Anthropic의 IPO 공시는 이해관계자 및 파트너로부터 $518 billion 규모의 인프라 투자 약정을 공개했다.

Anthropic's confidential IPO prospectus has revealed the financial scale of its AI ambitions, disclosing $518 billion in long-term computing and infrastructure commitments alongside rapidly growing revenue and substantial operating losses. The proposed listing, which could value the company at more than $2 trillion according to Reuters, would expose public investors to one of the largest infrastructure spending programmes in the technology industry.

Anthropic generated nearly $4.6 billion in revenue during 2025, approximately 12 times its 2024 figure. However, its operating loss widened to $8.06 billion, compared with $2.98 billion a year earlier. The company reported a net loss approaching $42 billion, although approximately $34 billion of this resulted from an accounting charge associated with increased valuations of financing instruments that could eventually be converted into Anthropic shares. This figure should not be confused with cash spent on operations.

Computing requirements account for much of Anthropic's expenditure. During 2025, the company spent $7.33 billion on computing and infrastructure, more than three times the previous year's figure and over half its total operating expenses of $12.65 billion.

The prospectus discloses at least $518 billion in future infrastructure commitments extending over approximately a decade. Around 80% of these commitments are either non-cancellable or require payment regardless of actual computing usage. The arrangements include at least $111.1 billion in obligations with Google, $110 billion with Amazon and $31.4 billion with Microsoft. Anthropic also has approximately $161.2 billion in Broadcom-related equipment lease commitments. The company expects access to approximately 3.5 GW of next-generation TPU-based computing capacity beginning in 2027, subject to continued commercial success.

Anthropic is moving beyond its original reliance on cloud services towards directly leased processors and dedicated data-centre infrastructure. Its prospectus identifies computing availability as a principal potential constraint on future growth.

Despite its revenue growth, customer concentration presents a commercial risk. Nearly a quarter of 2025 revenue came from just two customers, while several major clients lack long-term contracts. Anthropic held $20.28 billion in cash, cash equivalents and short-term investments at the end of December 2025.

The proposed valuation exceeds twice the $965 billion private valuation established during Anthropic's May fundraising round. The company submitted its IPO paperwork confidentially to the SEC in June, with Reuters reporting that the offering could occur after November's US midterm elections.

The prospectus devotes approximately 80 pages of its 261-page main body to risk factors associated with increasingly capable AI systems, including potentially harmful autonomous behaviour, model resistance to shutdown and systems manipulating or concealing information. CEO Dario Amodei has publicly called for greater caution in releasing increasingly powerful capabilities, yet Anthropic continues its commercial development programme, having introduced Claude Opus 5.5 on 22 September and Claude Sonnet 5.5 on 28 September.

For prospective investors, the Anthropic IPO brings together rapid commercial expansion, exceptionally large long-term infrastructure obligations and considerable uncertainty about the economics of increasingly capable AI. It provides a financial reference point for the wider industry, including rival OpenAI, whose own public listing remains under consideration.