Scaling agentic AI: From infrastructure to enterprise outcomes
summary
Argues enterprises move from experiments to production by building full‑stack AI infrastructure using NVIDIA Dynamo and Nemotron 3, standardizing GPU, networking and storage into composable cloud stacks with API‑driven, prebuilt templates and one‑click provisioning. Demonstrates autoscaling from one node (8 GPUs) to 512 nodes (~5,000 GPUs), enforces data residency and guardrails across regions, and emphasizes that most orchestration still runs on CPU so platforms must co‑provision CPU + inference clusters for compliant, cost‑efficient AI deployments.
AI-written summary · about
chapters (8)
- 0:00Welcome
- 2:43Cloud-native to AI-native apps
- 6:54AI infrastructure super cycle
- 11:53Abstracting infra for devs
- 16:09AI-native hotel apps example
- 21:08Automating GPU clusters
- 24:20Global GPU deployment
- 26:47Conclusion
related talks
28mScaling agentic AI: From infrastructure to enterprise outcomes – Kevin Cochrane
52mSimplifying the adoption of generative AI for enterprises
22mAI @Scale
12mDigital & Data/AI Strategy and Transformation
12mKeynote: AI Infrastructure
04m4 real-world AI deployments at enterprise scale
20mAccelerating AI pipelines w/ NVIDIA on AWS: An autonomous driving case
28mDeveloper Essentials for the New Era of AI