Generative AI and Data Center Decentralization #1

This report is the first in a three-part series analyzing the link between the proliferation of generative AI and the geographical diversification of data center locations throughout Japan.

#1        New demand driver: Generative AI

            1.1.  Generative AI as the new key driver of data center demand
            1.2.  Primary uses and use cases of generative AI
            2.1.  Characteristics of AI data centers
            2.2.  AI data centers’ tolerance for communication latency

Summary

The use of generative AI continues to expand across a wide variety of businesses and industries in Japan. One estimate suggests that, as generative AI becomes more widely used, it has the potential to add productive capacity to the Japanese economy equivalent to 27% of GDP . However, generative AI requires vast amounts of electricity. Power consumption by data centers is projected to reach 15x its current level by 2034, at which point it will account for over 5% of the nation’s entire electricity demand.
 
Building shells and facilities of AI data centers require high-level specifications that exceed those needed by conventional data centers. At the same time, AI data centers tend to have higher tolerance for communication latency, opening up the possibility of new development in regional areas that have not previously hosted data centers.