Nvidia's Bill Dally and Google's Jeff Dean on the Future of AI and the Importance of Inference
Tae Kim
· Key Context by Tae Kim
· March 23, 2026 at 12:06
· ⏱ 2 min read
| Read on Substack ↗
Summary
The article highlights a key observation from Nvidia's Bill Dally, who stated that 90% of the power in data centers is now consumed by AI inference, not training. This points to a significant shift in computational workloads as AI applications are deployed at scale.
•Nvidia's Bill Dally declared that AI inference accounts for 90% of data center power consumption.
•This indicates a major shift from AI model training to real-world application and deployment.
•The article frames artificial intelligence as a matter of sophisticated software and computation, not magic.