OpenAI Signals Aggressive Compute Expansion in Strategic Push Against Anthropic

OpenAI is doubling down on its infrastructure strategy, explicitly targeting rival Anthropic in a recent memo to shareholders. The company characterized its competitor as "compute constrained," arguing that Anthropic is operating on a significantly smaller growth trajectory than OpenAI. The rivalry comes as both firms, with a combined valuation exceeding $1 trillion, prepare for potential initial public offerings. As they compete for dominance in the large language model (LLM) space, they face intense pressure from cash-rich incumbents such as Meta and Google. According to the memo, OpenAI is targeting 30 gigawatts of compute capacity by 2030. In contrast, the company expects Anthropic to reach approximately 7 to 8 gigawatts by the end of 2027. OpenAI asserts that its ramp-up is not only ahead of the competition but is widening over time. OpenAI describes its approach as a "compounding advantage," where superior infrastructure leads to more capable models and lower per-token costs. This strategy is intended to drive higher revenue while allowing the company to offer free tools to millions of users. Meanwhile, Anthropic continues to gain traction in the enterprise sector. The company recently unveiled "Project Glasswing," a new cybersecurity-focused AI initiative designed for a select group of corporate clients.