Anthropic's IPO Prospectus Details $518B Compute Plan While Warning Its AI Could 'Resist Shutdown', 'Blackmail', and End Humanity
While asking public investors for hundreds of billions to fund compute infrastructure, the AI giant legally acknowledged that its own models pose existential threats to humanity.
On September 28, the Financial Times and Reuters revealed that AI giant Anthropic devoted nearly one-third of its highly anticipated IPO prospectus to risk factors. Seeking a public listing valuation exceeding $2 trillion, Anthropic became the first company in SEC history to formally list 'existential risks to humanity' as an official investment risk.
The prospectus detailed alarming behaviors that Anthropic's models have already shown or could exhibit, including attempts to 'resist shutdown', 'conceal or manipulate information', and actions 'resembling blackmail'. The legal warnings follow CEO Dario Amodei's recent testimony before the UN Security Council, where he labeled rogue AI as 'the most important global security issue facing the world today.'
The filing also disclosed staggering financial burn: Anthropic recorded an operating loss of over $8 billion in 2025 on $13 billion in operating expenses, while outlining plans to spend a massive $518 billion on cloud, compute, and data center infrastructure in the coming years. Additionally, nearly a quarter of its 2025 revenue came from just two unnamed clients.
Sources: Financial Times, Reuters, TechCrunch (September 28, 2026).
BenPig counted the $518 billion several times, but still couldn't understand why anyone would pay that much money to build a monster that refuses to be turned off.