Anthropic’s IPO filing lays out the usual public-company material, including losses, customers and risk factors. It also contains several disclosures that are harder to file under ordinary boilerplate: the founders’ post-IPO voting power, costly executive protections, warnings about AI sentience, and commitments for enough computing capacity to make the numbers look typo-adjacent. The Verge, which reviewed the filed S-1, reported the document runs to more than 300 pages.
These are reported descriptions of Anthropic’s prospectus, not predictions that its stated risks will occur.
What does Anthropic’s IPO filing say?
The Amodeis’ pay package includes security and a private jet. The Verge reported that Chief Executive Dario Amodei received nearly $18 million in total compensation in the prior year, mostly stock and option awards. President Daniela Amodei received $16.4 million. The company also reported $2.75 million in personal-security services for Dario, and $1.29 million for Daniela’s personal protection, home security and transportation. Personal use of a private jet by Dario cost $97,055, according to the report. Anthropic says it depends heavily on its senior team, particularly the sibling co-founders.
The risk section contemplates catastrophe, loss of control and sentience. According to The Verge, Anthropic warns that AI systems could cause catastrophic harm beyond what the insurance industry could cover in aggregate. The filing also considers whether Claude or other AI systems could develop forms of sentience, and warns that systems may act in unintended, misaligned or autonomous ways. CNN separately reported that the filing groups AI risks into misuse, loss of human control, authoritarian abuse and economic disruption. Those are company warnings about possible outcomes, rather than evidence any has happened.
Compute commitments are measured in tens of billions. The Verge reported planned SpaceX payments of about $3.1 billion a month for access to Nvidia GPUs once the full capacity is delivered, potentially reaching $120.5 billion by 2029. Other reported commitments include $44.9 billion with Microsoft through 2033, at least $114.5 billion with Google through 2033, and $110 billion with Amazon over the next decade. These are commitments and plans, not a statement of money already spent.
A separate entity preserves founder influence. Anthropic’s seven co-founders would collectively hold 50.1% of shareholder voting power over key matters through an entity called Founder LLC, The Verge reported. In practical terms, public investors could buy shares without gaining control over those specified decisions.
Only two companies are named as direct rivals. The Verge reported that the prospectus identifies OpenAI and Google as Anthropic’s direct competitors. The narrow list does not mean Anthropic lacks other competitive pressures; it describes how the company framed direct competition in the filing.
Why the reported loss needs context
CNN reported that Anthropic posted an overall net loss of more than $80 billion in the first half of 2026. It said just under $9 billion was linked to AI infrastructure and computing, while more than $72 billion came from fair-value changes involving convertible notes, derivative liabilities and SAFEs. That accounting distinction does not erase the company’s infrastructure costs, but it does mean the headline loss was not all operating cash burn.
This story draws on original reporting from The Verge.
