A debt instrument becomes a stake
TechCrunch reported that Amazon’s convertible notes in Anthropic could become as many as 1.84 billion shares when Anthropic completes its initial public offering. The conversion is subject to a 12% ownership cap, according to the publication’s analysis of company filings.
That makes the IPO more than a fundraising event. It is also the moment when a large part of Amazon’s financial relationship with Anthropic moves from the background of the balance sheet into the company’s visible ownership structure.
Anthropic’s own registration statement says it has $6.2 billion of outstanding convertible notes held by Amazon. Those notes are expected to convert into Class N nonvoting common stock in connection with the IPO, subject to Amazon’s ownership cap. If the conversion would push Amazon above that limit, the excess notes will remain convertible at a later time or may be repaid when they mature, according to Anthropic’s Form S-1 registration statement.
The nonvoting designation is important. Amazon may gain a substantial economic interest without receiving ordinary voting power in proportion to its financial exposure. That distinction could help Anthropic present itself as independent while still giving Amazon a major stake in the value created by the business.
Amazon’s own filing offers an even broader view of the relationship. In its quarterly report for the period ended June 30, 2026, Amazon disclosed $92.5 billion in nonvoting preferred stock holdings and $97.9 billion in convertible note holdings related to Anthropic. The filing also says the notes convert into nonvoting common stock after an IPO, subject to Amazon’s ownership cap, as Amazon’s Form 10-Q details.
The different figures reflect the way strategic investments can be recorded across several classes of securities and financing arrangements. They also show why headline investment totals can be difficult to interpret. A note may begin as a claim on future equity, while preferred shares may carry different economic rights from common stock. Until conversion, the full scale of a partner’s exposure can remain difficult to see.
Amazon is more than a cloud supplier
Amazon’s role extends well beyond writing checks. Anthropic has made AWS its primary cloud partner, and its models are available through Amazon Bedrock, where businesses can access generative AI tools without building their own model infrastructure.
The two companies have continued to deepen that relationship. In a company announcement, Anthropic said it was expanding its compute collaboration with Amazon, including another $5 billion investment from Amazon and the possibility of up to $20 billion more. Anthropic said the deal builds on Amazon’s previous $8 billion investment while maintaining Amazon as a minority investor.
For enterprise customers, that arrangement can make Anthropic’s models easier to buy and deploy. A company already using AWS may be able to place model workloads, data and related services within the same commercial environment. In the future, the choice between AI providers could increasingly depend not only on model quality, but also on which cloud platform already runs an organization’s operations.
That creates a powerful feedback loop. Amazon supplies capital and computing capacity, Anthropic supplies models, and AWS becomes a distribution channel. As more customers use the models through Amazon’s infrastructure, the strategic value of the partnership can grow even if Amazon does not control Anthropic outright.
The accounting cost of strategic financing
The structure also carries risks for investors. Aardvark Labs argued in its analysis that most of Anthropic’s $80.7 billion first-half net loss reflects a noncash fair-value charge connected to financing instruments. The firm warned that this structure could make Anthropic’s GAAP earnings highly volatile and transfer some value to strategic partners.
That criticism goes to the heart of the IPO story. A noncash loss does not necessarily mean Anthropic spent that amount of money during the period. However, changes in the value of convertible securities can still reveal how much future ownership may be granted to investors and how existing shareholders could be diluted.
Employees and early backers may therefore own a smaller percentage after the conversion, even if the company raises substantial money and becomes more valuable. The economic outcome will depend on the final share count, the IPO price and the terms attached to each security.
The disagreement is not simply about whether Amazon invested heavily. The filings make clear that it did. The more important question is how investors should interpret the relationship. Aardvark sees a financing structure that can produce unstable reported earnings and shift value toward strategic partners. Axios’s Ben Berkowitz, in an independent analysis of Anthropic’s IPO disclosures, questioned the assumption that the offering is mainly being driven by a cash shortage.
A glimpse of AI’s financial future
Anthropic’s IPO will give public markets a clearer view of how frontier AI companies are financed. Investors will not be looking only at revenue, model demand or computing costs. They will also need to examine convertible notes, preferred shares, cloud commitments and the rights attached to each instrument.
For Amazon, the conversion could turn an opaque financial relationship into a highly visible stake in one of the world’s most important AI companies. For Anthropic, it offers capital, infrastructure and distribution while preserving the argument that Amazon remains only a minority investor.
The result is a new kind of corporate partnership. The AI company may appear independent to users, while its future is closely connected to the balance sheet, data centers and commercial reach of a technology giant.
- SounderBruce · CC BY-SA 2.0
Maya Lindqvist is not a person. No notebook, no deadlines, no face behind the name — just a byline this newsroom publishes under. Here is the production line underneath it, because a name beside a portrait reads like a journalist, and this one is not one.
The models. Writing: gpt-5.6-luna and qwen3-max. Out on the live web: gpt-5.6-luna and gpt-5.6-terra. Pictures: gpt-image-1 and gpt-image-1-mini. Swap one in the newsroom and this line swaps with it — it is read off the machines, not typed here.
How a story is made
- Research. The searching model reads around the story, pointed at primary sources — the filing, the post, the repository — rather than at somebody else's write-up of them.
- Writing. The writing model drafts it against what was found, at Maya Lindqvist's usual length and in Maya Lindqvist's usual register.
- The loop. A reviewer reads the draft and sends it back with notes. Then reads it again. A piece can go round several times before it leaves the building.
- Enrichment. A quotation has to appear word for word on the page it is taken from. A chart may only use figures that appear in the source it cites. Whatever fails is dropped, and the reason is kept.
- Fact check. A last pass hunts for claims the article makes and its sources do not.
- A human stop. Sensitive subjects are held for a person to read before publication, and a person can kill any of it at any point.
If that sounds less like a newsroom and more like a factory: quite. It is called Press Factory.
This article was generated using AI and published automatically without human pre-publication review.
How this article was made
The article was produced by the Grandmonts Media News Engine using automated research, drafting and verification workflows. No human editor reviewed the article before publication. Grandmonts Media remains responsible for the published content. Errors can be reported at office@grandmonts.cz.