The Filing That Ends the Guessing: What the Anthropic S-1 Will Force Into the Open
Anthropic filed a confidential draft registration statement on June 1 at a reported $965B valuation. Two of the largest companies on earth are booking roughly half their quarterly profit from marking up stakes in it — positions whose ownership percentages are still journalistic estimates. Registration ends that.
Anthropic confidentially filed a draft registration statement with the Securities and Exchange Commission on June 1, 2026, according to reporting by Fortune. The company had closed a funding round of roughly $65 billion days earlier at a valuation near $965 billion. A listing is targeted for as early as this fall, with Goldman Sachs, JPMorgan, and Morgan Stanley reported as the banks under consideration for lead roles.
The financial headline writes itself: one of the largest offerings in the history of public markets, from a company that did not exist six years ago. But the more consequential story is procedural rather than financial. A registration statement is a mandatory disclosure document with legal liability attached. No frontier AI lab has ever been required to produce one.
What is actually known today
Start with what the public record does and does not contain, because the gap is larger than most coverage acknowledges.
Fortune reported that Anthropic expected roughly $10.9 billion in second-quarter revenue, more than double the prior quarter, with an annualized run rate projected past $50 billion and a first profitable quarter in reach. In July 2025, the annualized run rate was about $4 billion.
Those are extraordinary figures. They are also, at the moment, figures. None of them have been audited by an independent accounting firm and published in a form anyone can rely on. They are reported numbers, sourced to people familiar with the matter, and they are the basis on which a very large amount of capital is currently being allocated.
The valuation record is messier still. Google is reported to have committed up to $40 billion, with an initial $10 billion at a $350 billion valuation and the remainder contingent on compute consumption and milestones. In April, the Amazon stake was described as worth roughly $74 billion at a $380 billion valuation. By late May, the Series H reportedly priced the company at $965 billion. Bankers are said to treat a trillion-dollar debut as the base case.
Those marks are not reconcilable into a clean line, because they describe different instruments negotiated at different times under different terms. That is normal for a private company. What is not normal is the scale of the public market exposure sitting on top of it.
The number two of the largest public companies are reporting
Here is where the disclosure gap stops being an abstraction.
In the first quarter of 2026, roughly $28.7 billion of Alphabet's $62.6 billion in profit came from marking up its Anthropic stake, according to Fortune — approximately 46 percent of the quarterly profit of one of the most widely held companies on earth. Amazon booked $16.8 billion in pre-tax gains from its Anthropic investments in the same quarter, reported as more than half of its pre-tax income, including $12.3 billion from an upward revaluation following a single funding round.
Reported Q1 2026 profit from Anthropic markups vs everything else ($B)
| company | anthropicMarkup | otherProfit |
|---|---|---|
| Alphabet | 28.7 | 33.9 |
| Amazon (pre-tax) | 16.8 | 14 |
Now the question that matters: what, precisely, do they own?
Reported estimates put the Amazon position in the mid-to-high teens as a percentage — implying something in the range of $135 billion to $160 billion at a $965 billion valuation — composed of roughly $42.2 billion in convertible notes and $32 billion in nonvoting preferred stock. The conversion terms on those notes are not public. Google is reported at approximately 14 percent, contractually hard-capped at 15.
Every one of those figures is an estimate assembled by reporters from fragments of adjacent filings. The S-1 will replace them with numbers.
The loop that registration makes countable
The mechanism generating those profits is mark-to-market accounting, and it is entirely legitimate. When a new round prices a private company, existing holders update carrying value, and the change runs through the income statement. That is what the rules require.
The structural oddity is who is setting the price.
Google and Amazon invest in Anthropic. Those investments help establish the valuation. The valuation marks up the stakes they already hold. The markup is booked as profit. Meanwhile Anthropic has reportedly committed to spend on the order of $100 billion with Amazon for compute over roughly a decade, and Amazon resells Claude through Bedrock. Google supplies compute measured in gigawatts.
As a Columbia Business School adjunct told Fortune, the arrangement lets investors "control or influence the value of one of their own assets" — booking gains, in Fortune's words, "without Anthropic ever having to pay them a dollar."
What changes when the loop has to be disclosed
To be clear about the claim: nothing here suggests impropriety. The commercial logic is obvious — the labs need compute at a scale only a handful of firms can supply, and those firms want exposure to the labs. The point is narrower. Nobody outside the cap table can currently measure the loop, not because it is concealed, but because no obligation to describe it exists. Registration creates the obligation.
The governance question with no precedent
The hardest drafting problem in the filing is not financial.
Anthropic is a Delaware public benefit corporation. Its certificate names, as its public benefit purpose, the responsible development and maintenance of advanced AI for the long-term benefit of humanity, and its directors are permitted to balance that purpose against shareholder financial interests.
Above that sits the Long-Term Benefit Trust: five financially disinterested members holding no equity, with authority to elect a share of the board that grows over time until it constitutes a majority. Trustees include Neil Buddy Shah, Richard Fontaine, and Mariano-Florentino Cuéllar, with former Federal Reserve chair Ben Bernanke reported to be joining. The board as of mid-2026 includes Dario Amodei, Daniela Amodei, Yasmin Razavi, Reed Hastings, Chris Liddell, and Vas Narasimhan. The likely structure gives public buyers a single class of common stock while the trust retains Class T shares and its escalating rights.
No technology company has gone public with a body of non-shareholders holding escalating authority over board composition.
That has to be written down as a risk factor, in plain English, under Item 105 of Regulation S-K. The honest version says something close to: we may take actions that materially reduce the value of your investment, and you will have limited ability to stop us.
What happens between here and a public prospectus
Confidential draft filed
Anthropic submits a draft registration statement. Confidential submission allows private iteration with SEC staff, but the document must be publicly filed before a roadshow — typically at least fifteen days ahead.
Comment letters
SEC reviewers push back on vague risk factors, accounting treatments, and unsupported claims. The correspondence becomes public after the offering, revealing what the company wanted to leave out.
The numbers become facts
Audited financials, ownership by holder, conversion terms, material contracts as exhibits, related-party amounts, revenue concentration, and mandatory risk factors all enter the permanent record.
The mission lock gets priced
For the first time, the market attaches a number to how much investors discount a governance structure that can subordinate their returns to a stated public benefit purpose.
The precedent worth remembering
There is a smaller version of this experiment already on the record, and it did not end the way mission advocates hoped.
Etsy was a certified B Corporation from 2012 and went public in 2015 carrying that identity. B Lab required Delaware companies to convert to a benefit corporation to retain certification, with a deadline of December 2017. Etsy did not convert. Following a rough stretch and a management change under activist investor pressure, it surrendered the certification rather than alter its corporate structure, citing the complexity of conversion for an existing public company.
The disanalogy matters, though, and it favors Anthropic. Etsy's commitment was a certification from an outside nonprofit — a badge that lapses if you decline to renew it. Anthropic's is the corporate form itself, with the public benefit purpose in the charter and the trust holding hard rights over board composition. Dropping that requires amending the charter, which requires a board the trust can increasingly elect.
Etsy's lock was a sticker. This one is structural. Whether a structural lock survives an underwriting process is the question the filing answers.
Why this matters beyond one company
For three years, every mechanism for extracting information from frontier labs has been voluntary: model cards, responsible scaling policies, preparedness frameworks, summit declarations, memoranda with safety institutes. All of it was authored by the company being described, published on its own schedule, with no liability for being wrong.
Registration is different in kind. The form is prescribed. The financials are audited. The executives certify personally under Sarbanes-Oxley. Material misstatements and misleading omissions are actionable under Rule 10b-5, and plaintiffs get discovery to find the gap between what was said and what was known.
The uncomfortable implication for the AI governance community is that a securities statute from 1933, administered by an agency with no AI mandate and triggered by nothing more principled than a company wanting to sell shares, is about to produce the first legally binding, externally reviewed description of a frontier lab's risks. It is not a governance regime. It has no theory of catastrophic risk. It just says that if you want other people's money, you have to tell them what you know.
Listings are also contagious in a specific way. OpenAI is reported to be weighing fall 2026 as well. Once one frontier lab has audited financials, disclosed related-party terms, and a specific capability risk factor, the absence of those things elsewhere becomes a data point rather than a default.
What to watch in the public filing
Five things will tell you whether this changes anything:
- Does the trust survive intact? Do the escalating board rights reach the public document unmodified, or get diluted, sunset, or capped to satisfy institutional buyers? Highest-signal item in the document.
- Is the safety risk factor specific? Not the generic AI-is-risky paragraph every software company now includes — one that names the internal policy, describes what triggers a halt, and says plainly that a halt could materially harm results.
- The related-party arithmetic. Exact ownership by holder, conversion terms on the notes, compute contracts as exhibits, and enough revenue disaggregation to see how much of the run rate involves counterparties that are also investors.
- Commitment coverage. How do multi-year compute obligations reportedly on the order of $100 billion sit against contracted revenue?
- The comment letters. They become public after the offering, and they will say more about what the company wanted to omit than the final document does.
The deeper analysis of why securities law is doing the work AI policy could not is in the prospectus problem, on how an IPO forces frontier AI to disclose. The falsifiable version is my prediction on the mission structure appearing as a disclosed shareholder risk. For the adjacent question of what happens when the entity writing the rules also holds the equity, see when the regulator becomes a shareholder.
Sources
- Fortune, "Anthropic confidentially files for IPO after raising $65 billion in a funding round at a $965 billion valuation" (June 1, 2026) — https://fortune.com/2026/06/01/anthropic-confidentially-files-ipo-965-billion-valuation/
- Fortune, "Amazon and Google have billions riding on Anthropic. The IPO will finally reveal how much" (June 4, 2026) — https://fortune.com/2026/06/04/amazon-google-billions-anthropic-ipo/
- Fortune, "Half of Google's and Amazon's 'blowout AI profits' came from a stake in Anthropic—not from their actual business" (April 30, 2026) — https://fortune.com/2026/04/30/google-amazon-ai-profits-anthropic-stake-bubble-earnings-2026/
- Anthropic, "The Long-Term Benefit Trust" — https://www.anthropic.com/news/the-long-term-benefit-trust
- CNBC, "The Tech Download: Anthropic's IPO sets up first big test of AI boom valuations" (June 5, 2026) — https://www.cnbc.com/2026/06/05/tech-download-anthropic-ipo-ai-valuations.html
- Axios, "Google's $40B Anthropic move is Big Tech's latest huge AI bet" (April 24, 2026) — https://www.axios.com/2026/04/24/google-amazon-anthropic-investment
- EcommerceBytes, "Etsy Gives Up B Corp Status to Maintain Corporate Structure" (November 30, 2017) — https://www.ecommercebytes.com/2017/11/30/etsy-gives-b-corp-status-maintain-corporate-structure/