借助 Amazon SageMaker HyperPod 上的 UltraServer 加速 AI 工作负载.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.A I M 3 6 2Accelerate AI workloads with UltraServers on SageMaker HyperPodRekha Seshadrinathan(she/her)Senior Manager PMTPaulo Aragao(he/him)Principal WW Specialist

2、SA 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.The Brooklyn Bridge was the first steel wire suspension bridge and the longest in the world at the time.Brooklyn Bridge:A brief historyJohn RoeblingSteel wire 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Gene

3、rative AI assistantsGenerative AI agentsAgentic AI systemsFollow a set of rules Automate repetitive tasksAchieve a singular goalAddress broader range of tasksAutomate entire workflowsFully autonomousMulti-agent systemsMimic human logic and reasoningM O R EH U M A NO V E R S I G H TL E S SH U M A NO

4、V E R S I G H TEvolving into agentic AI4 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Rapid evolution of GenAI2022 202570%enterprises using GenAI in at least one business function*Agentic and reasoning systems drive new growth in inferenceCompute efficiency innovations in post-

5、training(e.g.,DeepSeek-R1)and inference(e.g.,disaggregated serving)$Billions invested in infrastructure growth,6 months cadence of new accelerators launch*Data source:McKinsey State of AI,Mar 2025 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AgendaChallenges with Gen AI develop

6、ment SageMaker HyperPodThe next generation of AI models and Amazon EC2 UltraServersDemonstrationCall to Action 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Access to compute is challengingOndemand capacityLong Term reservationsGuaranteed predictable availabilityLower cost that