Where the Internet Lives Anatomy of a Modern Data Center
Feb 4, 2026
Partha Ranganathan, Google engineering fellow who shaped modern data center design, explains how data centers evolved into ultra-efficient computational engines. He discusses how cloud scale and specialized accelerators boost efficiency. He explores AI-driven demands on memory, network, and chips. He also covers using AI to optimize cooling, scheduling, and hardware design.
AI Snips
Chapters
Transcript
Episode notes
Data Center As One Giant Computer
- Thinking of a data center as one giant computer simplifies design and scales efficiency across hardware, software, and facilities.
- Partha Ranganathan says this view enabled dramatic energy gains and a far more efficient cloud than predicted in the 1990s.
AI Workloads Reshape System Design
- AI workloads are math- and data-heavy and change the balance of compute, memory, and networking requirements.
- Partha Ranganathan emphasizes that AI demands a different hardware and system response than web or search workloads.
Accelerators Multiply Performance Efficiency
- Custom accelerators deliver far higher performance and energy efficiency for ML than general-purpose CPUs.
- Partha Ranganathan helped design accelerators that drive orders-of-magnitude improvements for video and ML tasks.