Three AI constraints: infrastructure, trust, data

Big thanks to Cisco for sponsoring this video and sponsoring my trip to Cisco Partner Summit San Diego 2025. This video is a deep dive with Jeetu Patel on why the real AI revolution is happening in infrastructure and networking, not just in GPUs or chatbots. Jeetu explains that we are massively underestimating how much AI infrastructure the world will need. Power becomes the core constraint, GPUs are the core asset, and networking is the force multiplier that lets thousands of GPUs act as one system. He walks through how we went from models on a single GPU → 4–8 GPUs in a server → racks with hundreds of GPUs (like NVL-72 with 500+ GPUs) → clusters of racks, and now “scale across” between data centers when power and real estate are scattered across different regions. The conversation then shifts to edge AI and Cisco Unified Edge: instead of doing all token generation in big data centers, some inference and token generation must move to the edge (branches, factories, hospitals, stadiums, stores) where data is created. Jeetu explains why edge devices need to be plug-and-play, remotely managed, and integrate compute, networking, security and observability in a single platform. He also introduces the idea that AI is now constrained by three big bottlenecks: • Infrastructure • A trust deficit (people don’t trust AI yet) • A data gap (models are mostly trained on human internet data, not on rich machine data) Jeetu explains how security becomes a prerequisite for productivity, not a trade-off, and describes Cisco’s work with Splunk, open-sourced time-series models, and machine data (logs, metrics, traces) to close the data gap by correlating machine data with human-generated data for better insights. Globally, he talks about the “token generation race” – how every country now cares about having enough AI token generation capacity because it directly links to GDP and national security. He cites huge infrastructure build-outs with partners like G42 in the Middle East, at gigawatt and trillions-of-dollars scale. Finally, Jeetu tackles the “AI will take my job” fear. He outlines three stages of thinking: 1. “AI will take my job.” 2. “Someone who uses AI better will take my job.” 3. “Without AI, I won’t be able to do my job.” His message to younger viewers: be excited, adopt AI as a companion, own your learning, and learn fast because AI compresses the time it takes to build skills. // Jeetu Patel’s SOCIALS // LinkedIn: / jeetupatel Website: https://www.cisco.com/ X: https://x.com/jpatel41 // David's SOCIAL // Discord: discord.com/invite/usKSyzb Twitter: www.twitter.com/davidbombal Instagram: www.instagram.com/davidbombal LinkedIn: www.linkedin.com/in/davidbombal Facebook: www.facebook.com/davidbombal.co TikTok: tiktok.com/@davidbombal YouTube: / @davidbombal Spotify: open.spotify.com/show/3f6k6gE... SoundCloud: / davidbombal Apple Podcast: podcasts.apple.com/us/podcast... // MY STUFF // https://www.amazon.com/shop/davidbombal // SPONSORS // Interested in sponsoring my videos? Reach out to my team here: sponsors@davidbombal.com // Menu // 0:00 - Coming up 0:33 - "Networking is sexy" 02:24 - Scale up, scale out and scale across explained 04:47 - Cisco and Nvidia partnership 05:55 - Cisco and G42 partnership // Addressing the AI bubble 08:11 - New Cisco Unified Edge 11:08 - Agentic AI in the future 13:05 - Huge demand for networking 13:57 - The three constraints 16:38 - AI in the real world 19:26 - How AI will take jobs away 21:38 - Conclusion Please note that links listed may be affiliate links and provide me with a small percentage/kickback should you use them to purchase any of the items listed or recommended. Thank you for supporting me and this channel! Disclaimer: This video is for educational purposes only.