Anthropic confidentially submitted its S-1 registration statement on June 1. Three days later, on June 4, the company published an essay titled "When AI Builds Itself" urging artificial intelligence laboratories to retain the option to slow down [1][2].

The juxtaposition is the story. A $965 billion IPO filing and a self-improvement risk warning within 72 hours. One event is a capital-raising mechanism. The other is a governance argument. Together they illustrate the central tension of the AI-state-power thread: companies that benefit from AI acceleration are now the ones defining AI governance.

The Safety Essay

Anthropic's essay argued that AI labs should preserve the ability to pause development if models demonstrate self-improvement capabilities that exceed safety thresholds. The essay did not propose specific enforcement mechanisms, counterparties, or verification standards. It called for the "option to slow down" — a framing that is aspirational by design.

The Wall Street Journal covered the essay as a safety story: "Anthropic Urges Global Pause in AI Development, Flags Self-Improvement Risk" [3]. SiliconAngle's headline captured the market's skepticism: "Filing IPO, Anthropic Says AI Slow — Fat Chance" [4].

Sawyer Merritt, a widely followed X account focused on technology and business, connected the dots in a thread: SpaceX's $1.25 billion per month compute contract with Anthropic — disclosed in the IPO filing — is the revenue story behind the safety headline [5]. The company calling for a pause is also the company whose revenue depends on accelerating compute through 2029.

The Revenue Structure

The S-1 filing disclosed the SpaceX relationship as a material customer concentration risk. Anthropic's revenue is heavily dependent on a small number of large customers, with SpaceX's $1.25 billion per month contract through May 2029 representing the largest single relationship [6].

The circularity is the structural point. Anthropic calls for the option to slow AI development. Anthropic's largest revenue source is a compute contract that requires Anthropic to scale AI development to fulfill. The essay is either a genuine governance proposal that the company's own business model contradicts, or it is a regulatory hedge — establishing Anthropic as the safety-conscious AI company in the eyes of future regulators while its revenue accelerates in the opposite direction.

MSM outlets treated the IPO and the essay as separate stories. CNBC covered the filing mechanics [7]. The Journal covered the safety essay. Neither publication connected the two events structurally.

The IPO Market Context

Anthropic's $965 billion target arrives during a week that also includes SpaceX's IPO roadshow beginning at a $1.75 trillion valuation [8]. The two largest AI-adjacent companies in the world are simultaneously pursuing public-market listings. The combined capital raise — potentially exceeding $200 billion — represents the largest single-week capital formation event in technology history.

The AI-state-power thread's active questions include whether public-market gatekeepers will impose governance conditions on AI companies that private markets did not [9]. The S-1 filing is the first test. If the SEC requires Anthropic to disclose the safety essay as a risk factor or a governance commitment, the essay's language becomes a regulatory artifact. If the SEC treats it as marketing, the essay remains aspirational.

SiliconAngle reported that the filing contained a notable passage: "Anthropic says AI slow — fat chance" — suggesting the company's own lawyers may have included the safety language as a risk disclosure rather than a corporate commitment [10].

What the Paper Reads

The paper's AI-state-power thread tracks how AI infrastructure becomes state infrastructure — who controls the compute, the data, the routing, and the governance [11]. Anthropic's week illustrates the governance problem from the inside. The company that publishes safety essays also signs billion-dollar compute contracts. The company that calls for a pause also files for an IPO that requires growth.

This is not hypocrisy — it is the structural condition of every AI company. Safety and growth are not opposites; they are the same business strategy at different altitudes. At the product level, safety features attract enterprise customers. At the corporate level, safety branding attracts regulatory favor. At the market level, safety narratives support premium valuations.

The question is whether the safety language becomes binding — whether Anthropic's essay produces a concrete governance proposal with verification, counterparties, or enforceable conditions. If it does, the essay changes the regulatory landscape. If it does not, it was a marketing document filed during IPO week.

The S-1 is now on the SEC's review clock. The essay is on the public record. The $1.25 billion per month SpaceX contract is on the balance sheet. The three facts coexist without resolution.

-- THEO KAPLAN, San Francisco