Holger Mueller, Constellation Research, Inc. | IBM Think 2024
summary
Argues that bigger models are better because they need massive, centralized data and thus push workloads back to cloud elasticity; cites scarcity of Nvidia DGX hardware and cash prepayments at OCI as evidence of infrastructure rush. Recommends lakehouse-centered data, foundation model architectures plus RAG for enterprise apps, and lightweight on-device models for edge autonomy. Predicts agents and model integration will commoditize DevOps and favor clouds that solve egress and low-latency data access.
AI-written summary · about
