Leo Leung, Oracle | Nvidia GTC 2024
summary
Oracle describes OCI as a purpose-built cloud for mission‑critical and AI workloads reporting a $6 billion run rate and expanding CapEx to meet GPU demand. The company runs model training and inference for customers including Cohere and Microsoft, embeds vector embeddings and MySQL HeatWave into the database, and supports massive instances such as a 30,000+ GPUs super cluster while partnering on Nvidia’s Grace Blackwell technology to optimize training and inference performance.
AI-written summary · about
chapters (11)
- 0:00Introduction to GTC 2024
- 0:21OCI’s cloud journey and growth
- 1:29Mission critical cloud and AI evolution
- 2:52Database as center of AI universe
- 4:21Vectors as features not markets
- 5:25Oracle’s complete AI stack
- 6:28GPU demand and silicon strategy
- 8:13Bigger AI chips and customer needs
- 9:20Training vs inference infrastructure
- 11:50Defining Oracle as hyperscaler
- 13:14Defending Oracle’s cloud quality
related talks
18mClay Magouyrk on AI, Flexibility, Speed and Security with OCI
23mScaling Up Your Cloud with Oracle Superclusters
01h 35mKeynote on Oracle's Vision and Strategy
12mTackle Any Workload with Oracle Database, AI, and Multicloud
07mCohere and Oracle’s Partnership in AI
16mCohere and Oracle Continue Improving AI for Enterprises
44mOracle Cloud Your Way: Multicloud, Dedicated, Alloy, Sovereign
24mOracle's vision for simplifying data-driven apps and analytics