AI and compute

OpenAI published an analysis quantifying the explosive growth in compute used for the largest AI training runs. Since 2012, the compute required for state-of-the-art models has doubled every 3.4 months—a pace far exceeding Moore’s Law, which had a 2-year doubling period. This has resulted in a cumulative increase of over 300,000x, compared to a mere 7x under Moore’s Law. The report positions compute as a primary driver of AI progress, suggesting that continued scaling will produce systems far beyond current capabilities. For developers building AI workflows, this trend has direct implications: training costs will continue to escalate, pressuring teams to optimize efficiency via model compression, distillation, or leveraging smaller models. It also underscores the importance of staying informed about hardware roadmaps and cloud pricing. The analysis is a reminder that while algorithms improve, the brute force of compute remains a dominant factor in AI breakthroughs.