4 real-world AI deployments at enterprise scale
summary
Over the last three to four years deployment replaces exploration: deep learning frameworks produce optimized libraries and graph compilers that connect down to hardware, yielding two orders of magnitude software speedups on existing machines. Second‑generation Xeon Scalable CPUs add on‑chip acceleration delivering 10x–14x improvements versus prior generations. Inference grows from ~1:1 to 1:4 training to inference and moves mostly onto CPUs, while purpose‑built ASICs—NNP‑L for training and NNP‑I for inference—are introduced to handle tensor arithmetic at enterprise scale.
AI-written summary · about
chapters (4)
related talks
09mThe AI stack from chip to analytics
50mAI acceleration at the edge
35mAI innovation begins with trust
06mUnleashing intelligence and data analytics at scale
30mAI: unleashing the next wave
28mScaling agentic AI: From infrastructure to enterprise outcomes
28mThe AI Tipping Point
26mKeynote: Artificial Innovation Accelerates Artificial Intelligence