(Oct 26): Anthropic PBC filed for an initial public offering (IPO), warning its technology carried existential risks even as it moved a step closer to being the first company focused primarily on developing the most advanced artificial intelligence (AI) models to go public.

The company behind Claude expects to begin trading on the Nasdaq Global Select Market under the symbol ANT, a filing on Monday with the US Securities and Exchange Commission showed. Anthropic filed confidentially in June for a listing. 

The IPO is being led Goldman Sachs Group Inc and Morgan Stanley, along with 20 other banks, the filing showed. Anthropic and some of its shareholders are offering stock in the IPO, according to the filing. The amount of shares they are offering will be disclosed in a later filing.

The company’s debut has become a focal point for the world’s hopes and fears over AI, which run the gamut from worries over job displacement and the substantial upfront spending the technology needs to other, more existential questions: Anthropic’s prospectus has flagged various “risks to humanity” including attempts to “resist shutdown” and “conceal or manipulate information”.

Anthropic had an US$80.7 billion (RM329.7 billion) net loss on revenue of US$16.3 billion in the first six months of 2026, versus a US$13 billion net loss on revenue of US$1.2 billion in the same period last year, the filing showed.

The company has mainly focused on serving the enterprise market. Anthropic’s net revenue retention, a measure of its ability to keep its existing customers and increase their spending over time, is about 900%, the filing showed.

Anthropic said 79% of its revenue came from customers outside of its top five customers for the third quarter of the year.

Anthropic is seeking to tap the public market’s ample funding capacity to maintain its lead over rivals, as AI companies burn through hundreds of billions of dollars to continue advancing their so-called frontier models. 

It is set to begin formally marketing for its IPO as soon as the week of Nov 9, Bloomberg News has reported. Anthropic expects to match or beat the size of SpaceX’s record-setting IPO, people familiar with the matter have said.

The company and its rival OpenAI are also facing mounting public scrutiny over AI safety after a string of high-profile hacking incidents by rogue agents, ramping up concern about the potential impact on Anthropic’s market debut.

Anthropic emphasised the potential for its AI to “behave in unintended, misaligned, or autonomous ways” in the filing. If its technologies are not developed with adequate safeguards and oversight, it could “pose catastrophic or existential risks to humanity, particularly if such systems were misused, replicated, weaponised, or allowed to operate with a high degree of autonomy”.

Consumption-based revenue rose from US$288 million, and 75% of the company’s revenue, for the year that ended Dec 31, 2024 to US$3.8 billion, and 83% of the total, for last year. Information technology made up US$13.1 billion, and 80% of revenue, between January and June of this year. For Anthropic, consumption-based revenue refers to customers that pay for its AI models as they use them, based on the amount of tokens and the type of AI model they use.

The company’s total indebtedness stood at US$6.2 billion as of June 30, and it said it could raise more debt going forward to meet its financing needs.

Anthropic said it has a revolving credit facility of up US$15 billion, which it increased in size from US$3.25 billion. Revolving loans work much like a credit card, allowing a company to borrow, repay and borrow again up to a set limit over the life of the facility. They are commonly used by companies to provide financial flexibility and ready access to liquidity when needed. The revolver will mature in 2031, the company said.

It added that, after obtaining an investment-grade credit rating from at least two of the three major credit rating agencies, the revolving loan could become unsecured.

The company spent US$8.9 billion on computing power and infrastructure in the six months that ended June 30, up 283% from the year prior. The company noted that these expenses declined as a percentage of revenue, seemingly responding to concerns about the pace of spending.

Just since the end of 2025, Anthropic entered into new agreements for computing power with minimum obligations of about US$417 billion. These agreements will generally start this year or next, with terms as long as 20 years.

Anthropic’s compute economics have improved as revenue has scaled. Compute and infrastructure spending was more than 10 times revenue in 2023, falling to 653% in 2024, 160% in 2025 and just 55% in the first half of 2026. In the second quarter alone, it fell further to 48% of revenue — meaning Anthropic was spending about 48 cents on compute and infrastructure for every dollar of sales.

Goldman Sachs touted its role in the coveted lead-left position in marketing documents sent to clients, after securing the same role for the record-setting listing of SpaceX earlier this year, a document seen by Bloomberg showed. Morgan Stanley, its rival, is the so-called stabilisation agent, a sought-after role that would see it engage in transactions that reduce volatility in the stock, the filing showed.

Uploaded by Tham Yek Lee