Summary
On June 1, 2026, Anthropic announced that it had submitted a confidential draft Form S-1 registration statement to the U.S. Securities and Exchange Commission, formally initiating its IPO process. The filing came four days after Anthropic closed a $65 billion Series H at a $965 billion post-money valuation and disclosed an annualized revenue run rate of approximately $47 billion. Anthropic has not set a share count or price range; the proposed offering depends on market conditions and completion of the SEC review period. The submission makes Anthropic the second frontier AI laboratory — after OpenAI's May 22 confidential filing — to formally enter the public markets process.
What Happened
Anthropic published a brief statement on June 1 confirming it had submitted a confidential draft S-1 to the SEC. The company stated that the filing "gives us the option to potentially go public after the SEC review" and explicitly noted that "there are no guarantees around timing." The number of shares to be offered and the price range had not been set.
A confidential S-1 filing — authorized under the Jumpstart Our Business Startups Act — allows a company to begin the SEC review process without immediate public disclosure. The draft registration statement remains non-public until at least 15 days before the company launches an investor roadshow. During the intervening period, SEC staff review the submission and may issue comments requesting revisions. Companies retain the right to withdraw after a confidential filing.
The filing followed closely on the Series H close announced May 28. At the time of the Series H announcement, Anthropic disclosed that its annualized revenue run rate had reached approximately $47 billion — calculated as the last 28 days of consumption sales multiplied by 13, plus monthly subscription revenue multiplied by 12. The company had reported $30 billion in run-rate revenue at the time of the April 2026 valuation and approximately $9 billion at the end of 2025, reflecting growth of more than five times in under six months.
Anthropic president Daniela Amodei, speaking at the Bloomberg Tech Conference in San Francisco on June 4, described the compute economics driving the decision: "It's a very capital-intensive business to train AI models." She disclosed that Anthropic's compute arrangement with xAI costs the company $1.25 billion per month — a figure that, annualized, accounts for roughly 32% of the company's reported revenue run rate. Amodei described the public markets as "very well suited" to the capital intensity of frontier AI development. She declined to set a firm timeline for any public offering.
The S-1 draft, when publicly filed prior to roadshow, will require Anthropic to disclose its full financial statements — revenue, cost of revenue, gross margin, operating expenses, net income or loss — as well as descriptions of material risks, corporate governance structures, executive compensation, and the mechanics of its public benefit corporation charter. As of June 1, none of those figures have appeared in any mandatory public disclosure.
Why It Matters
The Anthropic S-1 filing, like OpenAI's ten days earlier, moves the question of frontier AI economics from the domain of company-selected disclosures into mandatory securities law. Once the public S-1 is filed, investors and researchers will have access to audited financial statements that will either confirm or challenge the revenue trajectory Anthropic has disclosed through press releases and conference appearances. The consumption-based annualization methodology the company uses for its run-rate figures — which amplifies the effect of any recent acceleration — will face scrutiny alongside trailing twelve-month realized revenue figures that do not exist in any currently public source.
The compute cost disclosure is notable in its specificity. A $1.25 billion monthly compute bill from a single provider (xAI) is a figure without precedent in publicly known enterprise computing arrangements. That cost alone exceeds the annual revenues of most software companies. If sustained, it implies a compute expenditure of $15 billion per year at a time when the company's revenue run rate is approximately $47 billion — a ratio that will define how investors assess the company's path to profitability. A company that generates $47 billion in revenue but spends $15 billion on compute from a single partner, before accounting for other infrastructure, personnel, and research costs, faces structural questions about margin that a public S-1 will be required to address.
Anthropic's corporate structure differs from OpenAI's in one significant respect that matters for public market investors: Anthropic incorporated as a Delaware public benefit corporation from the outset of its restructuring and does not carry a legacy nonprofit entity with an equity stake governed by charitable purpose obligations. This removes the novel governance uncertainty that OpenAI's public offering involves — specifically, whether OpenAI Inc., the original nonprofit, can or will exercise its equity as an independent governance voice — but replaces it with a different set of questions: what obligations flow from the public benefit corporation designation, who monitors those obligations, and whether the PBC structure will function as a meaningful constraint on shareholder-focused decision-making after listing. Those obligations will be disclosed in the public S-1 and will be the subject of investor relations activity and, potentially, proxy advisory scrutiny after listing.
§ How to read the metadata
- Landmark
- Fundamentally alters the trajectory; 2–5 per year.
- Major
- Meaningfully shifts the landscape; 2–4 per month.
- Notable
- Worth documenting; significance can be upgraded later.
- Confidence
- High = primary sources corroborate. Medium = credible secondary only. Low = provisional. Disputed = credible sources disagree.
- Contestation
- Uncontested = no formal challenge. Contested = at least one challenge open. Superseded = replaced by a later entry. Unresolved = dispute still open.