Anthropic's prospectus reveals AI infrastructure "hard contracts": 80% of spending locked in with Google, Amazon, and Microsoft is "non-cancelable"

date
20:19 29/09/2026
avatar
GMT Eight
According to a confidential IPO prospectus seen by foreign media, Anthropic is expected to invest at least $518 billion over ten years with six partners to build AI infrastructure, a plan that ranks among the largest disclosed AI infrastructure commitments.
According to its IPO prospectus, Anthropic expects to invest at least $518 billion over a decade with six partners to build AI infrastructure, a plan that ranks among the largest disclosed AI infrastructure commitments. The AI lab said in the prospectus that about 80% of the amount is non-cancelable, or payable whether or not used. Anthropic told investors these commitments are necessary because access to computing power is becoming a key bottleneck in AI development, and future demand for advanced AI systems may exceed supply and will be "primarily constrained by compute availability." According to the IPO prospectus, Anthropic PBC's overall revenue in 2025 surged wildly to about 12 times the previous year, or nearly $4.6 billion. At the same time, the public offering by the AI application leader that developed Claude could value the company at more than $2 trillion, surpassing the roughly $1.77 trillion overall valuation of SpaceX, founded and led by Musk, at the time of its listing. Anthropic's planned fundraising scale of nearly $100 billion would also exceed SpaceX's roughly $85.7 billion total raised after the over-allotment option, all of which means Anthropic is poised to surpass SpaceX in one stroke and become the largest initial public offering in history. Locking in Google, Amazon, Microsoft and Broadcom The company said it plans to pay Alphabet's Google at least $111.1 billion, Amazon $110 billion, and Microsoft $31.4 billion over the next 7 to 10 years under long-term infrastructure service obligations, "whether or not used." The prospectus shows that Anthropic also bears about $161.2 billion in equipment lease obligations related to Broadcom, most of which are non-cancelable. Anthropic confidentially filed for an IPO with the U.S. Securities and Exchange Commission in June, but the document has not yet been publicly disclosed. Anthropic did not immediately respond to a request for comment on the document. The company said the payment period to Google is from April 2026 to July 2033, and the payment period to Amazon is from May 2026 to April 2036. "If our actual spending falls short, we must pay Google the difference," the company said, adding that the Amazon agreement has similar terms. The company added that the $31.4 billion commitment to Microsoft runs from November 2026 to May 2033 and is "non-cancelable unless Microsoft commits an uncured material breach." Anthropic also said the lease arrangement with Broadcom is non-cancelable for both parties unless there is a default. Additional partnerships with xAI and AMD The company also disclosed an agreement with Musk's xAI that could generate up to $84.5 billion in spending by 2029 for Nvidia compute capacity, most of which can be canceled with 90 days' notice. The document also shows that Anthropic's relationship with AMD is deepening. AMD has committed to buying up to $5 billion of Anthropic stock and providing AI compute capacity expected to exceed $20 billion. The total scale is comparable to OpenAI's "Stargate" project, a $500 billion AI infrastructure plan whose costs are expected to be shared by OpenAI, SoftBank, Oracle and MGX. Shift toward building its own infrastructure Anthropic said it is increasingly building its own AI infrastructure, shifting from a pure cloud model to dedicated data centers and directly leasing chips. The company lists its reliance on Amazon, Google and Microsoft as a major risk. While these three companies provide computing infrastructure and distribution channels, they are also developing competing AI models. The company said these three tech giants simultaneously play multiple roles as investors, customers, cloud providers, distributors and competitors, and their interests and motivations "may not be fully aligned with Anthropic's." "If our access to compute from third parties is reduced, repriced or terminated... our business, financial condition and operating results could be adversely affected," the company said.