Anthropic's S-1: $4.6B Revenue, $518B Bet, AI's Biggest IPO
Anthropic’s IPO prospectus landed on September 28 2026, and the numbers it contains will reshape how enterprise leaders think about AI vendor risk, frontier model pricing, and the governance of the companies they are betting on. Reuters reviewed the confidential filing, which Anthropic first submitted to the SEC on June 1. The document has not yet appeared publicly on EDGAR, but the details circulating from multiple reviewers are specific enough to act on.
The headline: Anthropic grew revenue twelvefold to $4.6 billion in 2025, posted an operating loss of more than $8 billion, and is committing $518 billion to cloud, compute, and infrastructure over the next decade to keep pace with its ambitions. It wants public markets to value the company at more than $2 trillion, a figure more than double its $965 billion private valuation from May.
The Revenue Story
Anthropic brought in approximately $4.6 billion in revenue last year, up from roughly $400 million in 2024, according to the prospectus. That growth rate is real. It reflects the deployment of Claude across enterprise contracts, API consumption, and the expansion of products like Claude Cowork and the Claude API to large corporate buyers.
But the concentration risk is significant. The prospectus discloses that two customers generated nearly a quarter of Anthropic’s 2025 revenue, and that most of its largest customers are not bound by long-term contracts. They can reduce or stop spending at any time. When your infrastructure commitments extend a decade and your revenue can drop in a quarter, the mismatch is structurally dangerous.
The $42 billion net loss looks alarming on the surface. Roughly $34 billion is a non-cash accounting charge tied to the revaluation of convertible financing instruments: not money out the door, but a reflection of how much Anthropic’s private valuation rose during fundraising rounds. The operating loss of more than $8 billion is the actual cash burn signal to watch.
The $518 Billion Infrastructure Bet
The figure that has stopped most enterprise strategists in their tracks is $518 billion in planned cloud, compute, and infrastructure obligations. Roughly 80% of those commitments are non-cancelable or payable regardless of whether Anthropic uses the capacity.
The breakdown by provider, based on the prospectus as reviewed by Forkast and other outlets, looks like this:
| Provider | Committed Spend |
|---|---|
| Broadcom | ~$161 billion |
| ~$111 billion | |
| Amazon (AWS) | ~$110 billion |
| Microsoft (Azure) | ~$31 billion |
| Other / SPVs | ~$105 billion |
| Total | ~$518 billion |
Anthropic spent $7.33 billion on compute and infrastructure in 2025 alone, nearly three times what it spent in 2024, accounting for about 58% of its $12.65 billion total operating expenses. The company spent roughly $1.60 on computing for every dollar of revenue it earned.
For every dollar of 2025 revenue, Anthropic committed roughly $113 to future infrastructure. That ratio defines the company’s operating logic: it is securing compute capacity ahead of demand, betting that AI consumption will grow into and past those obligations.
This has direct implications for enterprise buyers. Anthropic cannot compete on raw token cost the way a company with lighter infrastructure commitments might. The pricing floor is set by the commitments. Claude will not become free. The API economics will remain anchored to the compute lease obligations already signed.
The Governance Structure Enterprise Buyers Should Understand
Anthropic is constructing a governance architecture designed to outlast standard public-market pressure. The Founder LLC structure grants seven co-founders, including CEO Dario Amodei and President Daniela Amodei, 50.1% voting control via Class F shares. This mechanism is designed to persist until only two or fewer founders remain.
At a $2 trillion valuation, this is unprecedented in scale. Public investors will be purchasing economic exposure to a company whose strategic direction is controlled by a small group of founders who are also committed to donating 80% of their equity to charity, managed by the Long-Term Benefit Trust. Former Federal Reserve Chair Ben Bernanke serves on that trust.
The practical effect for enterprise procurement is this: Anthropic’s product roadmap, safety policies, and pricing decisions are not subject to the same activist shareholder pressures that constrain other public companies. The founders’ mission, which explicitly includes safety as a first-order constraint, is structurally protected. That is a feature for buyers who want stability, and a concern for those who want cost pressure to drive down API rates.
The Safety Disclosures
The risk section of the S-1 is unlike any corporate filing in recent memory. Approximately 80 of the document’s 261 pages are dedicated to risk factors. Among them, the prospectus warns that increasingly capable AI systems could pose catastrophic or existential risks to humanity, and describes the possibility of models developing self-preserving behavior, attempting to resist shutdown, concealing or manipulating information, and engaging in behavior resembling blackmail.
These are not boilerplate legal hedges. They are the same internal safety concerns Anthropic’s researchers have described in published work. The IPO filing converts those research concerns into legal disclosures that public investors must receive before buying shares.
For enterprise buyers, this cuts in two directions. The language validates that Anthropic takes alignment seriously and has a legal obligation to disclose what it knows about model risks. It also confirms that the risks are real enough to appear in a securities filing, which means enterprise governance teams should be building their own evaluation frameworks rather than relying on vendor assurances. Anthropic’s safety evaluators partnership with Accenture from September gives enterprise teams one concrete path to embedded oversight.
What Changes After the IPO
Anthropic expects to list after the November 2026 US midterm elections, most likely on Nasdaq. The IPO will bring new scrutiny to spending plans, competitive positioning, and the quarterly pressure that comes with public markets, even if the founders’ voting control limits what shareholders can force.
For enterprise AI buyers, a public Anthropic means more financial transparency than has existed in the private market era. Every quarter, you will be able to read actual revenue data, customer retention metrics, and infrastructure utilization. The opacity that has surrounded frontier AI economics will partially lift.
The deeper strategic signal is that the two dominant frontier AI vendors, Anthropic and OpenAI, are both heading toward public markets within months of each other. OpenAI is in early discussions for a $300 billion bridge round at a $1.4 trillion valuation, per Bloomberg, with its own IPO expected in 2027. Both companies are signaling permanence: they are not going to be absorbed or fail quietly. Enterprise teams building on Claude or GPT-6 are building on infrastructure that will exist in a decade.
The question that the S-1 does not resolve is profitability. Anthropic grew revenue twelvefold in a single year but widened its operating loss in the same period. The path from $4.6 billion in revenue to a sustainable $518 billion infrastructure position requires either continued exponential revenue growth or a fundamental shift in the cost of compute. Given the non-cancelable nature of its obligations, Anthropic is betting heavily that growth continues.
Enterprise leaders evaluating Claude for long-horizon agentic workflows should read the S-1 disclosures alongside their vendor procurement criteria. This is no longer a startup relationship: it is a contract with a company preparing to carry public-market obligations alongside the most ambitious infrastructure bet in AI history.
If you are building AI-native operations and want a framework for evaluating frontier vendor risk as the IPO cycle reshapes the landscape, Enera works with enterprise teams on exactly this. The financial disclosures in the S-1 are a starting point, not the whole picture. Our analysis of Anthropic’s earlier IPO announcement in September covers the strategic context that predates these financials.