Deutsche Bank Research Institute uses 70 years of AI history to frame today’s boom, emphasizing nonlinear progress, hardware constraints, slower enterprise adoption and valuation risk. Training compute has accelerated from 1.4x annually before deep learning to 4x, while data center electricity demand is projected to double by 2030.
AI at 70: 14 lessons from a lifetime of boom and bust
Deutsche Bank
Adrian Cox
Research
17 Pages
Key Takeaways
Training compute accelerates: AI training computation grew 1.4x annually before 2010 and about 4x annually in the deep learning era since 2010.
Cheaper compute, more demand: GPU computation costs fell more than 99% since 2006, yet data center electricity use is projected to double from 2024 to 2030.
Adoption still lags: Fewer than half of U.S. workers use AI at work, averaging 6% of working time and saving only 2% of hours.
Join our newsletter to have all of this content + Exclusive Newsletter Bonus Content delivered to your inbox every week
