In Good Company with Nicolai Tangen Arista Networks CEO: The AI Infrastructure Boom, Power Limits, and What’s Next
33 snips
Dec 31, 2025 
Jayshree Ullal, CEO of Arista Networks, shares her journey from electrical engineer to leading a crucial player in AI networking. She discusses how Arista's high-performance infrastructure supports the demanding needs of AI, emphasizing that power constraints, not hardware, are today's biggest challenge. Ullal highlights the evolution of networking toward AI-specific needs and explains the difference between AI and traditional cloud traffic. She also reflects on Arista's growth, its customer-focused culture, and the importance of adaptability in this fast-paced tech landscape.
AI Snips
Chapters
Transcript
Episode notes
Networking As The AI Information Highway
- Arista builds high-performance, highly automated networks that connect accelerators, users, and storage.
- Its extensible OS (EOS) unifies diverse data types and powers mission-critical scale and reliability.
AI Traffic Is A Different Class
- AI traffic is qualitatively different: far higher fidelity, intensity, durability and peak bursts than cloud front-end traffic.
- Arista brought Ethernet/IP standards into back-end AI fabrics that previously relied on isolated fabrics like InfiniBand or PCIe.
Power Limits, Not Chips, Are The Bottleneck
- Power is the main bottleneck for scaling AI data centers, not just compute or networking hardware.
- Finding sites with sufficient power and space can take three to five years today.