The neoclouds that built their businesses renting scarce GPUs are racing up the stack, trading raw compute for token-based inference, managed AI platforms and hybrid deployments that keep enterprise data close to home.
Neoclouds are cloud providers built specifically for AI and high-performance computing. They gained traction by acquiring and renting access to scarce GPUs when demand for accelerated compute was exploding. Now those same providers are seeking new ways to serve the market.
"They engaged in a speculative market," said Paul Updike, senior director of technical marketing engineering at Nutanix. "They bought billions of dollars' worth of GPUs, knowing that AI was going to be important. Now they're figuring out how to sell that capacity."
Only a few months ago, industry observers wondered whether neoclouds might be a temporary phenomenon.
"Do neoclouds exist in five years?" analyst Steve McDowell asked during a May 2026 interview with The Forecast. "They rose up because of a GPU scarcity problem that may not exist as we move forward."
But so far, the market is expanding rapidly. According to Synergy Research Group, neocloud revenues exceeded $25 billion in 2025, with Q4 revenues up 223% year over year. Synergy predicts the neocloud market will approach $400 billion by 2031, representing a sustained 58% compound annual growth rate.
Gartner predicts that neocloud providers will capture 20% of the $267 billion AI cloud market by 2030, pointing to performance, data sovereignty, competitive pricing and flexible deployment options as key differentiators. Neoclouds’ ability to procure and make available new hardware is attractive as more enterprises expand their use of AI.
In April 2026, financial news cable channel CNBC depicted neoclouds as both the most explosive growth sector and the riskiest edge of artificial intelligence investing today. Financial analysts point to their bare-metal AI workloads costing roughly 25% to 35% of prices from traditional hyperscalers. But unlike hyperscalers such as Microsoft Azure and AWS, which offer broad catalogs of general-purpose cloud services, neoclouds have focus on AI-optimized infrastructure and high-performance workloads, offering GPU-as-a-Service and other accelerated compute services.
How Are Neoclouds Expanding Beyond GPU Rentals?
While neoclouds are criticized for relying on a small group of tech giants and top-tier AI labs such as OpenAI, Meta and Anthropic, recent developments indicate some neoclouds are expanding their offerings to grow their customer base. Some are drawing big investments from chipmakers and building partnerships with hardware and software makers as they build managed platforms and on-demand inference services to their core GPU offerings.
Neoclouds are solidifying their role in the AI landscape by drawing investment from chipmakers and building closer relationships with them. In late 2024, AMD Ventures co-led a $333 million funding round for Vultr, a cloud provider that specializes in high-performance GPU and AI infrastructure. Vultr recently announced plans to make AMD's new Helios rack-scale systems and Instinct MI455X GPUs available through its cloud platform.
Capacity and supply constraints are driving enterprises toward specialized providers, according to Brayden Mahdavi, go-to-market lead and business manager for the neocloud segment within AMD's Compute and Enterprise AI group.
"They're so nimble and lean that they can help productize our technology in a way that is favorable for our end customers, who are facing challenges around total cost of ownership and other performance challenges," he said in an video interview with The Forecast.
Neocloud providers are giving customers more options for accessing and deploying AI infrastructure. A recent announcement of a partnership between Nutanix and neocloud provider ChronoScale mentioned the vendor's GPU-as-a-Service offering, which might be considered a "traditional" neocloud model. But the announcement also highlighted an offering called ChronoScale Token Factory, which provides pre-paid inference tokens backed by open-source models. ChronoScale Foundry, meanwhile, provides a managed platform for building, running and governing agentic AI workflows locally, with agents and enterprise data remaining inside the customer's boundary.
The marketplace is still pinning down its vocabulary as offerings evolve, with sometimes confusing results for customers. For example, the Token Factory offering, which provides on-demand inference for "bursty" and experimental workloads, is very different from the more generic industry term "AI factory," which describes the underlying compute architecture.
While the term "AI foundry" is broadly used to refer to on-premises, cloud or hybrid environments for developing and operating AI workloads, ChronoScale Foundry more narrowly refers to a specific AI platform designed to run inside the customer's own environment.
ChronoScale itself is extremely young. The company was established in May 2026, when Applied Digital Corporation announced the closing of a deal to contribute its cloud business to EKSO Bionics Holdings. The cloud business became a subsidiary of EKSO, which renamed itself ChronoScale Corporation. The company was formed to operate as an accelerated compute platform, purpose-built for AI training, inference and high-performance computing.
"The reality is that we're trying to fit all of this stuff into our heads, and it's getting a little confusing," Updike said. “Things in this space are moving faster every day.”
What Does It Mean for Neoclouds to Move Up the Stack?
Moving up the stack means neoclouds are no longer just renting GPUs; they are building managed platforms, on-demand inference services and enterprise partnerships that wrap specialized AI infrastructure in familiar management tools.
While the terminology is sometimes slippery, the reality is clear: Neocloud providers are offering customers new and different ways to run and manage their AI workloads.
"They're still offering GPUs, but they're doing more than that now," Updike said. "They're moving up the stack."
The Nutanix-ChronoScale partnership is one example. ChronoScale plans to use Nutanix software across its AI infrastructure platform to support functions including customer onboarding, tenant management, service automation, virtualized infrastructure and managed Kubernetes environments.
For customers, the planned offerings span several consumption and deployment models, including GPUaaS for predictable workloads, the on-demand Token Factory offering for less predictable workloads and the on-premises ChronoScale Foundry offering. Nutanix's role is in part to bring familiar enterprise infrastructure capabilities to highly specialized AI systems, Updike said.
"We're treating this exotic stuff as if it's normal," Updike said. "We're normalizing the exotic AI factory."
How Will Enterprises Govern AI Across Multiple Environments?
As enterprises deploy agentic AI across on-premises systems, neoclouds and other environments, consistent governance and a single control plane become essential to managing security, access and policy, explained Updike.
That sort of consistency becomes more important as enterprises move from relatively contained generative AI to the deployment of many autonomous agents. That shift puts enormous pressure on IT teams and their governance capabilities. According to Deloitte, 74% of companies plan to deploy agentic AI within the next two years, but only 21% report having a mature governance model for autonomous agents.
Unlike chatbots that merely produce answers, AI agents may interact with enterprise applications and data, make decisions and take actions that have a real business impact. As organizations deploy these systems across their own infrastructure, neoclouds and other environments, IT teams will need ways to consistently apply security policies, access controls and governance across environments.
The Nutanix-ChronoScale partnership is designed to address this challenge by bringing Nutanix offerings such as Agent Gateway and Private Inferencing to ChronoScale's AI capacity, with the goal of providing customers with a single control plane spanning on-premises and ChronoScale environments.
Evolution in the space is ongoing and it remains unclear which services and delivery models, if any, will emerge as an industry standard. For the moment, though, GPU rentals have become just one offering, alongside emerging models like on-demand inference service, managed platforms and partnerships designed to connect specialized AI infrastructure with other components of enterprise IT.
"We're watching all of that shake out in real time," Updike said.
Related:
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.
© 2026 Nutanix, Inc. All rights reserved. For additional information and important legal disclaimers, please go here.
Related Articles
