Anthropic Confidential Filing Reveals Customer Concentration Risks
Anthropic’s private IPO filing underscores a high degree of client concentration and large infrastructure commitments that are anticipated in the future, the report said. The AI startup has noted significant revenue growth in tandem with heavy investment on compute and cloud/infrastructure.
Despite having promised to invest in cloud and infrastructure worth tens of billions of dollars next year, Anthropic PBC is spending money at an unprecedented rate. After claiming its examination of the AI company's IPO prospectus, which has not yet been released, a media house reported on 29 September.
Despite a twelvefold increase in revenue, Anthropic lost $42 billion in year on $4.6 billion in revenue, a 425% increase from 2024. Spending on compute and infrastructure increased thrice from 2024 to 2017, reaching $7.33 billion.
Recent Expenses of Anthropic
Operating losses exceeded $8 billion for the business, and that doesn't even account for write-downs of various liabilities, most of which are associated with prior fundraising. Reportedly, the IPO prospectus details commitments totalling $518 billion for infrastructure, cloud computing, and future commitments. Until now, the figures have obscured the scale and profitability of one of the largest and most consequential AI enterprises in the world.
So, it's clear that developing and operating cutting-edge AI systems is incredibly expensive. Now, the US midterms in November are looking more and more like the likely time for Anthropic's IPO. Equity analyst and Ross Report author Ross Hendricks raised concerns about the sustainability of Anthropic's massive future infrastructure commitments. Ross posted on X, “Just a casual mention of raising half a trillion in a single year, really? Does anyone believe this? Let's say they pull $100B in equity from the IPO... and then what, $400B from the debt markets?”
Anthropic Launches its Latest AI Models
These financial information surfaced at the same time that Anthropic unveiled their new Claude Opus 5.5 and Claude Sonnet 5.5 models last week. The flagship model for complicated coding, agentic workflows, research, and knowledge work at Anthropic is Opus 5.5, which comes with a context window of 1 million tokens. Tasks demanding greater reasoning and judgement are best tackled via anthropic reasoning. This places it in a league of its own against frontier models like OpenAI's GPT-6 Astra.
Anthropic has released their newest mid-tier frontier model, Claude Sonnet 5.5. According to recent research from benchmarking website Artificial Analysis, Opus 5.5 achieved the highest possible score of 58 on its Intelligence Index when tested at full throttle. Six out of ten index benchmarks are led by the model. Opus 5.5 is just slightly higher than GPT-6 Astra. In its maximum-effort testing, Artificial Analysis ranked Sonnet 5.5 as the second-highest-scoring model with a score of 56, just two points below than Opus 5.5.
The media has reported that the introduction of Anthropic to the general market would most likely be delayed until after the US midterm elections in November. Big Tech, sovereign wealth funds, and venture capital firms have been funding the AI race up until this point, but the sale would bring public investors into the battle.
