential access to AI model weights, computational priority, and contractual lock-in that competitors cannot easily replicate.

