Anthropic began its investor roadshow on November 10 with terms that would make it one of the largest listings in US history: 825 million Class A shares at $70 to $78 each, raising up to roughly $62.5 billion, trading on Nasdaq under the ticker ANT. Pricing is expected November 18, with trading the following day. At the top of the range the fully diluted valuation reaches about $1.4 trillion.

That is well below the $2 trillion figure circulating in September, but roughly 45% above the $965 billion valuation from the May Series H round. Within a day of the filing, bankers were reportedly discussing pricing above the range, which would push the number toward $1.5 trillion.

The suppliers are also the buyers

Alongside the public offering sits a $26 billion private placement at the IPO price, in non-voting Class N stock. Google and Nvidia have each committed roughly $10 billion, AMD about $2.5 billion, with Broadcom, Micron, Microsoft and TSMC at $1 billion or less apiece.

Read that structure twice. The companies writing the checks sell the chips and cloud capacity that Anthropic buys. The capital goes out, and a large share comes back as future procurement. For the suppliers this is a reasonable way to lock in their biggest AI customer. For anyone trying to size independent demand, it makes the two buckets hard to separate.

What the money is actually for

Anthropic has told investors it plans to commit roughly $900 billion to compute and infrastructure over the coming years. Even at the high end of a $100 billion raise, that covers about a ninth of it. The rest has to come from revenue, debt and partners.

Which puts the whole story on the revenue curve. The company reported third-quarter revenue of about $17 billion, up roughly 46% sequentially, with an operating loss of about $567 million. Annualized revenue reached roughly $81 billion by October, and revenue in the first nine months of 2026 was about seven times the whole of 2025.

For comparison, OpenAI told investors its third-quarter revenue rose about 50% to $10 billion, against an adjusted operating loss of about $10.9 billion.

One of these companies is close to operating break-even; the other is still burning more than a dollar for every dollar it earns. Both, however, are shifting toward multi-year enterprise contracts and prepaid compute, which stretches the gap between when revenue is recognized and when cost lands. Single-quarter margins are becoming a poor lens.

What this means if you are not an investor

Three practical takeaways.

First, list prices are unlikely to spike. The first quarter after listing is a scorecard, and raising prices is the fastest way to lose the customers you just won. Layered pricing, cache discounts and batch endpoints are more likely than headline increases.

Second, supplier entanglement deepens. When Google and Nvidia are simultaneously shareholders and vendors, the negotiating position of a mid-sized customer depends on how much the vendor needs the anchor accounts.

Third, do not read valuation as capability. $1.4 trillion is a bet on future cash flow, not evidence about your task. The three numbers that matter in a model comparison are still pass rate on your own workload, cost per thousand calls, and how fast you get help when something breaks.

The listing is a public price on AI infrastructure, not a technology milestone. It will shape compute pricing for the next two years. It will not decide whether your particular problem gets solved.