The Claude maker's S-1 shows soaring revenue, a massive loss, and decades of largely non-cancelable spending on servers, buildings, and power
Anthropic has finally opened its books. The first thing inside is a bill that could total nearly $900 billion.
The AI lab behind the Claude model filed its S-1 registration statement with the SEC on October 26, 2026. The filing lays out future payment obligations for servers, data centers, buildings, and power that stretch across one to 20 years. Most of them are largely non-cancelable, which is a polite way of saying the checks are due whether the AI boom keeps booming or not.
What the filing actually shows
The headline number breaks into layers. The firmest layer is approximately $417 billion in minimum payments for compute infrastructure, data centers, and leases.
On top of that sit an additional $125 to $150 billion in projected capacity costs covering 2028 through 2031. Those are expected costs rather than locked minimums. Broader obligations push the total commitments toward the nearly $900 billion mark.
The partner list reads like a who’s who of Big Tech. Broadcom accounts for $161 billion, Google for $114.5 billion, Amazon for $110 billion, and Microsoft for $45 billion, each over various terms. SpaceX/xAI also appears among the company’s key partners.
Those four named commitments alone add up to well over $400 billion. That concentration matters, and the filing flags it: Anthropic relies heavily on a small number of providers to keep growing.
Anthropic describes its approach to locking in these deals as “bold but sane.” The company also acknowledges that its commitments might be viewed as irresponsibly large within the industry.
Revenue up, losses way up
The growth story is real. Anthropic reported revenue of $16.3 billion in the first half of 2026, up from $1.16 billion in the same period a year earlier.
That works out to a roughly 1,300% year-over-year increase.
The loss column is equally dramatic. Anthropic posted a net loss of $80.7 billion for the first half of 2026.
Most of that loss came from non-cash accounting charges, according to the filing. Non-cash charges are expenses recorded on paper that do not reflect money actually leaving the bank account in that period.
The road to Nasdaq
Anthropic anticipates listing on Nasdaq under the ticker ANT. Its potential valuation has been estimated between $1.5 trillion and $2 trillion, supported by heavy investment from major technology players.
The structure of the deals creates an unusual web of relationships. Google, Amazon, and Microsoft are partners and suppliers, but they also run competing AI efforts of their own. Broadcom, the largest single commitment on the list, supplies the chip side of the equation.
In practice, much of Anthropic’s spending flows back to the same companies that back it. That circularity is not unique to Anthropic, but the S-1 puts precise figures on it for the first time.
What this means for investors and the AI race
For prospective shareholders, the core question is simple to state and hard to answer. Can revenue keep compounding fast enough to cover obligations that run for up to two decades?
Partner concentration adds another layer of risk. When a handful of suppliers account for the bulk of your capacity, their pricing decisions, supply constraints, and strategic priorities become your problem too. The filing itself treats this reliance as a material factor.
Watch three things as the IPO approaches. First, whether the valuation range holds up once institutional investors digest the obligations. Second, how Anthropic explains the gap between its firm $417 billion and the broader nearly $900 billion total. Third, whether second-half 2026 revenue keeps pace with the trajectory that made the first half so eye-catching.
Anthropic calls its strategy bold but sane. Public markets are about to decide how much weight to give each of those words.
