#577: My Dream "home lab"

#577: My Dream "home lab"

From David Bombal by David Bombal

May 22, 2026 · 28 min

About this episode

David Bombal takes listeners on a tour of Cisco's AI data center lab, discussing the technology and infrastructure behind massive GPU clusters.

Join me for an exclusive, behind-the-scenes tour of Cisco's purpose-built $20 million AI data center lab in San Jose. AI is revolutionizing the tech industry, but running massive 10,000 GPU clusters can cost up to $175 million a year—and a mere 5% network bottleneck can flush $8 million down the drain. In this video, we break down the hardware, software, and networking topologies required to build, secure, and scale massive AI data centers. I sit down with Cisco engineers to explore scale-up domains with NVIDIA H200 and AMD MI350X servers, scale-out rail-optimized topologies, and the massive Cisco Silicon One G300 ASIC delivering 102.4 Terabits per second with innovative liquid cooling. We also dive into the Ethernet vs. InfiniBand debate for scaling to 100,000 GPUs, the power of Linear Pluggable Optics (LPO) to reduce power draw, and how Cisco used AI to safely refactor 500,000 lines of legacy code. Whether you are building an AI network today or upgrading your engineering skills for the future, this is video can help you learn. Big thanks to Cisco for sponsoring my trip to Cisco Data Center and Operational Control Center in San Jose. // Will Eatherton SOCIAL // LinkedIn: /…

People in this episode

Host: David Bombal

Topics covered

  • AI data centers
  • Cisco technology
  • networking topologies
  • GPU clusters
  • data center hardware
  • software engineering

Keywords

  • AI
  • data center
  • Cisco
  • GPU
  • networking
  • hardware
  • software

Sponsors

Cisco

Mentioned in this episode

Products: NVIDIA H200, AMD MI350X, Cisco Silicon One G300, Linear Pluggable Optics (LPO)

More episodes of David Bombal

Explore listener stats, chart rankings, contacts and more on the David Bombal podcast page.