OpenAI’s CFO Sarah Friar set a fresh benchmark for AI investments: measuring "useful intelligence per dollar." The metric assesses the real economic impact of AI by comparing the value it generates against all associated costs, including indirect expenses like retries or human oversight.
Traditional software spending focused on user counts or licenses, but Friar argues that AI success hinges on outcomes, not just access. A thousand employees with AI tools means little if those tools don’t complete meaningful tasks effectively.
The new scorecard breaks down AI performance into four key factors: task completion rate, cost per task, accuracy, and how well the value scales without ballooning costs. This approach reflects growing concerns as enterprises face unexpectedly large AI bills, raising CFOs’ alarms and pushing demand for financial accountability.
OpenAI’s revenue growth illustrates the AI investment surge: jumping from $2 billion in 2023 to $6 billion in 2024, with forecasts above $20 billion for 2025. Friar, joining in mid-2024, realigned the company’s financial strategy to focus on proving real returns amid over $100 billion spent on infrastructure.
The compute power behind AI also surged dramatically, from 0.2 gigawatts in 2023 to 1.9 gigawatts in 2025 comparable to the output of two large nuclear reactors dedicated solely to running AI models.
This rapid growth impacts related domains like crypto, where decentralized networks such as Render and Akash compete as AI infrastructure alternatives. Finance teams in these sectors may soon need to adopt Friar’s "useful intelligence per dollar" metric to demonstrate value to enterprise clients.
