No 1-on-1 meetings: How Jensen Huang runs NVIDIA | Lex Fridman Podcast Clips

Jensen Huang describes NVIDIA's unique approach to building complex AI systems through what he calls 'extreme co-design.' This involves optimizing across multiple layers simultaneously: software architectures, chips, systems, system software, algorithms, and applications. The challenge extends beyond just CPUs and GPUs to include networking chips, switches, power delivery, and cooling systems, all of which must work together efficiently in power-hungry AI systems. Huang emphasizes that company organization should reflect the product being built, criticizing the standard 'hamburger' organizational charts used by most companies regardless of their output. His management structure reflects this philosophy - he has 60 direct reports, mostly with engineering backgrounds, who are experts in different areas like memory, CPUs, optics, GPUs, and algorithms. Rather than conducting one-on-one meetings, which would be logistically impossible, he uses group problem-solving sessions where all relevant experts participate. This approach works because when discussing any component (like cooling or networking), other team members can immediately identify conflicts or dependencies with their areas (power distribution, memory, etc.). Team members are expected to know when to contribute and are held accountable if they miss opportunities to provide valuable input. This collaborative approach enables the extreme co-design that NVIDIA requires to build integrated AI systems.