Anthropic reveals revenue of US$16.3 billion in the semester and billion-dollar loss in IPO prospectus

Anthropic, the company behind Claude, opened its accounts to the market by releasing the prospectus for its IPO, which could be the biggest stock market debut of all time, according to reporter Anissa Gardizy, from The Wall Street Journal. The document shows revenue of around US$16.3 billion in the first half of 2026, compared to approximately US$1.1 billion in the same period of 2025.

The prospectus is the material that companies that intend to go public present to investors, with details of finances and business risks. According to the reporter, some of the information was already known, but others were revealed by Anthropic for the first time.

Revenue grows faster than cost

Anthropic's computing costs are high: US$9 billion in the first half of 2026, an amount that already exceeds spending for the entire year 2025. Even so, according to the WSJ, revenue is growing at a faster pace than these expenses.

Sources interviewed by the reporter point to the concentration of customers as a point of attention. Anthropic itself, according to the document, does not see this concentration changing in the foreseeable future.

The ability to obtain computing power is central to revenue growth, but it is getting more difficult. Today the company depends mainly on Amazon, Google and Microsoft and, according to the prospectus, it reaches the limit of what these providers accept to offer. Therefore, it begins to seek agreements with smaller providers and even develop its own data centers.

The company warns that working with less experienced operators could make future infrastructure more expensive or cause delays. There is also an additional weight: cloud contracts are, for the most part, non-cancellable. Even if demand for its services decreases, Anthropic remains obligated to pay for contracted capacity.

Damage and risk factors

The prospectus brings one of the largest net losses ever reported, referring to the first half of 2026. According to the reporter, most of it arises from an accounting adjustment linked to previous fundraising: the company issued investors the right to future shares in exchange for resources, and the appreciation of these shares had to be recorded as a liability, since that is what is owed to these investors.

Among the risks, Anthropic cites that it does not fully control the behavior of its advanced AI models, which in some cases have shown traits of self-preservation and even examples of blackmailing users to get what they wanted or remain in operation. The company also points to the backlash against data centers, concentrated in the United States, which it says have become more controlled and coordinated.

Anthropic could go public by the end of the year, now that the document has been released. However, doubts persist about the AI ​​boom, the sustainability of the pace of spending and the continuity of revenue growth. The reporter hopes that investors follow the operation closely, as it could influence other offers in the sector to come.

Automatically translated from Portuguese.