Anthropic IPO Filing Reveals Soaring Revenue But Even Higher Costs

Anthropic’s initial public offering filing offers one of the clearest looks yet at the economics of a frontier AI company – a business growing at extraordinary speed while spending heavily on the computing infrastructure required to build and operate its models.

The Claude maker filed its S-1 registration statement with the U.S. Securities and Exchange Commission today and plans to list on the Nasdaq under the ticker “ANT.” The preliminary filing does not yet specify the number of shares Anthropic plans to sell or an expected price range.

Anthropic reported revenue of $16.3 billion in the six months ended June 30, 2026, up from $1.16 billion in the comparable period a year earlier. Full-year revenue had increased from $386 million in 2024 to $4.59 billion in 2025, according to the filing.

The growth has come with high infrastructure costs. Anthropic reported $8.95 billion in compute and infrastructure spending during the first half of 2026, compared with $2.33 billion a year earlier. Total operating expenses approached $17 billion.

Such spending is unsurprising for the frontier AI company, according to Juan Londoño, a policy analyst at the Cato Institute.

“Compute power is to artificial intelligence what gas is to a car engine,” he said. “In the same way that cheap energy would bring efficiency gains throughout the economy, access to cheap compute will determine the AI industry’s ability to grow. Building as much infrastructure, namely, data centers, as quickly as possible should be one of the top policy priorities to win the AI race.”

Anthropic reported a net loss of $80.75 billion for the first six months of 2026, but $72.4 billion of the loss came from changes in the fair value of convertible notes and other financial instruments as the company’s valuation increased. A more meaningful figure is its operating loss, which was $693 million in the first half, compared with $2.14 billion in the year-earlier period.

The numbers show that Anthropic is reinvesting its revenue into research, staff and compute, which in turn should produce more capable models that attract additional customers.

Anthropic describes these as two self-reinforcing dynamics – an “improvement loop” that advances Claude and an “investment loop” that helps finance continued frontier AI research.

The IPO, expected sometime in November, is being underwritten by Goldman Sachs and Morgan Stanley as the lead investment banks.

An enterprise-first business

Anthropic describes itself first as a research lab, but its commercial strategy is heavily oriented toward enterprises. Claude is sold through models, a developer platform and applications, with the company seeking to embed the technology more deeply into customer workflows as its capabilities improve.

Rather than depend on one chipmaker or cloud provider, Anthropic says it treats compute as a single pool that can be allocated dynamically according to where additional capacity will be most productive. Its multi-chip architecture spans AWS Trainium, Google TPUs, Nvidia GPUs and AMD GPUs.

That approach is intended to let Anthropic optimize workloads for price and performance while reducing dependence on any single computing architecture. It also reflects how central access to computing capacity has become to the economics of frontier AI.

Anthropic has arranged additional infrastructure financing as it expands. The filing says a March 2026 agreement with Broadcom covers specialized AI accelerators and related components and gives Anthropic the ability, if needed, to issue up to $42 billion in convertible promissory notes to Broadcom or a designated financing partner to finance obligations under the arrangement.

The company plans to use proceeds from the IPO for working capital, operating expenses and capital expenditures, including investments in compute and infrastructure, as well as hiring and retaining highly skilled employees. Some proceeds could also be used for acquisitions or investments, although Anthropic said it has no current agreements or commitments for material acquisitions.

Claude itself is increasingly part of Anthropic’s model-development process. The company said successive versions are used to run experiments, generate and process training data, automate evaluations and safety audits and write and review code used to develop subsequent models. Anthropic calls the recursive process “closing the loop.”

An unusual governance model

The IPO will also bring an unconventional corporate structure to the public market.

Anthropic is a Delaware public benefit corporation whose stated purpose is the responsible development and maintenance of advanced AI for the long-term benefit of humanity. Its Long-Term Benefit Trust generally has authority to elect and remove a majority of the authorized members of the board through a special class of stock.

At the same time, Anthropic’s seven co-founders will retain substantial voting control. A Founder LLC will hold a special Class F share carrying 50.1% of the voting power on matters submitted generally to shareholders, including election of directors other than those selected through the trust structure. Anthropic therefore expects to qualify as a “controlled company” under Nasdaq rules.

The filing also formalizes Anthropic’s public-benefit commitments. The company has contributed 846.8 million Class N shares – representing 5% of its fully diluted shares as of June 30 under the filing’s assumptions – to the Anthropic Public Benefit Fund. Its board may issue additional shares representing up to 2% each time Anthropic’s share price rises 50% above a previous milestone, beginning with the IPO price.