Bill Pearson, Intel & Taneem Ibrahim, Red Hat | Red Hat Summit 2026
summary
Red Hat claims to be the largest commercial contributor to vLLM and describes Red Hat AI Enterprise 3.4 as delivering GA support for Intel Xeon-based inference while shipping day‑zero models like Gemma 4. The stack includes the Open GenAI Stack (OGX) to unify competing model APIs so workloads switch models or hardware without rewrites, provides token‑budget visibility across cloud/on‑prem/edge for CFOs, and publishes quantization optimizations and model builds to HuggingFace for reuse. Agentic workloads increase CPU demand, shifting CPU/GPU ratio planning for production inference.
AI-written summary · about
chapters (14)
- 0:00Introduction to AI inference
- 1:12Inference complexity and cost
- 2:15Intel's AI inference roadmap
- 3:37Red Hat AI Enterprise explained
- 4:20CPU role in AI inference
- 6:52Red Hat's open source AI approach
- 8:38Intel's open source contributions
- 10:02Intel NVIDIA partnership
- 10:45Xeon-based AI future
- 11:51Operationalized AI governance
- 13:17Model optionality and catalog
- 14:20Day-zero model support
- 15:22AI execution best practices
- 16:08Closing remarks
