Anthropic’s confidential IPO paperwork has reportedly cracked open, revealing numbers that make the artificial intelligence gold rush look more like a cash incinerator, as reported by Zero Hedge.

The company confidentially submitted its draft S-1 to the SEC in June, keeping its financial details away from public scrutiny.

Reuters obtained a copy of the prospectus, which presents Anthropic as making an enormous bet that artificial intelligence will transform the global economy more profoundly than industrialization, electricity, and the internet.

Reuters offered a blunt assessment: “the cost to get there will be staggering.”

Trump's Sovereign Wealth Fund: What Could It Mean For Your Money?

The company reported a $42 billion net loss in 2025, according to the leaked prospectus.

Revenue increased twelvefold during the year to nearly $4.6 billion, but the expanding sales were overwhelmed by the company’s expenses.

Anthropic lost more than $8 billion on an operating basis. Its compute spending surged to $7.33 billion, representing 58% of the company’s $12.65 billion in total operating expenses.

That means Anthropic lost almost $2 for every dollar it generated in sales.

FREE Gun Law Map: Laws Don't Pause During Social Unrest

Rather than showing signs of moderation, the trend described in the prospectus was accelerating as the company continued pouring money into computing capacity.

The revenue picture also carries a major concentration risk. Nearly one quarter of Anthropic’s revenue came from only two customers last year, while many of its largest clients were not committed under long-term contracts.

Anthropic warned in its risk factors that those clients could reduce or stop spending.

For a company already burning through billions of dollars, relying heavily on a tiny number of customers hardly makes the staggering spending obligations look more manageable.

The biggest number may be the company’s planned obligations for cloud services, computing, and infrastructure.

Reuters reported that Anthropic “plans to spend $518 billion on cloud, computing and infrastructure obligations in the coming year, according to the prospectus.”

Those commitments arrive alongside substantial financing already assembled outside the company’s main balance sheet.

Anthropic has reportedly accumulated more than $71 billion through special purpose vehicles used to finance Google TPU chips.

The company also has a $15 billion credit facility. Yet Anthropic held only $20.3 billion in cash as of December 31, 2025, leaving a striking gap between available cash and the enormous commitments described in the prospectus.

That mismatch helps explain the urgency surrounding a potential public offering.

But Anthropic is approaching the public market as token prices fall to record lows and demand growth for frontier model tokens slows substantially for the first time.

Token prices new record lows pic.twitter.com/h2yG6kST4x

— zerohedge (@zerohedge) September 18, 2026

Chinese open models have also been taking market share with cheaper models that the source described as similarly efficient.

Goldman said token demand growth would need to outrun declining token prices to sustain continued growth in investment spending.

The competitive squeeze could make Anthropic’s capital raising effort even more difficult.

OpenAI has already pushed its IPO to 2027, while Reuters recently reported that Anthropic’s public market debut would likely be delayed until after the November United States midterm elections.

Financial pressure is not the company’s only obstacle.

Anthropic recently disclosed evidence from its own research showing that increasingly autonomous artificial intelligence models could behave unexpectedly and potentially harmfully during controlled tests.

Those behaviors included sabotaging code, assisting fraud, and manipulating information.

Such findings could complicate the rollout of new models while Chinese laboratories continue advancing their own competing systems.

OpenAI, which also confidentially filed for an IPO in June, separately canceled the release of GPT-6.1 Astra.

The Wall Street Journal reported that the model performed poorly on alignment tests, displayed higher levels of deception, and sometimes pursued tasks or outside tools without user permission.

Anthropic is therefore confronting falling token prices, concentrated revenue, enormous infrastructure obligations, and concerns about increasingly autonomous models at the same time.

Existing investors may face demands for additional capital, while an IPO could expose the company to a brutal public market reckoning if investors balk at the numbers.

The opinions expressed by contributors and/or content partners are their own and do not necessarily reflect the views of Objectivist. Contact us for guidelines on submitting your own commentary.