Originally posted on Data Center POST.
At the AI Infra Summit 2025, which was held at the Santa Clara Convention Center in California, gathered industry leaders to explore one of the most pressing questions facing digital infrastructure today: Will Neoclouds emerge as the next hyperscalers, or are they a transitional step in the evolution of AI data centers?
During a panel discussion, moderated by Ilissa Miller, CEO of iMiller Public Relations, the panel brought together three experts at the forefront of this transformation:
- Priya Kesavan, Principal, Azure AI Systems Strategy at Microsoft
- Ty Miller, Chief Commercial Officer at STACK Infrastructure
- Tom Traugott, SVP of Strategy at EdgeCore
Together, they unpacked how Neoclouds, which are often synonymous with GPU-as-a-Service and “AI Clouds”, are reshaping the market, competing with hyperscalers, and redefining infrastructure design
Defining Neoclouds
The panel began by clarifying what exactly Neoclouds represent. According to Tom Traugott, they can be defined as specialized providers delivering GPU capacity at scale, often without the legacy infrastructure constraints hyperscalers must contend with. This specialization enables faster adoption of technologies such as direct-to-chip liquid cooling and densified racks, positioning Neoclouds as agile disruptors.
Priya Kesavan emphasized that AI workloads differ fundamentally from general-purpose clouds. Unlike hyperscalers, who excel at managing heterogeneous workloads, Neoclouds thrive by focusing on homogeneous, AI-specific tasks. This makes them more nimble, and not just a stopgap. “AI Clouds have a sustainable business model,” she noted, adding that hyperscalers increasingly look to collaborate with Neoclouds for their agility.
What wasn’t discussed is the guaranteed privacy that Neocloud providers could offer customers by providing them access to private cloud-based AI capacity to build and customize their own private-clouds, using solely private data to operate their businesses securely.
To continue reading, please click here.
