OpenAI CEO Sam Altman addressed concerns about artificial intelligence energy consumption during a panel discussion at Technische Universität Berlin on February 7, 2025, arguing that spending energy on AI research could solve humanity’s climate challenges more effectively than restricting the technology’s development.
According to Altman, “even if we have to use hundreds of megawatts or gigawatts on this problem if we can use AI to discover how to do efficient cheap fusion and then very quickly replicate the thousands of GWs of generating capacity that’s burn in carbon around the world that would be a huge win.” The OpenAI executive positioned AI energy consumption as a necessary investment in breakthrough scientific discoveries that could transform global energy production.
The discussion occurred during an event hosted by Technische Universität Berlin, OpenAI, and the Berlin Institute for the Foundations of Learning and Data (BIFOLD). Altman’s comments reflect growing debate within the technology industry about balancing AI development with environmental concerns as computational demands continue increasing.
Current AI energy efficiency metrics
Altman defended current AI energy efficiency, claiming modern language models operate with “incredibly efficient” performance per query basis. He compared AI systems favorably to human cognitive energy consumption, stating the models achieve high-quality responses using fewer watts per token than the food energy humans require for equivalent thinking tasks.
The OpenAI chief executive drew parallels to early internet concerns, noting that initial criticism of Google’s energy consumption overlooked the technology’s efficiency gains. “I remember a long long time ago when Google first came out there was a sort of moral panic because people said oh you’re like you know to ask that one question um you’re you’re using all this electricity in the data center,” Altman recalled during the panel.
According to Altman’s analysis, a single web search consumes significantly less energy than the alternative processes it replaces. He estimated that AI-powered queries use substantially less energy than driving to libraries for research, even accounting for increased usage frequency enabled by reduced friction.
[Read More…]
