Anthropic’s possible $2tn initial public offering is a financial story built on a technological question: how valuable does a frontier AI system become when it moves from research capability into routine enterprise infrastructure?
The Financial Times reported that several Anthropic investors expect the company to float at a valuation of $2tn or more, potentially in October. Anthropic has not set that valuation. It said on 1 June that it had confidentially submitted a draft Form S-1 to the Securities and Exchange Commission, while the number of shares and the offering price remain undecided.
The most important confirmed metric behind the optimism is Anthropic’s revenue run rate. In May, the company said it had crossed $47bn. TechCrunch reported that the measure had been about $9bn at the end of 2025. The increase tracks a period in which Claude has become more widely used in coding, enterprise workflows and other professional tasks.
A run rate is not the same as a full year of reported revenue. It takes a recent pace of sales and expresses it as an annual figure. For a technology business expanding this quickly, it can show current scale better than a trailing annual number, but it is sensitive to whether demand continues at the same speed.
Investors cited by the FT expect Anthropic’s annualised revenue to reach $100bn to $120bn by the end of 2026. The range is an investor forecast, not public company guidance from Anthropic. It therefore represents a hypothesis about adoption as much as a financial projection.
That hypothesis has produced extremely high valuation models. One investor argued that if growth were around 800% a year, a revenue multiple near 30 times could lead to a valuation around $3tn. The model is straightforward, but the science and engineering economics underneath it are not.
Frontier models require major computing resources. In announcing its $65bn Series H round at a $965bn post-money valuation, Anthropic said it would use funding to expand compute, products and partnerships as well as safety and interpretability research. The company has also described Claude as increasingly embedded in core enterprise operations.
This creates an unusual economic structure. Software can be distributed globally at enormous speed, but the underlying models require expensive training and inference infrastructure. The value of technical advances therefore depends not only on benchmark performance but on how efficiently they can be delivered to paying users at scale.
A public Anthropic would give investors a much clearer dataset for evaluating that relationship. A public S-1 should eventually show more about recognized revenue, costs, capital needs and risk factors. Those figures will help separate the technological adoption curve from the valuation multiple placed on top of it.
For now, the best-supported numbers are the $965bn May valuation and the $47bn-plus May run rate. The $2tn and $3tn figures describe what investors think those numbers could become if enterprise demand continues to compound at an exceptional pace. The IPO will test whether public markets agree with that view of frontier AI economics.
