AWS leads open-source momentum in AI infrastructure – News

The enterprise cloud is undergoing its most transformative phase yet — and AI infrastructure is being reshaped from the stack’s foundation to its interface. That was the consistent theme threading through theCUBE + NYSE Wired: AI + Cloud Leaders Media Week.

Held at the New York Stock Exchange, the event featured enterprise priorities across AI infrastructure, agentic frameworks, data orchestration and modernization. From open-source momentum to mainframe migrations, Amazon Web Services Inc. emerged as a central force, but it wasn’t the only player in the spotlight.

“I think Amazon is again showing their open-source chops by contributing to both projects,” said theCUBE Research’s Rob Strechay (pictured), referring to AWS’ dual support of the model context protocols and Agent2Agent protocols. “I think that’s going to be key to them long term.”

Strechay and theCUBE’s John Furrier closed out the week with a candid post-event wrap-up of theCUBE + NYSE Wired: AI + Cloud Leaders Media week, breaking down what theCUBE team heard in executive interviews with AWS, Intel Corp., Nutanix Inc., Caylent Inc. and others. Their exclusive conversations aired live on theCUBE, News Media’s livestreaming studio.

AI infrastructure meets cloud control

Amazon’s influence across open-source AI infrastructure was a top story of the week, according to Strechay. One of the event’s focuses was AWS’ backing of both MCP and A2A standards. This move deepens the company’s ties to the Linux Foundation and mirrors the early Kubernetes-era push to unify fragmented developer ecosystems.

“You had Swami [Sivasubramanian] up there talking about A2A, and I think it’s not an either/or type of thing,” Strechay said. “Amazon realizes that; Google realizes it by donating A2A to the Linux Foundation. What you see is that Amazon is, again, showing their open-source chops by contributing to both projects.”

This investment in interoperability has downstream consequences for data platforms, particularly as multi-agent frameworks become increasingly dominant, according to Strechay. Artificial intelligence doesn’t operate in a vacuum; it requires curated data, layered orchestration and frictionless catalog integration. That’s where AWS’ services, such as Bedrock, SageMaker and even S3 Vectors storage, enter the picture, along with enterprise strategies from Snowflake Inc. and Databricks Inc. that are racing to align.

“You look at some of the Aurora PostgreSQL stuff that they’re doing, EDB, another open-source Postgres on Kubernetes,” Strechay said. “People are looking at how they play in this new ecosystem, where the framework of agentic is sitting above them. I think that goes back to Swami talking about AWS S3 buckets for vectors … now we’re talking about tiering vectors, not just tiering storage.”

AI’s growing complexity is also reshaping AI infrastructure expectations. Legacy storage players must evolve or risk fading into irrelevance. Metadata-rich catalogs, open table formats and tighter proximity to compute are now essential. There are parallels between AWS’ modern control plane efforts and earlier innovations from VMware by Broadcom and Nutanix, which moved beyond basic storage into integrated data platform capabilities, according to Strechay.

“Now you have people looking at the catalogs as that control space and really that metadata semantic layer inside the catalogs and across catalogs,” he said. “Hence, you’ve got people such as Salesforce Inc. announcing their intentions to buy Informatica … that’s to unlock that walled garden to go cross-catalog.”

Caylent Inc., a key event participant, emerged as a bellwether for cloud-native readiness. The company’s strong uptick in database migrations is a signal that enterprises are repositioning data architectures in anticipation of AI growth, according to Furrier. That momentum is tied to AWS’ new AI-powered mainframe modernization push, Strechay added. Both cloud providers and global system integrators, such as Deloitte, must deliver more than just tooling.

“AWS talked about AWS Transform … where they’re using AI to move off mainframes,” Strechay said. “This is all the way down into the database level. It’s still not easy, and especially, you want to get your data in order before you do AI, because once you’ve plumbed it, you’re kind of screwed.”

There’s no AI transformation without cloud modernization and no competitive edge without execution, according to Strechay and Furrier. From open-source protocols to enterprise processing mapping, the companies moving the fastest are those turning experimentation into production at scale.

Here’s the complete video interview, part of News’s and theCUBE’s coverage of theCUBE + NYSE Wired: AI + Cloud Leaders Media Week event:

Photo: News