We have followed this story from the confidential S-1 through the November timeline. On November 10 it stopped being rumor. Anthropic published a short notice launching the roadshow for its proposed IPO, and for the first time we have hard terms.
This post is based only on that notice. I have not read the preliminary prospectus, so anything about revenue, margins or risk factors is out of scope here. Derived figures are my own arithmetic and are labeled as such.
What the notice actually says#
- Offering size: 825,000,000 Class A shares in total.
- Split: Anthropic sells 801,351,354 shares; existing stockholders sell 23,648,646.
- Greenshoe: underwriters get a 30-day option on up to 123,750,000 additional shares.
- Price range: $70.00 to $78.00 per share.
- Listing: applied for Nasdaq Global Select Market, symbol “ANT”.
- Bookrunners named: Goldman Sachs and Morgan Stanley.
- Status: the registration statement has been filed with the SEC but is not yet effective, so nothing can be sold yet.
The notice also repeats the mission line: Anthropic is a public benefit corporation focused on “responsible development and maintenance of advanced AI for the long-term benefit of humanity.”
The arithmetic (mine, not Anthropic’s)#
Multiplying the stated share counts by the range:
| Item | At $70 | At $78 |
|---|---|---|
| Company’s primary shares (801.35M) | ~$56.1B | ~$62.5B |
| Selling stockholders (23.65M) | ~$1.66B | ~$1.85B |
| Full 825M base deal | ~$57.8B | ~$64.4B |
| Greenshoe (123.75M) | ~$8.7B | ~$9.7B |
Gross figures, before underwriting fees. If the greenshoe is fully exercised, the company’s share of a deal this size would climb well past $60B at the top of the range. I cannot derive a valuation from this notice, because it does not state how many shares will be outstanding after the offering. Anyone quoting you a market cap today is either reading the prospectus or guessing.
One more thing the notice does not say: how Anthropic plans to use its proceeds. It states only that the company receives nothing from the selling stockholders’ shares. For the use of proceeds you need the prospectus.
Why a developer should care about a stock offering#
I’m a Claude partisan, and I think the IPO is good news for people building on Claude Code: a company with tens of billions in fresh capital can keep funding compute, safety research and the kind of long-horizon agent work that makes spec-driven development viable. But enthusiasm is not a vendor-risk policy. Three practical consequences:
1. Public-company disclosure is a feature for buyers. Once ANT trades, quarterly reporting will expose margins, concentration and compute commitments that today we only see in leaks. Procurement teams that have been asking “is this vendor durable?” get audited answers instead of vibes.
2. Pricing pressure will be visible. A listed company answers to quarterly numbers. Expect closer scrutiny of discounting and plan changes. We already saw fast-mode deprecations and tier shuffles arrive on short notice; budget for more of that, not less.
3. Governance claims will be tested. The structure (public benefit corporation, the Long-Term Benefit Trust, and the Public Benefit Fund) is part of why many of us trust Anthropic’s safety posture. The roadshow notice does not mention the trust or the fund; the details live in the prospectus. If your leadership cares about governance as a selection criterion, have someone actually read that section rather than relying on blog summaries, including this one.
A short checklist before pricing#
None of this requires you to change tools. It requires you to make your dependence explicit:
- Write the spec, own the spec. The strongest hedge against any vendor is that your requirements live in versioned markdown, not in a chat history. If a model, price or policy changes, your specs port.
- Pin and test model versions. Run your acceptance suite against whichever model your pipeline uses so a deprecation is a diff, not a crisis.
- Route through a gateway if you are large. Claude Code’s gateway and model-catalog settings let you control which models a team can reach and cap spend centrally.
- Read the prospectus risk factors yourself. Export controls, government designations and compute concentration are the items that can actually interrupt service.
What we still don’t know#
- The final price. The range is a marketing instrument; pricing happens after the book is built and can land outside it.
- Whether the deal upsizes, and whether the greenshoe is exercised.
- The share count after the offering, and therefore the valuation.
- Use of proceeds, lock-up terms and the dual-class mechanics, all of which sit in filings I did not review.
I’ll revisit this once the deal prices and ANT starts trading. Until then, treat every number above as an offering term, not a result.
Sources#
- Anthropic, “Anthropic launches roadshow for proposed initial public offering”, November 10, 2026
- Anthropic, “Anthropic files S-1 with the SEC for proposed initial public offering”, October 26, 2026 (listed on the newsroom; not read for this post)




