
Switched On Silicon States: US Data Center Expansion in the AI Era
May 14, 2025
Helen Kou, Head of US Power at BloombergNEF, and Nathalie Limandibhratha, Senior Associate at BloombergNEF, dive into the booming AI data center market. They reveal the staggering forecast that data centers could consume nearly 10% of US electricity by 2035. The duo discusses how major tech firms dominate this landscape and the unique energy dynamics of AI workloads. They explore forecasting methodologies and the significant implications of these rapid developments on energy infrastructure and the transition to a lower-carbon economy.
AI Snips
Chapters
Transcript
Episode notes
Data Center Market Concentration
- Four hyperscaler companies (Google, Amazon, Microsoft, Meta) own about 50% of US data center capacity.
- These companies are increasingly building huge gigawatt-scale data center campuses.
Data Center Categories and Workloads
- Data centers are categorized by size, ownership, and workload, impacting design and power needs.
- Workloads range widely from cloud computing to AI training to crypto mining, affecting infrastructure.
AI Data Center Workload Types
- AI data centers split into AI training and AI inference workloads with different location and latency needs.
- Data centers often handle multiple workloads and can switch roles over time depending on tenants.
