Anthropic's IPO Window Just Hit a Federal AI Bill — and the Compliance Drag Is Not Shared Equally

The regulatory chain is short and unusually fast. On the evening of September 8, Jacob Coxon resigned from Anthropic and published a warning that AI could kill all humans by the end of the decade. Within hours, Anthropic's alignment-science lead Evan Hubinger confirmed the view, putting the probability above 10%. By September 9, Reuters and Axios had the story; by September 10, members of both parties were publicly demanding federal action.

Of the four vehicles, H.R. 9925 is the one with hard thresholds, hard deadlines, and real money attached. Its text, posted to Congress.gov on July 23, 2026, creates an Under Secretary of Commerce for AI Security and binds any developer meeting two simple tests to a front-loaded compliance regime. The structure forces independent verification, public reporting, and gives the Secretary of Commerce emergency shutdown authority over frontier deployments.

Regulation costs money. The companies with the deepest foundation-model exposure enter this cycle already spending 22-35% of revenue on AI capex, with free cash flow under pressure at two of them. A new compliance opex layer — third-party audits every 12 months, IVO assessments every six months, transparency reports before each deployment — stacks onto balance sheets that are already pricing in AI infrastructure spend years in advance.

The same bills that add opex to the foundation-model labs barely touch the picks-and-shovels layer. Independent verification and pre-deployment audits increase the cost of training a new frontier model — they do not decrease the compute required to train it. Training runs that meet the FRONTIER Act's 10^26-FLOP threshold still need every GPU and every HBM stack they were going to need anyway. Regulation keeps training compute demand on its existing ramp regardless of which bill passes.

Anthropic has confidentially filed its S-1 and is targeting an October listing at a valuation some bankers say could reach $2 trillion. Reuters, citing people familiar with the financials on August 15, 2026, tied that valuation to a 2028 revenue forecast of $190B-$200B and a Q2 2026 revenue projection of at least $10.9B — which would deliver Anthropic's first quarterly operating profit of about $559M. The path from Q2 2026 to 2028 implies an 18x revenue ramp at the same time the company is publicly cited in extinction-risk hearings. That sequence turns the IPO narrative from a clean 'first $1T AI listing' into a 'first $1T AI listing under active federal scrutiny'.