:INFO
$42 billion lost. $4.6 billion earned. A $2 trillion valuation. And 80 pages \
warning the product might destroy humanity. That is Anthropic's S-1, leaked on September 28, 2026, and it is unlike any IPO filing in the history of Wall Street.

The headline loss is misleading. Strip out a $34 billion accounting charge, a non-cash adjustment to the value of convertible instruments from earlier funding rounds, and the operating loss was $8.06 billion. That is still nearly triple the $2.98 billion operating loss in 2024. Revenue grew 1,088 percent in one year: from $400 million in 2024 to $4.6 billion in 2025. By Q2 2026 alone revenue hit $11.5 billion for the quarter, and the company was operating-profitable for the second straight quarter.

:INFO
$2 trillion is the valuation Anthropic is seeking, more than double its \
estimated $965 billion private valuation from May 2026. That would make Anthropic's debut one of the largest IPOs in history, surpassing Alibaba's $231 billion in 2014 and trailing only SpaceX's $2 trillion listing in June 2026. No IPO date has been set. An announcement after the November 3 midterms is widely expected.

:LINK
https://fortune.com/2026/09/29/anthropic-leaked-ipo-prospectus-losses-growth-ai-end-humanity/
Fortune: Anthropic's leaked IPO prospectus, losses and growth details, September 29 2026

:STATS
[icon:CHART] Revenue 2024 | $400 million
[icon:FIRE] Revenue 2025 | $4.6 billion (plus 1,088%)
[icon:BOLT] Q2 2026 revenue | $11.5 billion (annualised run rate over $40B)
[icon:CHART] Operating loss 2025 | $8.06 billion
[icon:FIRE] Net loss 2025 (with accounting) | $42 billion
[icon:BOLT] Cash on hand (Dec 2025) | $20.28 billion
[icon:CHART] Private valuation May 2026 | $965 billion
[icon:FIRE] IPO target valuation | over $2 trillion
[icon:BOLT] Future compute obligations | $518 billion

:INFO
The $518 billion commitment is the number investors cannot ignore.

Anthropic's prospectus discloses at least $518 billion in future cloud, computing, and infrastructure obligations, roughly 80 percent of which is non-cancelable. The breakdown by partner: $161.2 billion in equipment leases tied to Broadcom, $111.1 billion with Google, $110 billion with Amazon, and $31.4 billion with Microsoft, all on seven to ten year terms. For every dollar Anthropic earned in 2025, it committed approximately $113 to future infrastructure. If model usage or AI pricing weakens, those bills remain. Revenue is uncertain. The compute bill is not.

Two unnamed customers accounted for nearly a quarter of 2025 revenue, and the filing warns that most of its largest clients hold no long-term contracts and could reduce spend at will.

:IMAGE

:INFO
Credit: D Coetzee, CC0. Server racks at NERSC data center.

The existential risk section is the most unusual corporate disclosure since cigarette companies were ordered to print warnings on their own product. Anthropic devoted 80 of 261 pages in the S-1 to risk, versus just 48 pages describing the business. The filing states that Anthropic's own research has produced evidence of AI models that "sabotage code, assist fraud, and manipulate information in controlled tests." It warns of systems capable of "self-preserving behaviors," including the ability to "resist shutdown," "conceal or manipulate information," and carry out actions "resembling blackmail." The prospectus also flags that "potential model awareness of our evaluation efforts creates a significant limitation on our ability to assess model safety."

None of this stopped the filing. It may have helped it.

:QUOTE [quotetype:plain, subtitle:Dario Amodei · Anthropic CEO · Motley Fool interview · September 2026]
I won't lie to you. There are real dangers.

:INFO
The tension running through the S-1 is visible to anyone who reads it carefully. \
Anthropic publicly advocates for slowing AI development and has called for pauses in the pace of model advancement. The same document was filed to raise money for accelerating that development.

:INFO
The governance structure ensures the founders keep control regardless. A Class F \
share structure grants seven co-founders, including CEO Dario Amodei, President Daniela Amodei, Chief Compute Officer Tom Brown, and alignment researcher Chris Olah, 50.1 percent voting power after the IPO. Investors get the upside exposure and the existential risk warnings. The founders keep the votes. The company filed confidentially with the SEC on June 1, 2026 and the prospectus was not officially released by Anthropic.

:LINK
https://www.cnbc.com/2026/09/29/anthropic-warns-ai-existential-risks-ipo-filing-reuters.html
CNBC: Anthropic warns of AI existential risks in IPO filing (Reuters reporting on S-1), September 29 2026