Subvolts take Nvidia's CTO explains the complex hardware needed to build massive AI systems. Oct 28, 2025 - Developers
- Learn about the networking and hardware challenges in building and scaling large AI systems for training and inference.
- Business
- Understand how infrastructure limitations and advancements from companies like Nvidia impact the feasibility of deploying advanced AI.
This is a solid interview with a key executive from a leading AI hardware company.
Nvidia CTO Michael Kagan: Scaling Beyond Moore's Law to Million-GPU Clusters
Audio streams from Sequoia Capital. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Nvidia CTO Michael Kagan joins Sequoia Capital's Training Data podcast to discuss scaling AI infrastructure. He shares insights on building million-GPU clusters, highlighting the critical role of network performance alongside compute power. Kagan also touches on the shift to inference workloads and his vision for AI's future.
Imagine building a giant brain for AI. This episode explains the super-fast connections and massive computer power needed, not just the brain cells. It matters because this tech lets AI do amazing new things.
This episode offers a deep dive into the infrastructure challenges powering today's AI advancements. It's for anyone interested in the hardware and networking behind large-scale AI, including engineers and business leaders. Understanding these scaling issues is crucial as frontier labs like Google Gemini and OpenAI push the boundaries of model size and deployment, directly impacting their ability to deliver advanced AI services.
- HostSequoia Capital
- GuestMichael KaganCTO, Nvidia
- Guest: Michael Kagan, CTO of Nvidia
- Title: Nvidia CTO Michael Kagan: Scaling Beyond Moore's Law to Million-GPU Clusters
- Show: Training Data (an official AI show)
- Length: 42 minutes
- Event: Recorded live at Sequoia’s Europe100 event
Recorded live at Sequoia’s Europe100 event: Michael Kagan, co-founder of Mellanox and CTO of Nvidia, explains how the $7 billion Mellanox acquisition helped transform Nvidia from a chip company into the architect of AI infrastructure. Kagan breaks down the technical challenges of scaling from single GPUs to 100K and eventually million-GPU data centers. He reveals why network performance—not just compute power—determines AI system efficiency. He discusses the shift from training to inference workloads, and his vision for AI as humanity's "spaceship of the mind," and why he thinks AI may help us discover laws of physics we haven’t yet imagined.
Hosted by Sonya Huang and Pat Grady
- Moore's Law
- Michael Kagan
- The observation that the number of transistors on a microchip doubles about every two years, leading to increased computing power.
- GPU
- Michael Kagan
- Graphics Processing Unit, a specialized electronic circuit designed to rapidly manipulate and alter memory to accelerate the creation of images in a frame buffer intended for output to a display device. They are now crucial for AI training.
- inference workloads
- Michael Kagan
- The process of using a trained AI model to make predictions or decisions on new data, as opposed to training the model itself.
1 h 08David Silver and Lasse Espeholt of Ineffable Intelligence: A Superlearner for the Age of Experience
WithDavid Silver, Lasse Espeholt
1 h 041 h 0552 minMaking Cities Awesome: Peregrine’s Nick Noone & Ben Rudolph
WithNick Noone, Ben Rudolph
Questions people ask
- What is Nvidia's role in AI infrastructure?
- Nvidia, through acquisitions like Mellanox, has transformed into an architect of AI infrastructure, moving beyond just being a chip company.
- What are the main technical challenges in scaling AI?
- Scaling involves building massive GPU clusters, and network performance is as critical as compute power for AI system efficiency.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Sequoia Capital.