Article

OpenAI, led by Sam Altman, anticipates $17 billion in losses for 2025, with cumulative needs potentially reaching $150 billion before achieving breakeven, expected not before 2030. In 2024, OpenAI generated approximately $3.5 to $4 billion in revenue but incurred significantly higher costs, primarily in computing and infrastructure, with over 60% allocated to GPUs, dedicated data centers, electricity, and high-capacity networks.

Training a frontier model costs between $1 billion and $2 billion per cycle. Additional costs of 15-20% stem from cloud services, cooling, redundancy, and global distribution. The remaining budget is spent on research, specialized personnel, security, compliance, and governance.

The consumer market, with subscriptions around $20 monthly, generates minimal revenue per response, highlighting the economic challenges. In contrast, AI's value in business contexts can significantly exceed operational costs, suggesting that enterprise contracts and critical process integration are where profitability lies.