Anthropic, the developer behind the Claude artificial intelligence chatbot, is experiencing challenges in obtaining adequate computing power from Amazon Web Services, Google, and Microsoft. While these major technology firms supply the essential cloud infrastructure required to train and deploy advanced AI models, they are simultaneously competing with Anthropic in the broader artificial intelligence market and have been reluctant to fulfill all of its capacity demands.
This resource constraint emerges as Anthropic experiences rapid business expansion and prepares for a potential initial public offering. A draft public listing document indicates that the company could allocate upwards of $750 billion toward computing infrastructure in the upcoming years, which includes agreements currently under discussion.
To date, Anthropic has already pledged approximately $450 billion toward cloud computing contracts spanning the next seven to ten years. This planned expenditure features multi-billion dollar commitments with Google Cloud, Amazon Web Services, and Microsoft to guarantee the processing power required for model development and operation.
Despite these long-term arrangements, Anthropic notes that the providers have not supplied the full extent of the requested capacity. In response, the firm has turned to leasing independent data centers and partnering with smaller operators, though it acknowledges that some alternative providers possess limited operational histories and carry higher reliability risks.
During the first half of 2026, Anthropic spent nearly $9 billion on computing capacity and infrastructure. While the company generated $16.3 billion in revenue during the same timeframe, it recorded a net loss of $80.7 billion, driven largely by accounting adjustments associated with its fundraising mechanisms.
In a separate operational update, Anthropic revised its usage policy to permit the Claude assistant to conclude conversations if users engage in prolonged, unnecessary, and abusive conduct toward the AI. This adjustment stems from ongoing internal research into artificial intelligence welfare and moral status precautions.
The policy implementation is strictly limited to extreme instances of repetitive harassment and excludes standard user frustration, creative writing, or technical testing. The change has prompted broader discussions regarding human interaction etiquette with artificial intelligence systems, drawing varied commentary across the technology sector.