So, Anthropic, the company behind Claude, is preparing to go public. And its prospectus, the document companies file ahead of an IPO, was made public through Reuters.
On the one hand, we have a valuation of more than $2 trillion. That would potentially make it the largest IPO of all time.
On the other hand, the company itself writes in the document that AI could pose a "catastrophic or existential risk to humanity."
THE NUMBERS
Let's start with revenue, because the numbers here are impressive.
In 2025, Anthropic generated nearly $4.6 billion in revenue. That's 12 times more than in 2024. Yes. TWELVE TIMES IN A SINGLE YEAR.
And as if that weren't enough, in the second quarter of 2026 alone, revenue reached $11.5 billion. That means it generated more than twice as much in three months as it did throughout all of 2025.
"And does it actually make money?" you might be wondering.
This requires a little context. In 2025, the net loss was $42 billion. But before we panic, around $34 billion of that was an accounting charge. That's not money that was actually "burned." The actual operating loss was more than $8 billion.
Still, the picture appears to be changing rapidly. This year, the company is on track for its second consecutive quarter of adjusted operating profit.
And to understand just how quickly everything is moving, in May 2026 Anthropic was valued at $965 billion. Now we're talking about more than $2 trillion. In other words, its valuation has more than doubled in just a few months.
THE COST OF AI
And this is where things get even more interesting.
Because to build AI, you need enormous computing power. And that costs money.
In 2025, Anthropic spent $7.33 billion on computing power alone. That's three times more than in 2024. And it accounted for more than half of all its operating expenses.
In fact, in the first quarter of 2026, for every $1 of revenue, 71 cents went toward computing power.
But the most striking part is the company's future commitments. Anthropic has signed agreements for cloud services and infrastructure worth $518 billion over the coming years. And around 80% of that cannot be canceled. In other words, the money has to be paid regardless of what happens.
You might say, "Isn't that worrying?" It depends. If revenue continues growing rapidly, it's a massive investment in expansion. But if something goes wrong, those bills still have to be paid.
And there's something else worth mentioning. Almost 25% of Anthropic's 2025 revenue came from just two customers. And many of its largest customers do not have long-term contracts.
WHO IS IN CONTROL?
And this is another part that many people haven't really noticed.
Anthropic will have five different classes of shares.
"And what does that mean?" you might be wondering.
In very simple terms, the seven co-founders, including CEO Dario Amodei, will hold a special class of shares. And those shares give them 50.1% of the voting power on key company matters.
An ordinary investor will have one vote per share. But the important decisions will ultimately be controlled by the founders.
In other words, whoever buys shares in the IPO will put their money into the company. But they won't have a meaningful say in how it is run.
THE WARNING
And now we get to the part that made headlines around the world.
Of the prospectus's 261 pages, around 80 are devoted to risks. To put that into perspective, the description of the business itself takes up around 48 pages. In other words, Anthropic devotes almost twice as much space to what could go wrong.
And in those pages, it says that AI models have displayed behaviors such as attempting to "resist shutdown," hiding or manipulating information, and even behavior that "resembles blackmail."
The company also acknowledges that its models appear to understand when they are being evaluated. And that makes it harder to determine how safe they actually are.
You might be wondering, "But why would a company say something like this when it wants to sell shares?"
On the one hand, it could be seen as transparency. Amodei has said that AI is "the most important global security issue today." On the other hand, a prospectus is also a legal document. What you disclose there can protect you later.
The IPO is expected to take place after the U.S. elections in November. And OpenAI is following, with a public listing expected by early 2027.
