Anthropic filed its public S-1 on 26 October 2026. It remains private: the IPO price and share count are not set. The filing reports $16.286B of first-half revenue and $11.555B in Q2, compared with $4.585B for all of 2025. These are recognized revenues, not annualized run-rates.^1,2

The Rule of X compares a forward revenue multiple with growth and EBITDA margin. The public snapshot contains 156 companies, dated 26 October. SpaceX is a listed overlay and enters the fit once when it passes the active filters. OpenAI and Anthropic never enter the fit. Their default $2T values are Dealroom equity scenarios, not IPO prices; the toggle shows their last priced rounds instead.

Reported revenue and annualized run-rates

Anthropic’s first filing leaves Q3 revenue and its 31 October run-rate fields blank. The $16.8B Q3 revenue figure reported by the Financial Times on 14 October is a separate investor-report figure, not an S-1 disclosure. CNBC reported a $65B end-July annualized figure; the older $74.3B July tracker below is retained only as a dated estimate, not the latest actual.^9,7

CNBC reported OpenAI at approximately $50B of September annualized revenue net of partner revenue, compared with $68B including partners. That accounting distinction is not evidence of a sudden decline in business. Bloomberg reported a December expectation of at least $70B; it is a reported target, not booked 2026 revenue.^7,8

Anthropic’s filed revenue is gross GAAP revenue

The S-1 says Anthropic is the principal in cloud-marketplace transactions: it is responsible for the service and sets pricing. Gross recognition is therefore its GAAP policy, not a number needing a “GAAP-like” correction. We use gross revenue in every main chart. The former 27% haircut was an unverified sensitivity, not filed net revenue or an exact restatement onto OpenAI’s basis.¹

Anthropic reported Q2 operating income of $0.974B and adjusted operating income of $2.235B. Neither is EBITDA. Its first-half net loss of $80.748B included $72.388B of noncash remeasurement of financing instruments and $8.720B of other mark-to-market changes, while operating cash flow was positive $3.929B. The chart’s 0% forward EBITDA margin remains an explicit assumption; it is not inferred from those results.

The forward denominator is an estimate

The annual paths below retain the existing Dealroom workbook scenarios: SpaceX $45B / $140B / $220B, OpenAI $40B / $85B / $150B and Anthropic $57B / $170B / $280B for FY2026–2028. The new filings do not turn these paths into company guidance.¹⁰

For November 2026–October 2027 we use 2/12 of FY2026 plus 10/12 of FY2027, assuming revenue is flat within each year. This calendar-weighted Dealroom estimate is $124.17B for SpaceX, $77.50B for OpenAI and $151.17B for Anthropic. It is not an observed monthly integral. At the chosen values, the multiples are 15.7× EV / revenue, 25.8× equity / revenue and 13.2× equity / gross revenue respectively.

SpaceX: growth, cash needs and its largest customer

SpaceX’s Q3 revenue was $14.6B, up 161% on a year earlier. Super Intelligence, the segment that includes AI compute, brought in $7.56B. Connectivity, mostly Starlink, brought in $5.25B and Space $1.77B.³

The growth uses a lot of cash. In the first nine months SpaceX spent $48.1B on capex against $13.1B of operating cash flow. On 26 October it also agreed about $40B of GPU finance leases with Apollo, starting in 2027.⁵

One customer is a large part of revenue. An unnamed Customer B made up 34.5% of Q3 revenue; the filing does not say who it is. Anthropic’s S-1 describes up to $120.5B of spending with SpaceXAI through 2029, or $37.2B a year once capacity is fully delivered.^4,1

Our 2027 revenue scenario for SpaceX stays at $140B. At full capacity the Anthropic contract would be 26.6% of that, and 36.5% of the scenario’s Super Intelligence revenue. We do not add it on top: the scenario may already count it, so adding it would risk double-counting.¹⁰

• Super Intelligence covers AI infrastructure and applications, Cursor since its acquisition closed on 14 August, and X advertising.
• Margin on the chart: nine-month adjusted EBITDA of $13.5B on $27.1B of revenue, 49.8%. This is a historical proxy, not FY2027 consensus.
• The Apollo leases are subject to closing conditions and are not in the enterprise-value bridge.
• The $120.5B is a multi-year maximum, not annual revenue or guaranteed backlog. After noncancelable periods ending in January and March 2027, either side can cancel on 90 days’ notice.
• The 2027 scenario includes $102B of Super Intelligence revenue: the $140B total less $32B of Connectivity and $6B of Space.

An illustrative value by business

The workbook illustration applies assumed multiples to potential revenue in four businesses. It sums to $3.916T before any separate cash, debt or corporate-cost bridge. The inputs are not reported segment forecasts, a price target or a verified Morgan Stanley endorsement. This is a separate potential-revenue illustration, not an annual point on the company scenario.

What this means

The filings establish a larger recognized-revenue base and reveal financing and customer dependencies that a run-rate chart misses. They do not establish the future revenue or margins needed to support the chosen valuations. The Rule of X is a comparison tool, not a valuation verdict: change the revenue path, margin assumption or equity scenario and the conclusion changes.