AWS CEO Matt Garman kicked off AWS re:Invent for his first time as the hyperscaler’s leader with a clear focus on emphasizing AWS’s fundamental strengths in AI and cloud as well as its Generative AI (GenAI) offerings and capabilities, including a new suite of Foundation Models that will provide even more enterprise choice.
Garman opened his inaugural CEO address at the Venetian in Las Vegas by outlining AWS’ continued advancements in scalable cloud, flexible storage, and cost-effective AI chips, particularly its Trainium2 line (now generally available) and its next-generation Trainium3, which is due next year. AWS announced it was building “Project Rainier,” a supercomputer utilizing thousands of Trainium2 chips that will first be used by AI startup Anthropic (which recently received an additional $4 billion investment from AWS). In addition, Apple made a rare AWS appearance, with Benoit Dupin, Apple’s senior director of machine learning and AI, telling the keynote crowd that the company is utilizing AWS’s custom AI chips to power search services and is considering using them to pre-train Apple’s AI models.
Given that last year’s AWS CEO keynote was almost exclusively centered on GenAI from the start, Garman’s opening remarks on cloud and AI chips may have felt like an abrupt shift for re:Invent’s 60,000 attendees. However, when you consider the current GenAI market, it made sense on several counts. Garman, who took over the CEO post this past June, has held several AWS leadership roles including in AWS’s EC2 unit, the heart of its scalable cloud offerings, so he’s watched AWS’s core business grow from the ground up.
As Garman acknowledged, GenAI requires massive compute, storage, and processing resources, and AWS offers all of those and more as underpinnings for successful GenAI deployments. AWS’s advancements in those areas are also designed to rein in costs, boost price performance, and provide greater competitive choices for enterprise clients. Garman, using his keynote to remind the audience and the market at large of AWS’s cloud dominance and growing silicon business, is only positive and demonstrates an ongoing commitment to AWS’s fundamentals as the GenAI trend concurrently roars on.
Despite his opening remarks on the cloud’s “building blocks,” Garman’s opening keynote still shone a massive spotlight on new and expanding GenAI capabilities and offerings, which Garman said were based on constant feedback from AWS customers. “When you’re innovating, you really want to start with the customer,” he said.
In fact, Amazon CEO and former AWS CEO Andy Jassey joined Garman on the re:Invent stage to announce some of the biggest keynote news: Amazon Nova, a new family of Foundation Models for text, images, and video, which Jassey described as continuing AWS’s mission to provide “the broadest and best functionality anywhere” to empower its broad developer community and enable the creation of widespread GenAI-powered applications. AWS’s new models will compete with offerings and investments from other hyperscalers and providers. While AWS has some clear preferences, Jassey struck a more agnostic tone with the re:Invent audience. “The reality is that all of you are going to use different models, for different reasons, at different times – which, by the way, is the way the real-world works,” Jassey said.
The announcement on Amazon Nova was paired with additional new capabilities announced by Garman for Amazon Bedrock, AWS’s fully managed service building and scaling GenAI applications utilizing whatever model an enterprise needs. These new capabilities include:
- Model Distillation: Enterprises often struggle to select the right model based on their business need and cost requirements and then adapt a larger model to a smaller one for a specific use case – often because they lack the experience to do so. Amazon Bedrock Model Distillation automates a substantial part of this process, giving customers the benefits of a large model with the cost efficiencies and effectiveness of a smaller model for their intended use cases.
- Automated Reasoning: Accessible through Amazon Bedrock Guardrails, which promote Responsible AI principles while building GenAI-powered applications, this new feature is a GenAI safeguard that checks for accuracy, ensures greater transparency, and guards against AI hallucinations.
- Multi-Agent Collaboration: Last year, AWS unveiled Amazon Bedrock Agents, which allows business users with little or no coding background to create GenAI-based applications for complex tasks and digital services, for instance, specific customer support tasks such as order return. Bedrock now allows multi-agent collaboration to execute more complex workflows.
Ultimately, AWS’s latest announcements provide additional choices to developers and open further opportunities for enterprises to work with the right partner and take advantage of AWS’s ongoing advancements. As an AWS Premier Tier Partner, Persistent leverages its 30-plus years of engineering leadership as we collaborate with global enterprises on their GenAI strategies and deployments. Learn more about how we help clients harness the power of GenAI, and visit our site for additional content and insights from AWS re:Invent 2024.
