AI company Anthropic is increasingly struggling to provide the computing capacity it needs to develop and run Claude models. According to information obtained by The Wall Street Journal and leaked from the company’s S-1 application in preparation for its IPO, major cloud service providers such as Amazon Web Services (AWS), Google and Microsoft consider Anthropic not only as a customer but also as an increasingly powerful competitor. This situation makes it difficult for the company to access the infrastructure it needs, despite allocating a high amount of resources. While it is stated that Anthropic may spend over $750 billion on contracts for cloud services, this amount also includes agreements that are still being negotiated. The company is also turning to alternative data center options due to the uncertainty in its relations with large technology organizations.
Financial data shared by Reuters and the Financial Times in late September had previously revealed the size of Anthropic’s infrastructure expenditures. However, new documents cited by The Wall Street Journal show that the issue is not limited to increasing costs. AWS, Google and Microsoft provide high-capacity computing resources to Anthropic while also developing their own artificial intelligence models and services. Therefore, these companies are among the organizations with which Anthropic directly competes, in addition to being infrastructure providers that support its operations. As the processing power required to train artificial intelligence models and meet user demands increases, the differences of interest in these commercial relationships become more evident.
Anthropic predicts spending in excess of $750 billion on cloud infrastructure
Anthropic began renting its own data centers and working with independent data center operators because the capacity it obtained from major cloud providers was not sufficient. However, the company has some reservations about the reliability of alternative providers. Leaked documents point out that the limited operating history of third-party data center operators may create operational risks. The infrastructure experience and operational capabilities that large-scale providers such as AWS, Google and Microsoft have may not be available at the same level in smaller operators. Therefore, while Anthropic tries to diversify its cloud capacity, it also has to manage new risks related to service continuity and reliability of the infrastructure.
The company’s financial data also reveals the pressure created by infrastructure requirements. According to The Wall Street Journal, Anthropic announced a net loss of $80.7 billion in the first six months of the year. In the same period, its revenues amounted to 16.3 billion dollars. In addition, the fact that only two customers account for approximately one-fifth of total sales shows that revenues are concentrated on certain customers. Although revenues from large customers contribute to the growth of the company, changes that may occur in commercial relations with these customers can directly affect financial performance. Anthropic’s rapidly increasing infrastructure expenditures and customer concentration are among the topics that investors should evaluate before the IPO.
On the other hand, the company’s recent revenue increase is also noteworthy. According to information Anthropic provided to potential investors last week, third-quarter revenue reached $16.8 billion. This figure corresponds to a 45 percent growth compared to the previous quarter. However, the rapid increase in revenues does not eliminate the impact of high infrastructure costs on the company’s financial statements. Training AI models, developing new versions, and rolling out Claude services at scale requires constant computing capacity. Therefore, in addition to revenue growth, the extent to which cloud infrastructure expenses can be kept under control will also be decisive in terms of the company’s financial outlook.
The leaked S-1 document covers not only financial and operational issues, but also the legal responsibilities that artificial intelligence may cause. In the document, which is a continuation of the statements regarding serious artificial intelligence risks that were previously revealed to the public, attention is drawn to compensation claims that may arise from the harmful behavior of the systems developed by Anthropic. The company warns that the financial liabilities it may encounter in such events may exceed its current insurance coverage. These risks are linked to the legal and financial consequences that may arise if artificial intelligence systems produce erroneous or harmful outputs. Therefore, the uncertainties that Anthropic faces go beyond the provision of IT infrastructure and also include the responsibilities that may arise from the use of the technologies it offers.
Anthropic is reportedly aiming to go public in November. Despite this, the leaked S-1 application does not mean that the public offering is finalized or will be completed on the specified date. While the growth data presented by the company to investors gives an idea about the increase in demand for Claude services, high loss figures and long-term cloud contracts show that financial risks continue. In addition, the fact that the largest infrastructure providers are also competitors in the artificial intelligence market makes Anthropic’s capacity planning more complicated. The main challenge for the company will be to keep its infrastructure costs and operational risks at sustainable levels while accessing the computing resources that will meet the increasing user and corporate customer demand.