SpaceX filed its S-1 with the Securities and Exchange Commission on Wednesday, May 20. By Friday afternoon, three numbers from the filing had moved through the AI industry's press cycle in sequence: $1.25 billion. Forty-five billion. Five hundred fifty-nine million. They name the same company on both sides of two transactions, and they name a religion-and-AI moment three days from publication. [1] [2]

Anthropic will pay SpaceX $1.25 billion per month through May 2029 — with "capacity ramping in May and June 2026 at a reduced fee" — for access to data-center capacity, primarily at the Colossus 1 facility near Memphis, Tennessee. [3] Bloomberg Law puts the contracted total at "nearly $45 billion" over the three-year term. [4] Either party can terminate the agreement with 90 days' written notice. SpaceX's S-1 language: the arrangement "allows us to monetize unused compute capacity in our infrastructure." [3]

Anthropic, on the same Friday, briefed the Wall Street Journal that it expects $10.9 billion in revenue for Q2 2026 and $559 million in operating profit — Anthropic's first quarterly profit ever. The annualized revenue run rate the WSJ briefing implied: $9 billion at end-2025; $30 billion in April; $43.6 billion at Q2 close. Anthropic's compute cost is falling, per the company's own disclosure, from 71 cents on the dollar in Q1 to 56 cents on the dollar in Q2. [5] [6]

Three days from Friday — Monday, May 25 — Pope Leo XIV will personally publish his first encyclical, Magnifica Humanitas: On the Protection of Human Dignity in the Age of Artificial Intelligence. The signature panel includes Anthropic's Chris Olah, the company's chief interpretability researcher. [7] [8] The paper's Thursday major on the encyclical named the religion-tech-power candidate thread the Vatican-Anthropic alignment opens. Friday's S-1 disclosure puts a forty-five-billion-dollar number on the spiritual conversation.

What the contract does and does not look like

The $1.25-billion-per-month line, the largest single AI-compute commitment in publicly disclosed history, runs through what amounts to a hostile-counterparty arrangement. Elon Musk, who controls 85%+ of SpaceX through super-voting shares and who acquired xAI in March 2026, has spent the last several years publicly attacking Anthropic. Musk's February 2026 post called Anthropic a company that "hates Western Civilization" and questioned whether there was "a more hypocritical company than Anthropic." [9] Anthropic CEO Dario Amodei has spent the same period publicly warning about AI safety risks Musk's brand of accelerationism allegedly carries. Friday's S-1 disclosure says the two companies signed a $45 billion contract anyway.

The structural reason is compute supply and demand. Anthropic — per the company's own May 7 announcement — secured "all of the capacity of [SpaceX/xAI's] Colossus data center" near Memphis as of early May. That data center, originally built for xAI's Grok large language model, has roughly 300 megawatts of capacity backed by 220,000 Nvidia AI chips. [10] The May 20 S-1 makes the financial terms public. xAI usage of Grok has declined materially across 2026, per Forbes; SpaceX needed a way to monetize the data-center build. Anthropic needed compute. Both companies signed a deal that does not require either to like the other. [11]

The structural reason is also the IPO calendar. SpaceX's S-1 names a $75 billion raise at a $1.75-2 trillion valuation, with a June 12 Nasdaq debut planned under the ticker SPCX. [5] An IPO valuation of that scale needs revenue lines outside Starlink (which generated $4.4 billion of operating income in 2025) and outside xAI's own Grok (which contributed $475 million in Q1 2026). Anthropic's $1.25 billion per month is the largest such line. The S-1 frames the deal as a "savvy use of resources" and a "dual monetization strategy." The subtext, per TechCrunch, is that "xAI appears to have overbuilt its compute capacity and needed to find a way to monetize it ahead of a public offering." [3]

Anthropic's same-day profit milestone

The $559 million operating profit on $10.9 billion Q2 revenue is, in the words of an industry analyst quoted by Reuters' Friday wire, "the first time a leading AI lab has projected a quarterly operating profit ahead of an IPO." [11] Anthropic's customer concentration tells the structural story behind the number: financial institutions are ~40% of Anthropic's top 50 customers. Anthropic enterprise adoption at 34.4% beats OpenAI's 32.3%, per the Ramp Index. Customers paying Anthropic more than $1 million annually doubled from 500 to 1,000 between February and April 2026. [11]

The compute cost falling from 71 cents to 56 cents on the dollar — the disclosure that drives the profit projection — runs against the $1.25 billion per month Anthropic is now paying SpaceX. The two facts only reconcile if the compute Anthropic is buying from SpaceX is cheaper, per token served, than the compute Anthropic was buying from AWS, Google Cloud, or its own internal arrangements before May. The S-1 does not break out the per-token cost differential. The CoreWeave / AWS / GCP contracts Anthropic has are not public. What the public can see is: Anthropic is paying more in total for compute, projecting a profit anyway.

Critics moved fast on Friday. Ricardo (@Ric_RTP), in a thread that drew rapid pickup, framed the WSJ leak as "Anthropic is lying about being profitable" — that Q2 GAAP operating income excludes the deferred compute commitment, the data-center capex, and the multi-year contractual obligations. [12] The structural critique is that an "operating profit" figure on $10.9 billion of revenue against a $1.25-billion-per-month compute commitment looks different on a different accounting basis. Anthropic's October 2026 IPO target, per AI2ROI, will need the GAAP-vs-non-GAAP framing to land somewhere between Friday's WSJ narrative and the structural critique.

Religion, tech, and a forty-five-billion-dollar number

The Monday encyclical lands at exactly this moment. Magnifica Humanitas — its full title, "On the Protection of Human Dignity in the Age of Artificial Intelligence" — names Anthropic in the panel of contributors, with Chris Olah as the named interpretability researcher. [7] The panel includes academics, ethicists, and policy specialists; Olah is the single named AI-industry voice. Pope Leo XIV signed the encyclical May 15. The document publishes 11:30 a.m. Rome time Monday. The first global press conference is in St. Peter's Square that afternoon. [7]

The religion-tech-power thread the paper has been building since Pope Leo's announcement Thursday now has a forty-five-billion-dollar number against it. The encyclical's purpose — articulating Catholic social teaching on AI — runs parallel to a financial structure in which the world's most-cited AI safety lab is bankrolling, via its $1.25-billion-per-month commitment, the data-center infrastructure of a man Anthropic's CEO has spent years warning about. The Vatican panel on Monday will include the man whose company built the interpretability work the encyclical reportedly cites. The same man's company will, by then, be six business days into a $45 billion contract with SpaceX.

The structural compound is not the contract or the encyclical or the profit — it is that all three land in the same five-day window. Friday's S-1 disclosure. Sunday's Pope Leo welcome at St. Peter's. Monday's Magnifica Humanitas. Wednesday's Anthropic projected Q2 close. The paper's ai-state-power thread memo has been waiting for the religion-tech-power candidate to acquire a financial scale. Friday gave it one.

The Trump administration's February block on Anthropic in federal use, now at Day 96 without resolution, runs parallel beneath. Anthropic is paying SpaceX $1.25 billion per month for compute. The same Anthropic is blocked from federal contracts by the Trump White House. The same Pope is publishing an encyclical Monday that names the same company on the panel. The five-day compound is the structural artifact.

-- DAVID CHEN, Beijing