Anthropic secretly files for october ipo in ai duel with openai
Anthropic, the startup behind Claude, has begun quiet conversations with Goldman Sachs, JPMorgan and Morgan Stanley about a blockbuster October listing that could raise more than $60 billion, sources familiar with the matter tell TechBloom. The move ignites a Wall Street sprint between the two most powerful language-model labs on Earth.
The race is on
Dario and Daniela Amodei, the siblings who walked out of OpenAI in 2021 to build a "safer" rival, are now shopping the same bankers their old employer is courting. Both companies want the same underwriters, the same October window, the same bragging rights as the first pure-play large-model maker to trade on Nasdaq. The difference: Anthropic already has a federal judge on its side.
Last Thursday the company won an emergency injunction blocking a Pentagon edict that labeled it a "substantial risk" to national supply chains. Defense officials had invoked a Cold-War-era authority normally reserved for Chinese telecoms, arguing that Anthropic’s refusal to let generative ai near mass surveillance or weapons guidance systems posed a threat to American security. The judge disagreed, calling the ban a potential multi-billion-dollar gut punch to U.S. innovation.
Valued at $38 billion in February after a $30 billion funding round co-led by Abu Dhabi’s MGX, Anthropic has burned through cash at hyperscale speed. It committed $50 billion to custom U.S. data centers stuffed with Nvidia H100s, Amazon Trainium and Google TPU v5p chips—hardware its investors supplied in deals worth tens of billions. The IPO proceeds would repay that silicon debt and bankroll the next cluster already under construction in Texas.
Why wall street can’t say no
Bankers smell the same pheromone that surrounded Google in 2004 and Meta in 2012: a proprietary moat built on transformer architecture and eye-watering compute scale. Anthropic’s pitch deck, glimpsed by two sources, claims Claude 3.5 already outperforms GPT-4o on SWE-bench coding tasks while hallucinating 30 % less. The kicker: enterprise customers in finance and healthcare pay triple the API price for that reliability, pushing annualized recurring revenue past $1.2 billion in May—up 400 % year-on-year.
Yet the prospectus must also disclose the Pentagon flare-up. National-security clauses could allow future administrations to yank export licenses or cloud credits, a regulatory asterisk that never haunted OpenAI. Convincing mutual-fund managers to look past that risk is why Goldman and JPMorgan are demanding a 5 % greenshoe and a dual-class share structure that cements the Amodeis’ voting control.
No final board vote has been taken. Markets could convulse, regulators could pounce, the rivals could yet merge instead of list. But the paperwork is circulating, the NDAs are signed, and the October clock is ticking. The first ai duopoly to ring the opening bell will set the valuation benchmark for every transformer startup still trapped in venture-capital limbo. Anthropic just grabbed the inside lane.