The first pure-play AI software company to hit public markets is already causing collateral damage before pricing a single share.

The Summary

The Signal

This is the inflection point where AI moves from infrastructure bet to direct exposure. Anthropic's 240,000-word prospectus lays bare what it costs to stay competitive in frontier models: not millions or billions, but hundreds of billions in locked-in commitments before you know if the revenue will materialize. The math is stark. At current revenue pace, it would take 14 years to cover the compute bill.

More than 11,900 customers now spend at least $100K annually, driving revenue per employee to $6.6M, triple the typical SaaS company. But here's the uncomfortable part: 5% to 7% of 2026 revenue came from customers violating usage rules, primarily through model distillation. That's not a rounding error when you're asking for a trillion-dollar valuation.

"The filing shows $265.6 billion in computing bills it must pay even if AI demand slows."

The governance structure tells you everything about who this IPO is really for. The Long-Term Benefit Trust and founders control voting through Class B and Class T shares, leaving public shareholders with economic exposure but zero influence. Insiders can sell up to 10% of locked shares on day one, an unusual provision that signals either confidence in sustained demand or awareness that the window might close.

Goldman and Morgan Stanley are leading the offering, with roadshow marketing starting the week of November 9. The timing matters. Chip stocks rose 2.4% on the filing, and even copper prices rallied as traders bet on AI infrastructure demand. But the Financial Times noted that valuing Anthropic on historic numbers is "a fool's errand" given the step-function changes in model capability and market size ahead.

The risk factors read like a sci-fi screenplay. Anthropic warns its models could pose existential risks to humanity, a disclosure that might normally tank a prospectus but here signals the scale of what's being built. The company paused frontier model training over security fears, suggesting internal safety debates are real, not PR theater.

CEO Dario Amodei was paid $18M last year, standard for a company scaling this fast. More interesting: Anthropic predicts AI will surpass human intelligence soon and control computers directly, with Nobel-level breakthroughs expected by 2030. Wall Street remains skeptical.

The derivative markets moved faster than traditional finance. MetaMask now offers Anthropic pre-IPO perpetual contracts via Hyperliquid, while OKX offers 10x leveraged bets on the valuation. Leveraged ETFs have already been filed. This is crypto's answer to access: tokenize exposure before shares even price.

The Implication

If Anthropic prices anywhere near $2T, it resets the market's willingness to fund compute-first business models with decade-long payback horizons. Every AI lab with a frontier model and a credible safety story will point to this as the comp. The 30% drop in Neocloud tells you the rotation is already happening. Investors are choosing picks-and-shovels or direct model exposure, not the messy middle.

Watch what happens to the committed compute numbers as deployment scales. Anthropic holds $10.8B in contracted future revenue, but the $637B in obligations dwarf that. Either revenue grows exponentially or those commitments get renegotiated. The filing also notes dependence on TSMC and HBM suppliers as central risks, meaning geopolitical supply chain disruption could crater model training before any existential AI risk does.

For crypto natives, this is the moment tokenized equity markets prove their speed advantage. Pre-IPO perps and leveraged ETFs gave retail traders exposure before institutional allocations even closed. That's the Web4 edge: bet on the future while TradFi is still reading the prospectus.

Sources

Crypto Briefing | Financial Times Tech | BeInCrypto | Protos | CoinTelegraph | The Defiant | Decrypt | CoinDesk