Neoclouds are multiplying as AI labs and enterprises hunt for dedicated graphics processing unit capacity, but standing up a cluster is only the start. For smaller providers, managed services support from hardware partners increasingly decides whether that capacity stays healthy and profitable after go-live.

Specialized AI clouds are carving out a role alongside the hyperscalers, particularly where sovereign cloud requirements dictate where data lives and who controls it. A young provider cannot staff every layer of a complex GPU deployment on its own, according to Tom Richer (pictured), executive vice president of Alpha Compute Corp.

“You can cobble together a lot of different providers, but then when stuff goes wrong, the finger-pointing starts,” he said. “You just want that integrated solution provider who can ensure success, help you with all of the build-out, then all of the … managed services to keep it going healthy and to keep everything performing in line with [service-level agreements].”

Richer spoke with theCUBE Research’s Dave Vellante at the Dell Technologies New York Symposium, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed managed service support, sovereign AI infrastructure and neocloud financing. (* Disclosure below.)

Managed services support anchors a sovereign neocloud

Alpha Compute moved into AI infrastructure less than a year ago after starting out in digital assets, and it now leads with sovereign and confidential computing powered by Dell Technologies Inc. and Intel Corp. hardware. Its approach, which the company calls sovereign by design, starts at the silicon and keeps data in-country for customers subject to rules such as the EU AI Act and the General Data Protection Regulation, Richer explained.

“We start with the hardware. Think of a stack. All the way down to the chip level, we are creating enclaves,” Richer said. “We’re a Dell partner, so we use Dell technology as well as Intel technology.”

That stack is already running in production. The company’s 72-node St. Louis cluster pairs Nvidia Blackwell systems from Dell with InfiniBand networking and direct liquid cooling inside a retrofitted brownfield data center, and Alpha Compute has outlined plans for additional Blackwell clusters. Alpha Compute announced a binding term sheet for a $31.9 million nonrecourse loan facility secured by Nvidia B300 hardware.

“There’s kind of a trinity there, though. You need the off-taker, a good off-taker,” Richer said. “But also you need that kind of land, power [and] shell.”

Once a cluster is live, Alpha Compute owns the client relationship. Tickets, remediation and hardware returns run through a joint team with Dell, which has made AI factories central to its enterprise strategy. Alpha Compute’s clients to date are AI labs that want the provider managing everything below the operating system. Enterprises are likely to ask for more, according to Richer.

“With Dell as our managed services partner, we can go above the [operating system] on day two,” Richer said. “The powerful thing with Dell and their integrated approach is we can go to any level.”

That depth of managed services support also shapes how Richer sees neoclouds competing. Rather than replacing public cloud, purpose-built providers with no “noisy neighbors” or multitenant sharing will take on the most demanding and sensitive GPU workloads, he explained.

“I really see neoclouds now accompanying and fusing with public cloud for enterprise choices in that hybrid manner,” Richer said. “For those superintelligence workloads that are really demanding, enterprises can send those sensitive, intense high-performance computing workloads over to the neocloud, so they can work in concert.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the Dell Technologies New York Symposium:

(* Disclosure: TheCUBE is a paid media partner for the Dell Technologies New York Symposium. Neither Dell, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE