Anthropic Files Confidentially for IPO at a $965 Billion Valuation

Photo by Aerps.com on Unsplash

Most coverage of the Anthropic IPO frames it as a victory lap for the Claude maker. It isn't. Anthropic filed its draft prospectus while suing the U.S. Department of Defense, while burning $1.25 billion a month on a single data center contract, and while sitting on a valuation that nearly tripled in roughly twelve weeks. The filing is less a celebration than a forced move.

According to CNBC, Anthropic submitted its confidential S-1 to the Securities and Exchange Commission on June 1, 2026, in what the network described as setting up a potentially historic share sale. TechCrunch reported the same day that share count and pricing have not been set, and that the confidential route lets Anthropic begin the process without immediately publishing financials, risk factors, or voting-power structure.

The 12-Month Revenue Curve That Forced This Filing

The numbers are the story. Fortune reported that Anthropic's annualized run rate was $4 billion in July 2024. CNBC put the May 2026 figure at $47 billion, up from $10 billion in annual revenue the prior year. That is more than a 12x expansion in under two years, and it is the kind of curve that public-market investors will pay a steep premium to access before it flattens.

The near-term forecast is steeper still. Citing Bloomberg reporting, Fortune noted that Anthropic expects $10.9 billion in Q2 2026 revenue, more than doubling quarter-over-quarter, and has told investors its annualized run rate will pass $50 billion by the end of July 2026. The same reporting indicates the company is on pace for its first profitable quarter.

A private company posting those figures has roughly two options. Keep raising mega-rounds at ever-larger valuations from a shrinking pool of investors who can write nine-figure checks. Or go public and tap the deepest capital market on the planet. The S-1 is the second option.

How the Anthropic Valuation Got to $965 Billion in 90 Days

The Anthropic valuation arithmetic is worth reading slowly. Fortune reported that the company was valued at $380 billion in February 2026 and $965 billion by May 2026. That is a $585 billion increase in roughly three months, larger than the entire market capitalization of all but a handful of public companies.

The vehicle for that re-rating was the Anthropic Series H funding round. TechCrunch reported that the $65 billion raise was co-led by Altimeter Capital, Dragoneer, Greenoaks, Sequoia Capital, Capital Group, Coatue, and D1 Capital Partners, and that it closed less than a week before the IPO filing. Seven co-leads is unusual. It signals that no single firm could absorb the check size alone, and that demand had to be sliced across the largest crossover funds in the market.

That $965 billion mark is the headline number in any Anthropic vs OpenAI IPO comparison, because it tops the $852 billion post-money valuation OpenAI reached in March 2026 after raising $122 billion, according to TechCrunch. OpenAI is preparing its own offering. For a brief window, the smaller of the two frontier labs is the more valuable one on paper.

The $1.25 Billion-a-Month Bill Hiding Inside the Prospectus

Here is the line that did not make most headlines. CNBC reported that Anthropic signed a deal to use compute at SpaceX's Colossus 1 data center in Memphis, Tennessee, and that per SpaceX's own prospectus, Anthropic will pay $1.25 billion per month through May 2029. Multiply that out. It is roughly $15 billion a year, or about $45 billion across the contract's life, going to a single facility owned by a competitor's affiliated company.

That single commitment dwarfs almost every operating expense at any technology company in history at the same revenue stage.

It is not the only one. Fortune reported that Anthropic committed more than $100 billion to Amazon Web Services technology in April 2026 to train and run its Claude AI agent. Stack the two together and Anthropic has pre-committed compute spending that exceeds the GDP of most countries before a single retail investor sees the offering.

This is the part that public investors will need to underwrite carefully. The Anthropic Claude IPO filing arrives at a company whose cost structure looks less like a software business and more like a heavy-industry firm renting refineries. The margin profile depends entirely on whether inference prices fall faster than compute commitments compound. If they do, the math works. If they don't, the run-rate growth gets eaten by the data center bill.

The Pentagon Lawsuit That Could Reshape the S-1's Risk Section

The non-obvious risk in this filing is political, not technical. Fortune reported that Anthropic is in active litigation with the U.S. government after Secretary of Defense Pete Hegseth designated the company a supply-chain risk. Anthropic has sued to reverse the designation, and Fortune noted the outcome could jeopardize billions in revenue.

Few companies file confidentially while suing a cabinet department. The S-1 risk-factor section will have to disclose this in detail before the document goes public, which gives Anthropic a window to either resolve the dispute or frame it on its own terms.

The dispute appears to be downstream of an earlier safety decision. Fortune reported that Claude Mythos Preview was held back from release in April 2026 over concerns it was, in the publication's words, "too good at hacking into software," a safety-first move that also triggered conflict with the Trump administration. Read together with the Pentagon designation, the picture is of a lab whose safety posture and its largest potential customer are pulling in opposite directions. Public-market investors do not usually buy companies in open conflict with the federal government. They will here, but the discount or premium attached to that risk is the most interesting unknown in the entire offering. For ongoing coverage of how regulators are approaching frontier labs, see our AI Industry articles and AI Safety articles.

What a Confidential Filing Actually Buys Anthropic

A draft registration is not the same as an S-1 made public. TechCrunch explained that confidential filings let companies begin the IPO process without publicly disclosing financials, risks, or voting-power structure until they file a full S-1.

Three practical things flow from that.

• Anthropic can negotiate with the SEC on disclosure language for the Pentagon lawsuit and the SpaceX and AWS contracts before any of those terms hit the public record.
• The company can test pricing with underwriters and adjust the offering size without each revision becoming a news cycle.
• Anthropic controls the timing of the public reveal, which matters when the largest competitor in the category is preparing its own filing.

Think of it as a closed-door dress rehearsal before opening night. The audience eventually sees the show, but the company gets to fix the lighting first.

Why This Lands Differently Than the 1999 Internet IPOs

The lazy comparison is dot-com. It is also wrong. The 1999 cohort went public on traffic metrics and forward-looking narratives, often with single-digit-million revenue. Anthropic is filing with a $47 billion run rate, a path to profitability inside a single quarter, and customer contracts large enough to anchor a public utility.

The closer historical parallel is Google's 2004 IPO, which arrived at a company already printing real cash with a dominant position in a category that nobody else had figured out how to monetize at scale. Anthropic is not Google. The cost structure is far heavier and the moat is contested by OpenAI, Google itself, Meta, and an open-weights tier that keeps closing the quality gap. The point is that the financial profile is closer to a mature platform IPO than a speculative one.

The consumer signal matters too. CNBC reported that Claude jumped to the No. 1 slot on Apple's chart of top U.S. free apps in late February 2026. That ranking is meaningful because Anthropic spent years positioned as the enterprise-first, safety-focused alternative to ChatGPT, and the App Store leaderboard suggests the consumer flywheel finally caught. Readers tracking the broader category shift can follow our AI Tools articles for related coverage.

What to Watch When the Full S-1 Goes Public

Four things will determine whether the offering prices above or below the $965 billion private mark when the document is unsealed.

1. Gross margin disclosure. Run-rate growth is dazzling. The gross margin after compute costs is the number that will set the multiple. The SpaceX and AWS commitments will show up in that line.
2. Customer concentration. Frontier labs often have a small number of enterprise customers driving a large share of revenue. The S-1 will name them, or at least name the percentage thresholds.
3. Voting structure. Anthropic's long-term governance includes a Long-Term Benefit Trust. How that body's authority survives a public listing is a governance question that institutional investors will price.
4. The Pentagon lawsuit status. Resolution, settlement, or escalation between now and the public S-1 will move the offering's valuation more than any product announcement.

The practical implication for anyone watching this filing: ignore the headline valuation. The number that matters is the one buried in the cost of revenue section, because that is where Anthropic's $1.25 billion-a-month bet either pays off or punishes everyone who bought at the IPO price. Read the footnotes first.