The origin of the term neocloud remains foggy, but they arrived in a flash. Neocloud companies sprang up like wildflowers in 2023 and 2024 led by state-of-the-art, GPU-powered crypto-mining data centers that switched into AI factories. They rose after traditional hyperscalers went on a buying spree that triggered an ongoing global chip shortage. While they were a quick fix for big AI model builders and early GenAI innovators, these “bare metal” businesses are playing a critical role in the AI boom.
A neocloud (or newcloud) is a specialized cloud infrastructure provider that supports high-performance computing (HPC), artificial intelligence (AI), and machine learning (ML) workloads. They focus on GPU-as-a-Service (GPUaaS) and Bare-Metal-as-a-Service (BMaaS), delivering raw, scalable GPU computing power.
By January 2026, the neocloud market was pegged by one research firm at $35.22 billion and projected to reach nearly $240 billion in five years. However, some analysts suggest that neoclouds may be temporary because they’re merely filling a supply chain crunch. They argue that as hardware availability normalizes and traditional cloud service providers evolve, neoclouds will get squeezed. They face the same risks that independent bare-metal operations faced in the early days of cloud computing, when larger, general-purpose cloud providers caught up to meet market demand, according to Steve McDowell, Principal Analyst at NAND Research.
"The neoclouds today…look a whole lot more like an advanced rack space," McDowell told The Forecast, highlighting the structural vulnerabilities of the current model.
"They’re running full CPUs, full clusters. Do neoclouds exist in five years? They rose up because of a GPU scarcity problem that may not exist as we move forward."
To survive, McDowell said neoclouds will need to evolve dramatically “as we move into the age of inference.”
Look Back and Ahead
In late 2022, the cryptocurrency marketplace was in tatters. The so-called “Crypto Winter” had driven the price of Bitcoin down by around 75%, and an untold number of altcoins and NFTs were rendered essentially worthless, almost overnight.
Many over-leveraged crypto speculators who had hoped to cash in on the digital gold rush were washed out. But the infrastructure providers, who were suddenly left with racks full of largely idle GPUs, saw a new opportunity on the horizon: the burgeoning AI boom.
“These people had the capability to run giant data centers built for high-performance computing,” said Paul Updike, a senior director of technical marketing engineering at Nutanix, in an interview with The Forecast.
“It turns out, the infrastructure and hardware engineering skills you need for crypto are the exact same infrastructure and hardware engineering skills you need for AI.”
The result of this pivot was the emergence of neoclouds. These specialized cloud providers offer GPU capacity for training, fine-tuning, and running AI models, and often position themselves as lower-cost alternatives to hyperscalers such as Microsoft Azure and AWS. Also, unlike hyperscalers, neoclouds focus narrowly on high-performance AI compute rather than offering a broad menu of enterprise cloud services.
Some observers have dismissed neoclouds as a fad that is destined to die out once the supply of GPU infrastructure begins to keep pace with supply. But Updike sees them as an early stage of an evolution to more mature AI-first datacenters and cloud providers, capable of supporting production inference, multi-tenancy, and enterprise-grade AI operations.
“Neoclouds started out selling an environment, and all customers cared about was whether it was capable and secure,” Updike said.
“But once you switch away from training, and you’re running an AI-first business, you need these enterprise capabilities, and that means that neoclouds will need to become real service providers. That’s a different problem to solve, and it requires a software infrastructure that is different, as well.”
Extinction or Evolution?
While some think that neoclouds will largely disappear over the next few years, others believe that they have the potential to disrupt the entire cloud computing industry.
“Hyperscalers are at real risk of losing market share as more enterprises switch to lower-cost, specialized competitors,” wrote tech analyst David Linthicum in an April 2026 article. “Without a meaningful course correction—most importantly, a drastic reduction in AI-specific pricing—AWS, Microsoft, and Google face the prospect of ceding entire segments of the market to their smaller rivals.”
Linthicum went on to note that hyperscalers will likely buy up promising neoclouds. Still, he added, these new infrastructure providers have already forced “a long-overdue reckoning on price, performance, and customer service” while giving customers more choices.
Beyond cost and GPU access, Updike noted, neoclouds may also appeal to companies with data sovereignty concerns.
“Localized neoclouds provide a level of sovereignty that you don’t have with the hyperscalers,” he said.
Matt Kimball, a principal datacenter analyst covering servers and storage for Moor Insights & Strategy, told The Forecast that he was initially skeptical of neoclouds but now sees a future for them.
“I see the path,” he said. “They built a very specific infrastructure for a very specific set of customers and use cases, and it works really well for them. And if you believe that inference is going to be the size that it’s going to be, they’re really well prepared to do that better than the major [cloud service providers] today.”
Still, Kimball said, he expects the space to shake out somewhat over the next half-decade.
“I think there’s something like 200 neoclouds right now that are being tracked,” he said. “Five years from now, there aren’t going to be 200 neoclouds. You’re going to see a lot of these fade away.”
Taking the Next Step
To fully mature into AI-first data centers, Updike said, neoclouds will need to develop true multi-tenancy, as well as the enterprise-grade software infrastructure needed to support production inference workloads. He likened the current neocloud model to a beach house rental, with one customer running a large training job to completion across a dedicated environment. The shift to inference requires something closer to an apartment building, with multiple tenants sharing the same infrastructure simultaneously.
“In the world of inference, where you’re actually running the models and doing the work, you’re now in a space where you’re not going to consume as many nodes,” Updike said.
“You may even burst into nodes and come back down out of nodes. It’s a lot of jobs running simultaneously that you have to manage and schedule and make sure that they’re going to play nice together.”
Neoclouds currently operate somewhat like colocation datacenters, according to McDowell.
“They’re renting full clusters, and it’s mostly bare metal,” McDowell told The Forecast. “I don’t know that they're sharing a lot of resources, but as we move into the age of inference, they're going to have to solve for that.”
According to McDowell, Nutanix is well-positioned to fill this gap.
“Neoclouds have a vast number of resources, and they are going to need a control plane that helps manage workloads across those resources. That’s what Nutanix excels at.”
Related:
Editor’s note: Learn about the Nutanix Enterprise AI solution and AI Gateway service that provides a unified, secure inference endpoint lets enterprises use cloud-hosted models (and token credits) alongside private LLMs with consistent authentication, observability, and token-based rate limiting.
Calvin Hennick is a contributing writer. His work appears in BizTech, Engineering Inc., The Boston Globe Magazine and elsewhere. He is also the author of Once More to the Rodeo: A Memoir. Find him on LinkedIn.
Ken Kaplan contributed to this story. He is Editor in Chief for The Forecast by Nutanix. Find him on X @kenekaplan and LinkedIn.
© 2026 Nutanix, Inc. All rights reserved. For additional information and important legal disclaimers, please go here.
Related Articles
