Enterprise infrastructure is undergoing one of its most significant evolutions in decades, driven by the need to support dynamic, data-intensive workloads. As organizations adapt to modular platforms, reimagined processes and increasingly decentralized architectures, the traditional model of information technology is being reshaped from the ground up.

During this week’s theCUBE Pod Episode, theCUBE analyst John Furrier (pictured, left) and Dave Vellante (right), talk about how industry leaders across compute, networking and storage are accelerating efforts to build enterprise infrastructure that can keep pace with emerging demands, especially those created by real-time decision-making, agent orchestration and token-driven systems. From Taiwan’s semiconductor supply chain to Dell Technology Inc.’s disaggregated architecture and Vast Data Inc.’s pivot to data platforms, a unified theme is clear: Enterprise infrastructure is becoming the battleground for competitive advantage.

“The war in the infrastructure is scaling up. That’s going to be the area to watch,” Furrier said. “Scale-out’s pretty much on a nice run. You’re seeing a lot of the clusters scaling out nicely. The action will be in scale-up.”

How new design models are reshaping enterprise infrastructure

Enterprise infrastructure is undergoing a philosophical and architectural rethink as legacy systems give way to modular, intelligent platforms. Rather than layering automation onto outdated workflows, leaders are embracing flexible, distributed designs built for real-time responsiveness, data context and high-speed adaptability, transforming infrastructure into a foundation for clarity, not complexity, according to Vellante.

“The narrative is that you don’t need all these big GPUs,” he said. “Not everybody needs the latest and greatest GPU.”

In this new model, infrastructure must support both speed and adaptability. Removing inefficiencies at the systems level and designing for logic-based decision flows opens the door to better outcomes and sustained performance, Vellante added.

“People call it the semantic layer, but sometimes that’s confused with even semantic search, and we’ve sort of been playing around with the data harmonization layer, but that’s kind of geeky,” he said. “The whole point is you take structured data, unstructured data, different query types, different storage formats and you’re able to make sense out of it all. This is a better way to do it. Let’s eliminate 10 steps.”

Taiwan’s role in the future of compute and connectivity

One of the clearest examples of this transformation is unfolding in Taiwan, with industry leaders driving advances in semicustom supercomputing and chiplet design. The infrastructure conversation now goes beyond cloud, embracing supply chain resilience, regional innovation and tightly integrated compute, memory and networking as a unified, future-ready strategy, Furrier explained.

“Jensen [Huang’s] keynote [at Computes 2025] was significant because it really marks a continuation of the leadership of the Taiwanese ecosystem of companies. That’s because the holy trinity of server storage and networking is in play,” he said. “It’s the epicenter of the computer ecosystem, and it’s a launchpad for new markets, especially in Asia.”

These developments are also enabling new deployment models, such as on-prem orchestration frameworks that integrate closely with proprietary data sets. That shift is critical for regulated industries and organizations prioritizing latency, compliance and control.

“Those enterprises, they don’t have the skillsets most of them, and they don’t have solutions coming out of the vendors. The vendors are like HP and Dell largely doing hardware and are putting the Nvidia stack on top,” Vellante said. “Nvidia wants everybody to buy the whole stack, but customers … don’t want to buy the whole stack. There’s a 12- to 18-month window for Dell, HPE, Lenovo, Supermicro, etc., to build out those solution stacks and help enterprises adopt on-prem AI.”

At Dell Technologies World, clarity amid complexity

At Dell Technologies World 2025, Dell signaled a bold shift from legacy conventions with its streamlined “Power” lineup — PowerScale, PowerProtect and PowerStore — focused on performance and consistency. But the bigger message was flexibility: Dell’s disaggregated architecture gives customers scalable, hybrid-ready infrastructure built for wherever their data works best, according to Furrier.

“The question is Vast is out-marketing Dell period, no doubt about it. They have to get their act together, and from what I saw at Dell Technology World, they are,” Furrier added. “The action at Dell is the data layer.”

Dell is positioning itself to challenge hyperscalers not by mirroring their cloud platforms, but by offering enterprise infrastructure solutions that balance on-premises control with cloud-like elasticity. This battle is being waged at the data orchestration layer, where access, latency and scale matter most, Furrier added.

“This disaggregated architecture, they are going to fight tooth and nail against Vast because Vast thinks they can run the table and they are doing well with the hyperscalers and the big customers. But they have to win the enterprise,” Furrier said. “I’d be hard-pressed to think that’s going to be uncontested by Dell.”

Alongside Dell, Vast Data is moving aggressively. The company is evolving into a unified enterprise platform with a Snowflake-like model that combines high-speed storage with low-code orchestration. Its approach aims to streamline infrastructure complexity and improve the manageability of large-scale workloads, Vellante pointed out.

“They started with the data store, which was the scalable of object parallel sort of file system. That’s kind of what they were known for,” he said. “Then they introduced this thing called InsightEngine, which is embedding microservices for that whole Symantec context enrichment … and then this thing called AgentEngine. They’re playing in that game and everything’s exposed through a unified namespace. You’ve got to be impressed with what they’ve done and the value that you’re getting.”

Watch the full podcast below to find out why these industry pros were mentioned:

Bill Tai, venture capitalist, athlete, adjunct professor
Jensen Huang, founder and CEO of Nvidia
Jeff Clarke, COO and vice chairman of Dell Technologies
Jony Ive, former SVP of industrial design and chief design officer at Apple
Pete Sonsini, venture advisor at New Enterprise Associates
Charlie Kawwas, president of Broadcom
Michael Dell, chairman and CEO of Dell Technologies
Jeff Denworth, co-founder of VAST Data
Bob Evans, former COO of Oracle
Jackie McGuire, principal analyst at theCUBE Research
Sergey Brin, former president of Alphabet
Arvind Krishna, chairman and CEO of IBM
Rob Strechay, principal analyst at theCUBE Research
Martin Kon, president and COO of Cohere
Jamie Dimon, chairman and CEO of JPMorgan Chase
Sam Altman, co-founder and CEO of OpenAI
George Gilbert, principal analyst at theCUBE Research
Brian J. Baumann, founder of NYSE Wired and director of capital markets, technology at NYSE

Here’s the full episode of this week’s theCUBE Pod:

Don’t miss out on the latest episodes of “theCUBE Pod.” Join us by subscribing to our RSS feed. You can also listen to us on Apple Podcasts or on Spotify. And for those who prefer to watch, check out our YouTube playlist. Tune in now, and be part of the ongoing conversation.

Photo: SiliconANGLE