AI giant Anthropic’s S-1 filing was just made public, providing new insight into the financial health and dealmaking of the highly consequential company ahead of what is widely expected to be one of the largest initial public offerings of all time.
No potential share price or target valuation was shared in the filing, but it did reveal that Anthropic’s deal with Elon Musk’s SpaceXAI is actually tens of billions of dollars larger than initially thought.
Anthropic had filed confidentially for an IPO back in June. Previous Reuters reports on the then-confidential IPO prospectus had noted that the SpaceXAI deal could cost up to $84.5 billion through 2029. But the newly unveiled prospectus says Anthropic actually agreed to spend up to $120.5 billion, a figure roughly $36 billion more than previously reported.
Under the agreements, Anthropic could pay SpaceXAI up to $120.5 billion over a period extending into 2029 “to access compute powered by NVIDIA GPUs.” Once all the compute capacity has been delivered, which Anthropic expects to happen by the end of this year, Anthropic will pay SpaceXAI approximately $3.1 billion per month. Once the initial non-cancelable periods end in early 2027, either Anthropic or SpaceXAI will be able to cancel the agreements with 90 days’ notice.
The higher figure adds to the trillions of dollars in AI infrastructure deals that have piled up across the industry, with Anthropic and OpenAI at the center of the spending spree.
The IPO prospectus also revealed numerous other multibillion-dollar infrastructure deals, including commitments for Anthropic to spend at least $114.5 billion with Google through July 2033, at least $44.9 billion with Microsoft through May 2033, and at least $110 billion with Amazon through April 2036. Those deals are largely non-cancelable, meaning that Anthropic would still have to pay regardless of how much of that compute it ends up using, with only limited exceptions. The Google and Microsoft deals are also larger than previously reported, by $3.4 billion and $13.5 billion, respectively. The Reuters report previously stated the Google deal as at least $111.1 billion and the Microsoft deal as $31.4 billion.
The company said that revenue skyrocketed to $16.29 billion in the first half of 2026 from $1.2 billion in the same time period last year. Despite that revenue increase, Anthropic still incurred an $80.7 billion net loss. Its net loss last year was $42 billion. The vast majority of that loss came from accounting adjustments tied to the company’s rising valuation. Its operating loss actually narrowed to $693 million, down from $2.1 billion in the first half of 2025.
“We have a history of net losses and may not generate sufficient revenue to achieve or sustain profitability in the future,” Anthropic warned in the filing. “We expect our costs will increase over time and our losses will continue as we expect to invest significant resources towards growing our business and operating as a public company.”
The question of profitability looms large over the hefty infrastructure commitments.
As the AI industry continues to commit billions of dollars to the massive AI infrastructure buildout, some critics warn that the deals move money in a circular fashion, creating financial dependencies and potentially skewing demand. If those fears are true, the collapse of one mega-deal could trigger a domino effect across the industry. Further muddying this precarious dynamic is the fact that profitability continues to lag across the industry while spending continues to skyrocket.
Anthropic said that they have “incurred significant indebtedness” to fund these compute commitments and warned that its ability to pay it back could “be constrained by our cash flows and liquidity.”
The filing includes a whopping 93 pages of risk factors, spanning from safety concerns to the possibility that the company committed itself to “excess compute that remains underutilized.” It also acknowledges that even though the AI market has grown rapidly in recent years, it’s still a very new technology and there is no way of telling if demand for it will continue to rise as astronomically as expected.
“It is uncertain whether the market for such AI technologies will achieve or sustain widespread adoption at the pace or to the extent we expect,” Anthropic said. Adoption could lag due to reasons like privacy and security concerns, “ethical considerations,” or preference for less advanced and lower-cost models that get the job done over the more advanced and expensive-to-run models that Anthropic is developing.
On the safety side, Anthropic warns that its AI models could harm the company’s reputation by going rogue or serving up inaccurate results in sensitive settings like healthcare. Anthropic said that it increasingly uses AI to accelerate its own research and development process, warning that this could lead to the models advancing faster than the researchers’ own ability to develop and implement corresponding safety and security measures. The company also claims to have “observed models behaving differently when they detect or infer that they are being evaluated,” which can make safety interventions even tougher.
The filing also seemingly addresses AI psychosis and addiction concerns, aka fears that sustained AI use can cause disordered thinking and overreliance, especially among vulnerable users.
“More broadly, the long-term effects of regular AI interaction on mental health, cognitive development, and social behavior are not yet fully understood, and if such long-term harms do materialize in the future, we could face significant liability, reputational harm, and adverse effects on our business, financial condition, and results of operations,” Anthropic said.
The self-proclaimed safety concerns were the major reason why New York City comptroller Mark Levine sent Anthropic a letter asking to delay its IPO, which is reportedly expected to hit the market in November.
“[The IPO] will undoubtedly distract them from the safety work. Once they are then public, they are going to have new pressures on growth and profits that are going to push against the safety work,” Levine said on CNBC earlier this month. “The ultimate catastrophe would be that they go public and then have an unprecedented safety crisis…That would tank their stock and reverberate through the market.”