Anthropic has put a price on its IPO: $70 to $78 a share for 825 million Class A shares. The amended prospectus shows what that buys. Public investors get about 6% of the shares and roughly 5.5% of the general vote. Meanwhile, seven of the company's chip, memory and cloud partners commit $26 billion alongside them.
Anthropic launched its roadshow on November 10, filing a second amendment to its S-1 registration statement with the Securities and Exchange Commission the same day. The headline numbers are large. At the top of the range, the 825 million shares would be worth $64.35 billion. CNBC calculated that the company could be valued at up to $1.405 trillion on a fully diluted basis, up from the $965 billion valuation it announced in May. Citing people familiar with the plans, CNBC also reported that Anthropic intends to price the deal on November 18 and begin trading on Nasdaq under the ticker "ANT" on November 19.
The price range was the main blank in the earlier drafts of the prospectus. The more lasting story is in the rest of the amendment. It shows who is underwriting Anthropic's expansion and how little control a public shareholder will have over the company.
What the offering actually contains
According to the filing, Anthropic is selling 801,351,354 new shares, and existing stockholders are selling another 23,648,646. Anthropic gets no proceeds from the stockholders' shares. The underwriters also have a 30-day option to buy up to 123,750,000 more shares from the company.
At $70, $74 and $78, the shares Anthropic sells itself would raise about $56.1 billion, $59.3 billion and $62.5 billion before underwriting fees (VibeCodAI calculation: shares offered by the company multiplied by price). The company estimates that, at the $74 midpoint, combined net proceeds from the IPO and a concurrent private placement will be about $84.6 billion. That rises to about $93.7 billion if the underwriters exercise their option in full.
The prospectus states that once the offering is complete, new investors will hold about 6.3% of the Class A and Class N shares outstanding. Existing holders will keep about 91%.
The $26 billion that comes from Anthropic's own supply chain
The private placement is the most revealing part of the amendment. Seven companies have agreed to buy $26.0 billion of non-voting Class N stock at the IPO price, contingent on the offering closing. At the $74 midpoint, the filing allocates the money as follows:
- Google: about $10.0 billion
- Nvidia: about $10.0 billion
- AMD: about $2.5 billion
- Broadcom: about $1.0 billion
- Micron: about $1.0 billion
- Microsoft: about $1.0 billion
- TSMC Partners: about $500 million
Each of these companies sells Anthropic something it needs to train and run Claude: accelerators, custom chips, memory, cloud distribution or the fabrication behind them. At the midpoint, the placement accounts for roughly 30% of the new money Anthropic raises in this transaction, before fees (VibeCodAI calculation: $26.0 billion divided by $26.0 billion plus $59.3 billion).
Some of that money was committed earlier. The filing says Google's purchase funds "a portion of its previously committed investment obligations" under a March 2026 note purchase agreement. AMD's $2.5 billion counts against an equity commitment of up to $5.0 billion tied to a July 2026 hardware supply agreement for AMD Instinct MI450-based Helios rack systems. Once the placement closes, Anthropic can draw only up to $2.5 billion more from that commitment. At least $12.5 billion of the $26 billion therefore comes from deals made before the IPO, now settled in public-company stock.
Anthropic's ties to Broadcom go further. The prospectus describes a March 2026 lease arrangement for specialized accelerators, with payments that total approximately $125.2 billion excluding interest. The lease cannot be cancelled during its term except in a default. The filing also says Anthropic can issue up to $42.0 billion of convertible notes to Broadcom or a designated financial partner.
Investment from suppliers is not new in AI. Chipmakers and cloud providers have been investing in the labs that buy their products for years. Anthropic's arrangement is unusually visible and unusually large, and it is now attached to a listed security. That cuts both ways. The suppliers' money signals confidence, and it helps lock in the capacity Anthropic needs. It also means that a meaningful share of the demand for Anthropic's stock comes from companies whose revenue depends partly on Anthropic's spending. Public buyers should treat that demand differently from the independent price discovery a book of fund orders provides.
Why the buyers need the money: compute
The financial statements explain why the company is raising so much. Revenue for the nine months ended September 30, 2026, was $33.2 billion, up from $2.5 billion a year earlier. Compute and infrastructure expense rose 329% to $18.8 billion, which is 57% of revenue. A year earlier it was 172%. Over the same nine months, the operating loss shrank to $1.26 billion from $5.22 billion.
That is a real improvement in operating economics. The headline net loss of $104.0 billion for the period tells a different story, and most of it is not cash. It is driven by a $94.3 billion charge for the change in fair value of convertible notes, derivative liabilities and SAFEs. The filing attributes changes of this kind largely to increases in the company's own valuation. A rising stock price makes Anthropic's convertible obligations to investors more expensive on paper. Investors comparing per-share losses should read the operating line, not just the bottom line.
As of September 30, Anthropic reported $95.6 billion in cash and cash equivalents and $36.3 billion in short-term investments. It said it intends to use the new proceeds for general corporate purposes, "including investments in compute and infrastructure."
What a Class A share does not buy: control
Anthropic is a public benefit corporation. Its stated purpose is "the responsible development and maintenance of advanced AI for the long-term benefit of humanity." The amendment sets out a share structure that puts that purpose, and the founders, ahead of outside shareholders on most questions of control:
- Class A (the shares sold in the IPO): one vote per share.
- Class N (held mainly by strategic partners, including the supplier placement): no vote except where the law requires one. These shares generally convert to Class A when sold to an unaffiliated third party.
- Class F: a single share held by an LLC owned by Anthropic's seven co-founders, including Dario and Daniela Amodei. It carries a fixed 50.1% of the total voting power on matters put to stockholders generally, including the election of directors who are not Class T directors. The filing says future share issuances will not dilute it.
- Class T: held by the Long-Term Benefit Trust, which the filing describes as a body of "financially disinterested trustees." It generally lets the trust elect and remove the Class T directors, who will be a majority of the board.
- Class E: issued to employees other than the co-founders. It votes only on limited matters and has no economic rights.
Here is what that means for the public float. Class A holders together control the 49.9% of general voting power that Class F does not. The 825 million IPO shares are about 11.1% of the roughly 7.43 billion Class A shares the filing expects to be outstanding after the offering and placement. That gives IPO buyers about 5.5% of the general vote, before any greenshoe shares and before any future Class N conversions (VibeCodAI calculation: 825,000,000 ÷ 7,427,991,791 × 49.9%). The trust separately chooses most of the board. Anthropic says it will be a "controlled company" under Nasdaq rules, which exempts it from some of the exchange's governance requirements.
None of this is hidden, and multi-class structures are common among founder-led technology listings. Anthropic's version is distinctive because control is split between founders and a mission trust rather than held by founders alone. The company presents that split as a safeguard for its public-benefit purpose. For an investor, it means their influence goes through the share price, not through the ballot.
Counterpoints and open questions
There is a reasonable bullish reading of the same documents. Revenue grew roughly thirteenfold year over year in the nine-month comparison. Compute costs fell sharply as a share of revenue. The supplier placement can be read as the companies with the best view of AI demand putting money behind Anthropic at the public price, not at a discount. CNBC also reported, citing people familiar with the deal, that the book was already covered across the range on an allocable basis.
Several things are still unknown. The final price will not be set until pricing. CNBC's sources say Anthropic is aiming for about 20% retail allocation, well above the 5% to 10% typical of IPOs. The prospectus itself refers only to "potential allocation to retail investors of a minority of shares." The filing also describes staged lock-up releases. The longest restrictions, up to 365 days, apply to the co-founders and certain large holders. Those schedules will shape how much supply reaches the market in the first year.
What to watch
The first signal is where the deal prices against the $70 to $78 range, and whether the underwriters exercise their option. After that, the questions will be whether compute spending keeps falling as a share of revenue once the Broadcom lease and AMD capacity come into service, and how the Class N stock held by suppliers moves into the voting float over time. The roadshow is selling a frontier AI company. The prospectus describes something more specific: a company whose largest suppliers are also its shareholders, and whose public owners will have a small say in how it is run.
Sources: Anthropic, "Anthropic launches roadshow for proposed initial public offering" (November 10, 2026); Anthropic, PBC, Amendment No. 2 to Form S-1 (filed November 10, 2026); CNBC, "Anthropic sets IPO price range of $70 to $78" (November 10, 2026).