AWS Re:Invent: New Resources to Innovate with Artificial Intelligence in the Cloud

A raft of new AI-powered cloud instances will make it easier to build and train AI models.

Bob is the managing editor of BizTech magazine.

Amazon Web Services announced several new products and features to its cloud platform designed to help organizations innovate with artificial intelligence (AI) in the cloud. The company, which has the most popular cloud platform by market share, made the announcements at AWS re:Invent, its annual show for developers, analysts, journalists and others.

The event, being held in Las Vegas, attracted about 60,000 in-person attendees, plus thousands more watching certain sessions online, according to Matt Garman, one of AWS’ original executives who became its CEO in June. “AWS re:Invent has something for everyone,” Garman told attendees during his keynote address. “But at its core, it’s a learning conference dedicated to builders, and specifically to developers.”

Click the banner below to receive exclusive industry content when you register as an Insider.

The company’s announcements this week are designed to extend what Garman described as its lead as the provider of the world’s most robust and extensive cloud computing options, many of which are found on Elastic Compute Cloud (EC2), its flagship cloud platform. Garman described a raft of new tools available to developers, especially those designed to help them build and train AI models.

“We offer more compute instances than any other provider, and it all starts with EC2,” Garman said. He noted that, in total, AWS offers 850 instance types: “What that means is you can always find the right instance type for the workload that you need. And we work hard to make sure you always have the latest technology, so you always have instances with the latest advancements from Intel, AMD or NVIDIA.”

LEARN MORE: Master the art of cloud management with CDW’s Inscape platform.

How AWS Works with (and Competes with) AI Chip Makers

Those three companies — especially NVIDIA — are the world’s most prominent producers of microchips for the AI era.

“Today, by far the biggest compute problem out there is generative AI, and the vast majority of generative AI workloads today run on NVIDIA GPUs,” he explained. “AWS is by far the best place to run graphics processing unit workloads, and part of the reason is that AWS and NVIDIA have been partnering for 14 years. In fact, NVIDIA actually chose AWS as the place to run its own AI workloads.”

On EC2, AWS will launch P6 instances on NVIDIA’s new Blackwell line of GPUs and will also launch new AMD instances. Perhaps more important from AWS’ perspective, Garman also announced the general availability of instances using AWS’ own AI chip, Trainium 2. In a release, AWS said, that Trainium 2 chips “offer 30-40% better price performance than the current generation of GPU-based EC2 instances.”

Garman said the company is now hard at work on its Trainium 3 line, which will double Trainium 2’s computing power when it becomes available next year.

AWS is by far the best place to run graphics processing unit workloads, and part of the reason is that AWS and NVIDIA have been partnering for 14 years.

Matt Garman CEO, Amazon Web Services

How AWS “Works Backward” by Starting with Customer Needs

Trainium is AWS’ effort to carve a slice of the AI chip market for itself. When the company began development in 2018 on its first-generation AI chips, it realized that that the trend toward generative AI would be big. Even so, the idea that AWS, known for cloud storage, would leap into chip manufacturing “was a pretty controversial idea,” Garman acknowledged. “It seemed crazy that we would go develop our own silicon, but we were convinced we could deliver real, differentiated value to customers.”

By 2022, it had launched its first Trainium chips. “We knew it was early, the software was early and it wasn’t going to be perfect for every workload,” Garman said. “But we saw enough interest that we knew we were on the right track.”

He described the company’s efforts in the chip business as consistent with its long history of “working backward” — asking customers what they need, then figuring out how to build it. A good example is its history in the financial services industry, Garman said.

“We went to New York in 2006 and outlined for the big banks our vision for how cloud could change how they run their technology,” he recalled. “And they said, ‘You know, we think it’s very unlikely that any of our production workloads are ever going to run in the cloud.’ They gave us a whole list of reasons: compliance, audit, regulation, security, encryption. All these things were reasons why they said they’d love to, but they’d never run on cloud. We could have said, ‘OK, that’s not going to work, let’s move on.’ But that’s not what we did. And I’d actually like to thank those big banks because they really helped us to understand what it takes to support large, regulated customers inside the cloud. And we spent the next decade ticking every one of those things off the list.”

Today, financial services are further along than most other industries in their cloud migrations, according to new research by CDW. And the vast majority are AWS customers.

To learn more about AWS re:Invent, visit our conference page. You can also follow us on the social platform X at @BizTechMagazine to see behind-the-scenes moments.