ShanghaiJune 23, 2026 /PRNewswire/ — June 23, 2026, the AWS China Summit was held in Shanghai. During the summit, Matt Garman, Global Vice President of Amazon and Co-President of AWS Asia Pacific, pointed out that the continuous improvement of model capabilities and the growing maturity of Agentic engineering systems have formed a mutually reinforcing flywheel, pushing Agentic AI past the inflection point of explosion. To help enterprises seize this opportunity, AWS provides a complete Agentic AI technology stack—from AI infrastructure, models, data and knowledge, Agentic platforms, to Agent applications—empowering businesses to achieve measurable business outcomes. Additionally, Garman emphasized AWS’s consistent core advantages in data capabilities and globally trusted infrastructure, dedicated to providing customers with a cloud tailored to the needs of the AI era. At the summit, AWS officially released the “Enterprise Production-Grade Agent Development Guide White Paper,” offering comprehensive guidance from methodology to best practices, helping more enterprises avoid detours in Agent development and benefit from projects earlier.
Matt Garman, Global Vice President of Amazon and Co-President of AWS Asia Pacific
Garman stated: “We are at a historic moment—the inflection point of Agentic AI’s explosion has arrived. AI Agents are gradually becoming one of the main entities in the new era of production relations. In the past, humans were the main actors in production relations, and all technologies merely provided tools. In the future era of Agentic AI, humans and AI Agents will collaborate, using tools to complete production and value creation. This is an unprecedented paradigm shift in enterprise organization and value creation. Agentic AI is not just a technological innovation but a business transformation. AWS will leverage its advanced cloud and AI services and its unwavering ‘customer obsession’ philosophy to help enterprises navigate the challenges of AI transformation, seize historical opportunities, and become innovation leaders in the AI era.”
Matt Garman, Global Vice President of Amazon and Co-President of AWS Asia Pacific
The Inflection Point of Agentic AI Has Arrived, Driving a Reshaping of Business Models
With the continuous iteration of generative AI technology, the exploration of Agent applications is entering a new phase. On one hand, large models are making breakthroughs in reasoning capabilities, code generation, and multimodal understanding, constantly crossing new thresholds of intelligence. On the other hand, the systematic engineering capabilities of Agents—covering Prompt Engineering, Context Engineering, and Harness Engineering—that translate model capabilities into stable, deliverable business outcomes are rapidly maturing. These mature engineering practices, in turn, provide real-world feedback to model providers, clarifying scenario needs and boundary-breaking directions, collectively driving the arrival of the Agentic AI explosion inflection point. Faced with a dizzying array of AI innovations, enterprises need a clear AI landscape map to identify their needs. To this end, AWS has drawn a five-layer AI technology stack map, comprehensively covering everything from AI infrastructure to Agent applications, while integrating security, effectiveness, performance, and cost across all layers to help enterprises achieve measurable business results through AI.
The rapid development and application of Agentic AI are profoundly reshaping enterprise business models. Unlike traditional technology projects, when building Agentic AI projects, enterprise decision-making must shift from “choosing technologies and tools” to “defining the business outcomes to be achieved,” driving technology implementation in reverse. Data will also transform from static assets into core strategic assets that continuously drive Agentic AI to create value for enterprises. Additionally, as Agentic platforms become the dividing line between proof of concept and production, the collaboration of hundreds or thousands of Agents with human employees, and among Agents themselves, must rely on the empowerment and management of a unified Agentic platform. During this process, clear authorization systems, traceable decision-making mechanisms, and automated auditing capabilities—such as trust and governance mechanisms—will serve as “accelerators” for the safe, large-scale operation of Agents. Furthermore, the importance of role division and collaboration within organizational teams becomes more prominent, requiring enterprises to establish new roles, mechanisms, and responsibility systems to coordinate and regulate the collaboration between humans and Agents.
Building a Full-Stack AI Technology Stack to Accelerate Enterprise Agentic Business Transformation
To help enterprises accelerate Agentic business transformation, AWS provides a full-stack AI service across five layers: AI infrastructure, models, data and knowledge, Agentic platforms, and Agent applications. At the AI infrastructure layer, AWS offers the latest GPU instances and its self-developed Trainium AI acceleration chips, leveraging Amazon SageMaker AI to help enterprises quickly build, train, and deploy their own AI models. At the model layer, Amazon Bedrock provides unified access to multiple leading large models, supporting on-demand access to cutting-edge commercial and open-source models. Its unified API access, built-in security and governance capabilities, and stable production-ready operation enable enterprises to integrate multi-model capabilities into actual Agentic business systems. At the data and knowledge layer, AWS provides comprehensive AI-specific data services, including Zero-ETL, Amazon S3 Vectors, dedicated vector databases, Amazon Bedrock Knowledge Bases, and the newly released Amazon Context, allowing enterprise data to conveniently and effectively empower Agentic AI, transforming previously static data into AI-understandable and usable data and knowledge assets. At the Agentic platform layer, Amazon Bedrock AgentCore offers a complete set of capabilities from development to operation to iteration, supporting enterprises in uniformly and efficiently managing the full lifecycle of Agents. At the Agent application layer, AWS provides a range of out-of-the-box solutions, including Kiro, Amazon Quick, and Amazon Connect, helping enterprises in various fields such as software development, enterprise IT, knowledge workers, and customer service quickly apply Agent capabilities to actual business without building from scratch.
Based on different needs and scenarios, AWS offers enterprises three paths to apply Agentic AI products and services to solve business challenges and create value. Enterprises can directly purchase existing AWS Agent services, build customized Agents based on AWS’s full-stack AI capabilities, or obtain more diverse solutions through partner channels.
Addressing the pain point of lacking effectiveness evaluation standards when developing Agents, AWS has introduced a methodology called “Evaluation-Driven AI Agent Development Lifecycle,” covering the entire lifecycle management process, including “setting standards, development and implementation, effectiveness evaluation, deployment, continuous monitoring, and improvement cycles.” This methodology aims to help enterprises master a scientific evaluation system, thereby more reliably managing the entire lifecycle of Agents from development to operation. At this summit, AWS officially released the practical “Enterprise Production-Grade Agent Development Guide White Paper,” detailing the specific implementation steps of the evaluation-driven Agent development lifecycle. Additionally, AWS open-sourced an evaluation set and code for a simulated project on GitHub, helping enterprises quickly translate the methodology into hands-on capabilities they can try out.
Release of AWS ‘Enterprise Production-Grade Agent Development Guide White Paper’
Two Decades of Deep Innovation in Technology: Providing a Cloud Tailored to the AI Era
As a global leader in cloud computing, AWS leverages continuous innovation to provide strong, stable underlying support for its full-stack AI capabilities, committed to offering enterprises a cloud tailored to the needs of the AI era. Building on two decades of expertise in the data domain, AWS is comprehensively empowering enterprise data applications in the AI era, covering every stage from data generation, flow, aggregation and governance, to usage. Its open data architecture enables data to be written once and used everywhere; its excellent data foundation combines extreme performance with cost-effectiveness, reducing vector storage costs by up to 90%; and its trusted data capabilities ensure end-to-end unified governance from data, models, to Agents.
Furthermore, AWS’s globally distributed infrastructure features three key characteristics: security and trustworthiness, stability and reliability, and efficiency and cost-effectiveness. In terms of security and trustworthiness, AWS achieves hardware-level physical security and isolation through the Nitro System, with end-to-end encryption and network isolation, supporting 143 security compliance certifications globally. In terms of stability and reliability, AWS builds cloud infrastructure with exponentially leading overall availability in the industry through designs with at least three independent Availability Zones per region, dedicated transoceanic cables, and a product R&D team that fully practices DevOps, providing a stable and reliable operating environment. In terms of efficiency and cost-effectiveness, AWS comprehensively optimizes cost-performance through services like self-developed chips, helping enterprises achieve efficient output with reasonable investment.
Today, AWS has helped numerous enterprises accelerate Agentic AI applications. XPeng Motors has built an enterprise-level AI programming and Agentic work platform called “Lingxi” using AWS’s leading AI and cloud technologies and services, enabling zero-code delivery and covering the entire R&D lifecycle. Since its launch, the platform has achieved an AI code coverage rate of over 70% in the AI application center. Insta360, leveraging its self-developed AI capabilities and a decade of imaging technology expertise, relies on AWS’s five-layer Agentic AI architecture to provide a cloud-based “one-stop intelligent video creation” capability—Moment Pro. This enables the entire process from shooting to high-quality output in less than a minute without any manual editing, significantly enhancing the user experience of its cameras. Moonshot AI’s Kimi, using AWS’s global infrastructure and leading AI technology, provides model services to global enterprises, serving corporate clients and accelerating its own global business expansion. Cheetah Mobile is among the first Chinese enterprises to run production-grade Agents on Amazon Bedrock AgentCore. Its launched EasyClaw overseas enterprise edition, leveraging AWS’s computing power and technology foundation, flexibly schedules multimodal models to achieve full-chain AI empowerment for cross-border e-commerce advertising—from content production, ad optimization, to data review—reshaping traditional overseas marketing models.