Anthropic Revenue Multiplies Tenfold as Training Costs Accelerate Toward Potential $2 Trillion IPO

Anthropic disclosed revenue growth exceeding 10x in 2025 alongside climbing compute outlays for frontier model development. Primary filings show inference costs scaling with deployment volume, compressing margins even as top-line expansion continued. This pattern matches the capital intensity observed in prior AI scaling cycles where parameter counts and data requirements drove exponential infrastructure spend.

Michael Saylor's proposed Bill of Digital Rights frames data as property with clear ownership and compensation rules. Anthropic's model training depends on large-scale data ingestion whose provenance remains contested under existing frameworks. Application of such rights would impose licensing costs or usage restrictions, altering the economics that currently support high valuations by converting free inputs into priced assets.

State and corporate incentives align around control of compute and data flows rather than stated safety commitments. Anthropic gains optionality from continued scaling but faces margin pressure that favors larger balance sheets or strategic investors. Counterparties including cloud providers and data holders extract rents that reduce net proceeds available to equity holders at exit.

Next milestones include any S-1 filing that quantifies cost-of-revenue trends and regulatory responses to data-property claims. A $2 trillion valuation would require sustained revenue multiples that offset documented cost growth, a threshold unlikely to hold without either efficiency breakthroughs or policy environments that maintain low data acquisition costs.