Hybrid cloud reshapes enterprise AI execution – News

Hybrid cloud is no longer just an infrastructure compromise — it’s increasingly the execution layer that determines whether enterprise artificial intelligence can move from promise to production.

As AI moves into production, hybrid cloud strategies are being reshaped by the realities of inference, distributed data and where GPUs actually live. Software solutions providers such as Hammerspace Inc. are part of a broader shift toward treating hybrid cloud as an operational model for seeing, moving and activating data across environments, not just a place to run workloads, according to Molly Presley (pictured), senior vice president of global marketing at Hammerspace.

“Organizations are trying to figure out, ‘Do we continue to do things the old way? Do we experiment with new ideas? How do we accomplish the goals that we have as things are changing so quickly?’” Presley said. “And in this idea of being ready for AI, a lot of what organizations are trying to do is figure out, ‘What is my infrastructure strategy?’”

Presley spoke with theCUBE’s Rob Strechay at the Future of Data Platforms Summit, during an exclusive broadcast on theCUBE, News Media’s livestreaming studio. They discussed how hybrid cloud is evolving from an infrastructure choice into a practical execution layer for enterprise AI, driven by inference, data mobility and cost pressure. (* Disclosure below.)

Hybrid cloud becomes an AI execution layer

Hybrid cloud strategies were once driven by cost, operational control and long-term infrastructure planning. Today, those decisions are increasingly shaped by where GPUs reside, where models run and how quickly data can be activated for AI workloads. The result is a shift toward flexibility, where execution matters more than location, Presley explained.

“Those hybrid decisions now when you’re talking about AI are more about where are my GPUs, where are the models? Where is the data I want to run,” she said. “That’s the change that I think we’re seeing as an organization … because of cloud, data is distributed in a lot of different places.”

Inference is emerging as the practical proving ground for enterprise AI value. Most organizations don’t own enough GPUs to support inference entirely on-premises, which makes cloud resources unavoidable. The challenge becomes how to move the right data to those resources without forcing a rewrite of existing hybrid strategies, Presley noted.

“What they tend to do … is move the data that’s going to be used for an inference job, so orchestrate it up to the cloud GPUs,” she said. “Then, once it’s there, they tend to just place it down in object storage in the cloud.”

Standards and cost pressure reshape AI data architecture

As AI workloads span environments and users, standards are becoming foundational rather than optional. Models often consume data they didn’t generate, which makes proprietary formats and tightly coupled systems a liability at scale. Open approaches to data access and description are increasingly necessary for AI to move beyond isolated use cases.

“In AI, the application, the data user, the model that’s using the data most likely isn’t what generated the data,” Presley said. “This is where standards are so incredibly important in AI.”

Cost pressure and procurement constraints are accelerating architectural change. With SSD shortages and rising prices, organizations are being forced to reassess existing capacity before buying more. That has renewed focus on using what already exists more efficiently, including underutilized storage attached directly to compute environments, Presley added.

“Until you use up all the capacity you’ve already bought, you don’t get to buy anymore,” she said. “That’s an architectural challenge because different applications are attached to and designed for the different storage systems.”

Here’s the complete video interview, part of News’s and theCUBE’s coverage of the Future of Data Platforms Summit:

(* Disclosure: Hammerspace sponsored this segment of theCUBE. Neither Hammerspace nor other sponsors have editorial control over content on theCUBE or News.)

Photo: News