Artificial intelligence initiatives — especially agentic AI-native applications — are shaking up the marriage of storage and compute.

Traditional online transaction processing workloads still demand very low latency, typically relying on row-store databases rather than object storage. But for agentic workloads the dominant constraint shifts from speed to storage scale and cost, according to Paul Copplestone (pictured), chief executive officer of Supabase Inc.

“What really is critical is [agentic workloads] generate a lot of data. What you want is extremely low cost data storage, and that’s where S3 comes in,” Copplestone told theCUBE. “What we’re seeing is this huge trend towards shifting database workloads just on top of S3 itself … This means you can get to petabytes or hundreds of petabytes of data.”

Copplestone spoke with John Furrier at AWS re:Invent, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how the need to support both traditional OLTP and new agentic, AI-native applications is creating a demand for open, synched storage environments.

AI-native applications and the shift to open storage architectures

As organizations scale generative and agentic AI, infrastructure bottlenecks are emerging as a primary constraint. A big part of that friction comes from how data is scattered across different storage systems. Using both traditional databases and S3-based analytics storage is driving demand for open, compatible storage formats that let multiple engines query the same data without constant copying, according to Copplestone.

“When you no longer need the data in your database, you delete it out and you’ve still got it in your data warehouse,” he said. “The queries that you would run are exactly in sync between the database and your data warehouse.”

Amazon Web Services Inc. is betting that the next phase of enterprise AI will involve long-running agents working across huge data sets, not just single prompts. That puts a premium on keeping low-latency databases and affordable, large-scale S3 storage working together. Supabase positions its new open PostgreSQL-plus-S3 architecture as a way to do exactly that, Copplestone explained.

“The way we see it is a lot of this kind of low latency data can get stored in Postgres,” he said. “Then anything that can tolerate higher latency, but you need a lot of it, sits in S3. Once you have these two primitives, you’ve got a complete platform.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of AWS re:Invent:

Photo: SiliconANGLE