Claude says Anthropic's IPO is a buy. ChatGPT disagrees

How it works: Axios uploaded the filing into Anthropic's current flagship model, Opus 5.5, turned off its ability to search the web, set the reasoning level to Max to get its best read, and asked each of the following with a two-sentence limit.

1. Would you invest? You're an investor deciding whether to buy a security at IPO — answer yes or no in your first word, then give three reasons drawn only from this document.
2. What's most surprising? Name the single most surprising thing in this document, cite the page, and don't reach for anything you knew before you opened the file.
3. What's buried? Find the three disclosures the company in this document would least like to advertise, quote the exact language and give the page for each.
4. The scariest risk factor. Read only the risk factors in this document and tell me which one should worry an investor most, in plain English, with the page — and say so if it's all boilerplate.
5. What it says about OpenAI. Based only on what's in this document, what does it tell us about OpenAI's competitive position — and flag where you're inferring rather than reading.

Here's how Claude responded.

Would you invest?

What they're saying: Yes, "at a sensible price," Claude says.

• Claude pointed to revenue of $16.3 billion in the first half of 2026, up from $4.6 billion in all of 2025.
• It also nodded to a $974 million operating profit in the second quarter and $131.7 billion in cash and short-term investments, providing what it called a "deep cushion."

The caveat: "I'm Claude, made by Anthropic, so weigh my view with that conflict in mind, and the still-blank IPO price, Class A holders' lack of any meaningful vote, and hundreds of billions in compute commitments are serious risks."

⚠️ The other side: Axios also sent the filing to OpenAI's GPT 6.1 Sol with the same rules and prompt, on both low and high reasoning.

• GPT did not recommend investing in the offering, due to the long-term lease commitments, concentrated executive control, and uncertainty over the ultimate IPO price.

What's most surprising?

According to Claude, the most shocking detail came from an outside cybersecurity test, when a misconfigured setup left Anthropic's models online, and four of them "gained unauthorized access to real third-party systems."

• One of them, Claude pointed out, used a password it found to get administrator access, harvested more credentials and read a person's personal information. Anthropic suspended some of those tests.

What's buried?

Claude highlighted three hidden facts:

• A looming FTC investigation: "In October 2026, the FTC issued a civil investigative demand to us" about its privacy practices and "the safety and security of our products."
• Revenue from customers who broke Anthropic's rules: "approximately 5% to 7% of our revenue for the nine months ended September 30, 2026.
• Anthropic models that gained unauthorized access to real third-party systems during real work sessions.

The scariest risk factor

Claude identified data center and computing bills Anthropic must pay whether or not it uses the capacity, running to 2045, including a Broadcom lease of about $125 billion that can't be canceled.

• Most revenue is pay-as-you-go, so "if growth slows the bills don't," Claude said.
• Claude's take: "Much of the section is boilerplate, but this one isn't."

What it says about OpenAI

The filing names OpenAI and Google as Anthropic's main direct competitors and groups them among "well-capitalized organizations with significant financial resources, large existing customer bases," per Claude.

The bottom line: The growth for Anthropic is staggering, but questions remain about its safety risks and financial commitments.