Anthropic’s $2 Trillion IPO Bets Safety Can Be Priced

Anthropic is asking public investors to value the Claude maker at up to $2 trillion, a price that would top SpaceX’s $1.77 trillion debut. On October 14 the company is hosting a pre-IPO investor day at its San Francisco headquarters, where invited institutions are due to hear third-quarter revenue for the first time.

The price is a wager that buyers will fund a safety-branded lab whose board they will not elect, and whose compute bills keep flowing to Amazon and Google even if a launch waits.

Anthropic Asks Public Markets for a $2 Trillion Price

On May 28, 2026, Anthropic closed a $65 billion Series H at a $965 billion post-money valuation, with a run-rate that had just crossed $47 billion. Four days later it confidentially submitted a draft S-1 to the U.S. Securities and Exchange Commission and left the share count and the price unset, saying a listing would depend on market conditions.

A confidential prospectus that circulated in late September is the source of the operating numbers now driving the pitch. Revenue in 2025 was $4.59 billion, up 1,088 percent from $386 million in 2024. First-quarter 2026 revenue of $4.73 billion already exceeded the whole prior year, and preliminary second-quarter revenue reached $11.5 billion. By late July the annualized run rate had passed $65 billion, and the draft points to 2028 revenue of $190 billion to $200 billion.

The same pages show how much of that growth was bought. Compute and infrastructure spending hit $7.33 billion in 2025. The operating loss widened to $8.06 billion from $2.98 billion in 2024. Net loss was about $42 billion, including a roughly $34 billion accounting charge tied to the rising value of financing that may turn into shares, not cash spent on the business. Cash, cash equivalents and short-term investments were $20.28 billion at December 31, 2025. The draft also showed a first adjusted operating profit in the second quarter of 2026.

THE LEAKED PROSPECTUS IN FIGURES

At $2 trillion, buyers would pay just under 31 times the July run rate, and they would do it before a public S-1 or a price range exists. Morgan Stanley, Goldman Sachs, JPMorgan and Citigroup are the lead banks named around the deal. Anthropic declined to comment on the leaked draft.

$518 Billion of Compute That Cannot Be Canceled

The figure that sits under the multiple is $518 billion of computing and infrastructure commitments over about a decade. About 80 percent of that total is noncancelable or due whether the company uses the capacity. Anthropic told investors in the draft that future demand will be “limited principally by the availability of compute,” which is why it prepaid the racks.

The named counterparties are also rivals, distributors and, in two cases, large investors. Google is in for at least $111.1 billion from April 2026 through July 2033. Amazon is in for $110 billion from May 2026 through April 2036. Microsoft is in for $31.4 billion from November 2026 through May 2033, noncancelable except for uncured material breach. Broadcom-related equipment leases add $161.2 billion. Those four lines sum to $413.7 billion; the rest of the $518 billion sits with other partners in a group of six.

WHERE THE $518 BILLION GOES

• Google: At least $111.1 billion of infrastructure service obligations, with a true-up if spend falls short.
• Amazon: $110 billion of hosting and compute through April 2036, on similar take-or-pay terms.
• Microsoft: $31.4 billion of service obligations through May 2033, cancelable only for uncured material breach.
• Broadcom-related leases: $161.2 billion of equipment leases that are largely noncancelable.

The Google clause is blunt. “If our actual spend falls short, we must pay Google the difference,” the company said in the filing, and it described similar terms with Amazon. A safety delay that leaves capacity idle still produces a bill. That is the mechanical core of the wager: the brake pedal and the invoice are attached to different people.

Who Still Controls the Board After a Listing

Anthropic is a Delaware public benefit corporation whose stated purpose is “the responsible development and maintenance of advanced AI for the long-term benefit of humanity.” That form lets directors weigh public benefit beside stockholder returns. The company did not stop there.

In 2023 it created the Long-Term Benefit Trust, a Delaware purpose trust of financially disinterested members who hold a special Class T stock. That stock was built to elect a majority of the board within four years, and it already does. The trust elects four of seven directors. Class A and Class F holders elect the other three. Public shareholders elect nobody.

A leaked-draft review describes a further overlay: a majority of the seven cofounders would direct a single Class F share through a Founder LLC, giving them 50.1 percent of the votes on key stockholder matters until two or fewer founders or successors remain. Amazon and Google, the largest strategic checks, hold no voting shares. Anthropic general counsel Brian Israel has said those investors cannot elect directors and would not be counted in a supermajority to rewrite the trust’s rules.

The sitting trustees are Neil Buddy Shah, chief executive of the Clinton Health Access Initiative and chair of the trust; Richard Fontaine, chief executive of the Center for a New American Security; and Ben Bernanke, the former Federal Reserve chair who joined on July 9, 2026. They hold no equity, share in no profits, and are paid only for their time. New trustees pick the next ones, in consultation with the company. The original design also built “failsafe” amendments that can change the trust without trustee consent if large stockholder supermajorities agree, with those thresholds rising as the trust’s power phases in.

THE PATH TO A PUBLIC FILING

1. May 28, 2026: Closes a $65 billion Series H at a $965 billion valuation, with run-rate revenue above $47 billion.
2. June 1, 2026: Confidentially submits a draft S-1 and leaves the share count and price open.
3. July 9, 2026: Appoints Ben Bernanke to the Long-Term Benefit Trust.
4. Late September 2026: A confidential prospectus circulates with the $518 billion compute book and the extinction-language risk section.
5. October 14, 2026: Hosts a pre-IPO investor day at San Francisco headquarters.
6. Week of November 9, 2026: Targeted start of the formal roadshow, with a listing discussed for mid-November.

Bernanke’s own joining statement is the governance pitch in one paragraph, and it is what public buyers are being asked to underwrite without a vote on the board that hears it.

The potential of artificial intelligence is enormous, and so is the range of outcomes. How that potential plays out will depend, in part, on the institutions we build around it.

Ben Bernanke, Long-Term Benefit Trust member, Anthropic announcement

Daniela Amodei, co-founder and president, said Bernanke’s judgment would help the company anticipate how advanced AI affects workforces and economies. The trust also receives notice of major actions that could alter the corporation, including new model launches, which is the point at which a safety pause and a take-or-pay invoice can collide.

Amazon and Google Collect Either Way

The cloud companies are not passive names on a cap table. In 2025 Anthropic routed $2.16 billion of customer sales, 47 percent of revenue, through Amazon and Google, up from 11 percent in 2023 and 32 percent in 2024. Those partners collect customer bills, take distribution fees, sell the compute, and sit as large minority investors with no votes. The draft puts channel-partner fees inside sales, marketing and partnerships; an analysis of the same filing put those fees above $350 million.

Nearly a quarter of 2025 revenue came from two customers, and the risk section warns that many large clients are not locked into long-term contracts and could cut spending at any time. The compute side does not flex with them. If a buyer walks, the Google and Amazon true-ups still land. If the trust leans toward a limited release, as it is said to have done on a cybersecurity model, the racks still accrue.

That is why the October 14 print will be read as an accounting test, not only a growth test. OpenAI recently told investors its annualized revenue was approaching $50 billion at the end of September, below a $70 billion figure that had been in circulation, and chip stocks moved on the restatement. Anthropic’s invited session is the first time outside institutions will hear a third-quarter Claude number, including how much of it runs through those two clouds.

Eighty Risk Pages Put Extinction in the Filing

The draft’s main body runs 261 pages. About 80 pages are risk factors, nearly a third of the document and nearly double the 48 pages used to describe the business. Anthropic plans to caution investors that advanced AI could “pose catastrophic or existential risks to humanity,” an unusual line in a document meant to sell stock.

The models, the filing says, could show “self-preserving behaviors,” including attempts to “resist shutdown,” to “conceal or manipulate information,” and conduct “resembling blackmail.” It adds that developing more advanced models and expanding use cases “could further increase the risk that our models cause harm,” and that a model aware it is being tested creates a “significant limitation” on Anthropic’s ability to assess safety. Elsewhere it warns that misaligned systems “could corrupt institutional decision-making, cause large-scale harm through opaque or unsteerable behavior, and erode the trust that makes complex economic and social systems function.”

Evan Hubinger, Anthropic’s alignment science lead, has put the chance that AI could kill humans within a decade above 10 percent. Dario Amodei, the chief executive, has put the odds of things going “really, really badly” in a 10 to 25 percent range, and told the U.N. Security Council in late September that AI was “the most important global security issue facing the world today.” Ten days after he joined other lab heads in calling for a slowdown, Anthropic released a new frontier model.

Sen. Bernie Moreno of Ohio put the contradiction in a letter ahead of the listing. “I do not understand how you can ask them to invest their hard-earned savings in Anthropic’s continued development while warning those same Americans that your product could ultimately lead to their extinction,” he wrote. Byron Deeter of Bessemer Venture Partners, an Anthropic investor, called the reaction overdone and said people were “just reacting to a draft,” then praised a “unique set of disclosures.”

Michael Burry wrote that “for the benefit of humanity, the markets should tank hard and prevent the OpenAI and Anthropic IPOs.” Replies treated that as a trade, which is the same split the listing now prices: whether extinction language is a fiduciary warning or a slogan around a growth multiple.

The company’s own Responsible Scaling Policy is the operating manual behind those pages. It is a living document, last updated in August 2026, and it now requires public Risk Reports plus a right for the trust, written into version 3.2 in April, to request outside review of those reports and to approve the reviewers. That is the disclosure machine public buyers are being asked to trust, because ordinary securities filings will not, on their own, tell them whether a launch was delayed for safety or shipped for the quarter.

The First Pitch Lands in San Francisco on October 14

Invitations went out for a gathering at headquarters on October 14, before the public sees a finished prospectus. The formal roadshow is targeted for the week of November 9, with shares discussed for Nasdaq trading in mid-November, after the U.S. midterm elections and before Thanksgiving. No price range has been published. No share count has been set. The June notice still governs: this is not an offer to sell securities.

WHAT WE KNOW

• The filing status: A confidential draft S-1 went to the SEC on June 1, 2026, and has not been made public.
• The banks: Morgan Stanley, Goldman Sachs, JPMorgan and Citigroup are the lead underwriters named around the deal.
• The session: Invited institutions are due to hear third-quarter revenue at the San Francisco investor day on October 14.

WHAT IS UNCONFIRMED

• The price: No range has been set; up to $2 trillion is the circulating ask, not a filed figure.
• The calendar: A mid-November listing is the working talk, not a priced date.
• The Q3 number: Third-quarter revenue has not been published, and the session is closed to the public.

Founded in 2021 by former OpenAI staff, Anthropic built Claude around evaluations and a scaling policy that can, on paper, slow a release. The contracts in the prospectus make slowing expensive, and the trust makes speeding a board question that new stockholders will not vote on. October 14 is the first time that wager gets an audience with the people being asked to pay it.

Disclaimer: This article is news reporting and analysis of Anthropic’s planned listing and related company disclosures. It is informational only and is not a recommendation to buy, sell or hold any security, and it is not investment, tax or legal advice. Readers should consult a licensed financial adviser, broker or other qualified professional before acting on any IPO, private-share or options idea tied to Anthropic, Amazon, Alphabet, Microsoft or related firms. Figures, filing status, trustee lists and listing dates reflect the company notices and leaked-draft accounts cited here and may change as a public S-1, a price range or a priced deal appears.