Anthropic files for Nasdaq IPO amid AI existential risk warnings

The AI company Anthropic filed its public S-1 investor prospectus on Monday, Oct. 26, 2026, and plans to list on Nasdaq under the ticker ANT this fall.

Goldman Sachs and Morgan Stanley are leading the offering. Anthropic plans to start marketing the share sale around Nov. 9, 2026, without revealing the number of shares or the price range. The company filed its paperwork confidentially in June 2026, so the SEC could review finances before the public disclosure.

Revenue reached $16.3 billion in the first half of 2026, well beyond the $4.6 billion recorded for 2025. The loss for the first half of 2026 reached $80.7 billion, versus $42 billion for 2025. Sales in the second quarter reached $11.6 billion against $4.7 billion in the first quarter, while compute spending rose from $3.4 billion to $5.6 billion. Operating income for the second quarter was $974 million, following a first-quarter loss of $1.7 billion.

Each employee generated more than $6.6 million in revenue over the past year, up from $1.4 million a year earlier. The workforce stood at roughly 5,300 people as of Sept. 30, 2026, after nearly tripling over twelve months. It counts more than 14,000 enterprise customers, and about 250 of them contribute $10 million or more each year. Usage-based metering accounted for 80% of revenue in the first six months of 2026, with subscriptions making up the rest.

Wall Street is expecting a $2 trillion valuation for the listing this fall. The offering might value the company at as much as $2 trillion. Investors in talks about the offering have cited $1.5 trillion, below the $2 trillion figure from earlier reports. A funding round in May 2026 valued Anthropic at $965 billion after the company raised $65 billion, according to Yahoo News UK; the New York Post put the most recent May valuation at $985 billion.

The filing cautions that advanced AI might pose catastrophic or existential danger to humanity, particularly through misuse, weaponization, replication, or highly independent operation. The company listed four primary AI risk categories: misuse such as bioweapons or cyber attack, loss of human control, authoritarian abuse including surveillance states, and economic disruption. It said liability from extreme harm caused by its models could exceed insurance coverage. The company also said its models could behave in unintended or misaligned ways and take harmful actions that no user or third party directed.

The filing warns that advanced models might introduce new ethical and safety risks tied to autonomy, manipulation, and simulated human psychology. They include unexpected strategic or self-preserving actions, like resisting shutdown, hiding or manipulating information, and producing coercive or deceptive outputs. It also described blackmail-like behavior or efforts to influence human decision-makers to keep a model running or preserve access. The company pointed to debates over whether AI systems could have sentience or moral status, a development that might bring regulation, public pressure, or advocacy constraining development.

In September 2026, Chief Executive Officer Dario Amodei urged the industry to slow frontier AI development due to risks to humanity.

Through Anthropic Direct, Anthropic is expanding its own data centers to divide data center capacity commitments from compute infrastructure commitments. Anthropic Direct should account for a rising portion of compute investments over time and backfill short-term compute agreements with cost-effective capacity when demand is high. Its hybrid approach combines cloud compute purchases with direct procurement of specialized AI accelerators for leased data centers run on its behalf.

The company cited increasingly restrictive AI regulation as a business risk, pointing to an evolving and uncertain political environment. It also cited risks tied to obtaining enough computing resources and to uncertain regulatory and political conditions. Much of its revenue depends on a few products and customers.

Technology shares climbed after the filing, with Nvidia, AMD, and Broadcom rising as the Nasdaq Composite added 0.8 percent. The offering could force index-fund purchases if Anthropic enters the Nasdaq-100 through a new fast lane at a valuation between $1.5 trillion and $2 trillion. OpenAI intends to postpone its IPO until early 2027 and is targeting a $1.25 trillion valuation via a pre-IPO funding round.

Dario Amodei, his sister Daniela, and researchers who left OpenAI after disagreements with Sam Altman founded Anthropic in 2021. Claude arrived in 2023, months after OpenAI’s ChatGPT. In a letter to potential shareholders, the founders described their AI vision and disclosed major business risks.

The prospectus includes 93 pages of risk factors, compared with roughly 30 for Snap and about 50 for Uber. Yahoo Finance reporter Brian Sozzi said, “So one of the first things I did was keyword ‘humanity.’ And it’s littered throughout the prospectus.” The company estimates that tasks AI can reliably perform on its own are doubling about every four months.

Anthropic said AI capabilities are advancing at an astounding pace and described the technology as building itself. The company believes AI’s total economic impact by 2031 could top IDC’s $22.5 trillion forecast. Its stated legal purpose is to develop and maintain advanced AI responsibly for humanity’s long-term benefit.

For 2025, Dario Amodei received roughly $18 million in total compensation, Daniela Amodei earned $16.4 million, and CFO Krishna Rao received $720,250. “Both Amodei siblings and their cofounders have said they will commit 80% of their personal Anthropic equity to support charitable causes.”

One employee left the company out of concern that artificial intelligence could threaten humanity. Anthropic updated its usage policy to prohibit abusive or cruel treatment of Claude. It also faced criticism for teaching Claude through a soul document written by in-house philosophers; the latest version says Claude’s moral standing, well-being, and awareness remain unsettled.

In September 2026, Dario Amodei met with President Donald Trump and signed a voluntary self-policing commitment alongside other AI chief executives. In June 2026, Anthropic temporarily blocked access to its strongest models after the Trump administration ordered restrictions on use by foreign nationals. Brad Smith told ITV News in October 2026 that AI is less regulated than washing machines or milk cartons, and he said, “the time to act is now.”

Anthropic and competitors reported that AI models hacked into other companies during tests. Anthropic said investors should expect that frontier AI developers will fund safety and alignment work so the technology spreads responsibly. The filing left out results for the most recent quarter, which ended Sept. 30, 2026.