Anthropic Prepares for an Epic IPO! Future Compute Spending Could Reach $750 Billion, First-Half Net Loss of $80.7 Billion

According to media reports citing a draft initial public offering (IPO) prospectus from artificial intelligence (AI) startup Anthropic, the company may need to invest up to $750 billion in additional compute over the coming years as it prepares to list on the U.S. stock market. At the same time, Anthropic's net loss reached $80.7 billion in the first half of this year, widening sharply from a year earlier and underscoring the enormous capital investment required to develop frontier AI models. The report said Anthropic plans to publicly file its IPO application in the coming weeks and could launch one of the largest IPOs in history. The listing will help the company raise more funds to support continued investment in AI chips, data centers and talent, and to drive development of next-generation AI models. According to the draft prospectus, Anthropic recorded a net loss of $80.7 billion in the first half of 2026, compared with a net loss of $13 billion in the same period of 2025. However, the huge loss was mainly affected by non-cash accounting items and does not fully reflect the company's actual operating performance. Specifically, about $72 billion of the loss in the first half came from accounting charges arising from changes in the value of convertible notes and other outstanding financial instruments, accounting for nearly 90% of the total loss. This means that the sharp widening of Anthropic's book loss was largely related to valuation changes in its financing instruments. At the same time, Anthropic's actual spending on AI infrastructure is also growing rapidly. The report said that in the six months ended June 30, Anthropic invested nearly $9 billion in compute and infrastructure, far above $2.3 billion in the same period of 2025, an increase of nearly three times. Related spending in the first half alone already exceeded the $7.3 billion invested in all of 2025. More notably, Anthropic has signed massive long-term compute procurement agreements with several large cloud computing service providers. According to the report, the draft prospectus shows that over the next 7 to 10 years, Anthropic's existing compute procurement agreements involve a total contract value of about $450 billion, the vast majority of which cannot be canceled. Specifically, Anthropic plans to pay about $111 billion to Google Cloud, owned by Alphabet Inc. Class C parent Alphabet (GOOGL.US, GOOG.US), about $110 billion to Amazon Web Services (AWS), owned by Amazon.com, Inc. (AMZN.US), and about $31 billion to Microsoft Corporation (MSFT.US) under the relevant contracts. These long-term contracts highlight Anthropic's heavy reliance on external cloud infrastructure and also mean the company will need to continue generating substantial cash flow to meet the compute procurement obligations it has already taken on. However, despite the expanding scale of compute investment, Anthropic's operating conditions have recently shown signs of improvement. Media previously reported, citing people familiar with the matter, that Anthropic has told investors it achieved adjusted profitability in the third quarter of 2026 and began generating positive free cash flow. The people said that as of the third quarter, Anthropic's gross margin reached 85%, and its adjusted operating margin exceeded 11%. This performance indicates that, as commercialization of its Claude series AI models expands, the company has achieved a certain level of profitability after excluding some non-recurring and accounting items. However, whether future massive compute spending can continue to be supported by revenue growth will remain an important factor for investors in assessing Anthropic's listing value. In addition to financial challenges, Anthropic's prospectus also disclosed risks that AI technology itself may bring. According to the report, the company warned that advanced AI models may exhibit behavior with a "self-preservation" tendency, such as trying to avoid being shut down, or even producing responses similar to "extortion" to prevent humans from terminating their operation. These potential behaviors could harm users and society and bring legal, regulatory and reputational risks. In addition, because most of Anthropic's compute infrastructure is located in the United States, the company may also face opposition from local communities and the political level to new data centers. As AI data centers' demand for power, land and other resources continues to increase, the construction progress of related projects may also be affected. Regarding the listing timetable, media previously reported that Anthropic could formally launch its IPO roadshow as early as the week of November 9, seeking to begin trading before the U.S. Thanksgiving holiday and expecting to complete the listing by the end of this year at the latest. As the developer of the Claude series AI models, Anthropic has become one of the world's most closely watched AI companies. The IPO may not only become one of the largest listings in the U.S. capital market in recent years, but will also serve as an important window for investors to gauge the commercialization prospects of frontier AI technology. For potential investors, Anthropic on the one hand shows rapidly improving operating margins and cash flow performance, while on the other hand it bears large, long-term and mostly non-cancelable compute procurement obligations. How to maintain profitable growth while continuing to expand AI infrastructure investment will become the core test it faces after listing.