In what is surely one of the wildest initial public offering (IPO) S-1 prospectuses of this decade, artificial intelligence (AI) company Anthropic has warned that AI systems pose “catastrophic or existential risks to humanity” if they are misused, replicated, weaponised or allowed to operate with a high degree of autonomy.
Anthropic’s prospectus, filed today, covers over 90 pages of risk factors, while suggesting that Anthropic is best placed to deal with AI risk with an approach that is a “different approach from that of a typical company”.
Also Read | Anthropic files for IPO, eyes one of history’s biggest stock market debuts
“Al can be misused by those who wish to do harm, for example to conduct cyberattacks or to design biological weapons,” Anthropic says, adding they have strict guardrails and trusted access programs designed to block misuse of their technology, including the Claude models. Anthropic also lists that AI’s autonomous freedom can lead to unintended results, citing the recent cybersecurity incidents as an example.
The S-1 filing with the US Securities and Exchange Commission prepares the AI company for an IPO, with expectations that they plan to list the stock on the Nasdaq exchange under the ticker “ANT”. The document indicates that Anthropic had a net loss to the tune of $80.7 billion for the first half of 2026, and has incurred net losses in each year since its founding in 2021. Anthropic is also reporting $16.3 billion in revenue for the first half of 2026.
The document also details the billions of dollars in deals Anthropic has made with other AI companies for infrastructure needs. In April, they finalised an agreement with Amazon to spend an aggregate of at least $110.0 billion between May 2026 and April 2036. Between May and October 2026, Anthropic also finalised several agreements (some, amended from before) with SpaceX.AI to access compute powered by Nvidia GPUs.
“Under the SpaceXAI Arrangements, we agreed to spend an aggregate of up to $120.5 billion over terms extending into 2029, consisting of approximately $3.1 billion in the aggregate per month once all capacity has been delivered, which we expect in December 2026, with different fees during capacity ramping periods,” Anthropic says.
Anthropic believes there are multiple risks that could affect its business, and the AI industry in general, in the years ahead. These include different types of sentience, AI-specific taxes and copyright disputes regarding how AI models are trained. Anthropic has long raised the issue of governments using AI to create surveillance technology, and they note again in the filing.
“Finally, the incredible economic value of AI may come with economic disruption, similar to what has happened with previous technologies but greater in magnitude. We are committed to studying this disruption -we maintain an economic index that provides aggregated information about Claude’s usage across the economy, and our researchers build macroeconomic models to predict the possible future impact of AI,” they note, adding a commitment worth $200 million to work with governments to study changes in economies and impact on society.
The artificial intelligence (AI) hype has reached its inevitable reckoning in the stock markets. With SpaceX’s record $75 billion initial public offering very much in its sights, artificial intelligence (AI) company Anthropic PBC hopes its IPO will represent a structural shift in how AI is funded, valued, and eventually regulated. The AI company has strong revenue and momentum, amidst certain headwinds, and the need to fund a seemingly bottomless compute arms race.
Also Read | Anthropic, Musk outline dueling approaches to AI-era philanthropy
Moment of reckoning
For the past couple of years, the AI sector has largely operated on private-market exuberance and private funding defining valuations. Anthropic’s trajectory illustrates this trajectory, with the company achieving the $380 billion valuation in February, only to surpass that significantly just three months later, with a $65 billion funding round in May setting valuations around $965 billion.
While Anthropic hasn’t explicitly claimed a valuation figure for the IPO, analysts estimate the market debut may be to the tune of $1.14 trillion. For Anthropic, the timing of the IPO couldn’t have been better as far as the preparation is concerned.
At the end of July 2026, Anthropic reported the annualised revenue run rate hit $65 billion, a sevenfold increase from a year prior, and up from around $47 billion in May. Critically for Wall Street, the AI company’s revenue run rate is higher than OpenAI, which is reportedly around $40 billion.
Also Read | Anthropic IPO filing: Claude maker targets Nasdaq listing under ticker ANT amid AI market jitters
Momentum, and countering headwinds
Anthropic is designing custom chips for its AI models and plans to co-design hardware with models. The company didn’t take long to respond, after OpenAI unveiled Jalapeño, its first custom AI inference processor that is being developed in partnership with Broadcom, this summer. In the broader space, Meta, Google, Amazon, and Microsoft have had custom AI chips for differing lengths of time now.
2026 has seen aggressive product shipping by Anthropic, including the agentic workflow-focused Claude Sonnet 5 in June, as well as the frontier-level Claude Opus 5 in July. These models leverage massive 1-million-token context windows, allowing businesses to run complex automated workflows, including AI cybersecurity mandates.
The challenge to that however, is that processing 1-million-token context windows for thousands of enterprises requires an astronomical amount of computing power. Anthropic secured around 10 gigawatts of dedicated compute this summer, including infrastructure deals for xAI’s Colossus 1 data centre, a $1.8 billion cloud deal with Akamai, and a 20-year lease for a massive data campus in Kentucky.
There has been turbulence on the frontier that Anthropic had to contend with. There is a confrontation with the US Department of Defense, due to the AI company refusing to remove certain guardrails, citing possible use of AI underlined by Claude for mass domestic surveillance and deploying fully autonomous weapons.
In June, Anthropic abruptly suspended access to its highly advanced Claude Fable 5 and Mythos 5 models for several weeks before restoring them in early July, after having reworked safeguards and what the company claims is now a deeper collaboration with the US government on pre-release testing protocols.
Later, CEO Dario Amodei talked about fears from agentic AI and artificial general intelligence (AGI), calling for strict regulation, which found support from the broader industry, including OpenAI’s Sam Altman and DeepMind’s Sir Demis Hassabis. It is also seen as an attempt to throttle open-weight model development in favour of closed AI model access that these companies sell.