AI Business

Enterprise AI Demands Full Cloud Stacks, Not Just GPU Capacity

As companies move AI agents into production, infrastructure providers face pressure to deliver complete platforms combining compute, networking, storage and security—a shift that favors integrated cloud players over bare-metal GPU specialists.

·9 min read
Enterprise AI is pushing cloud infrastructure beyond the GPU
Enterprise AI is pushing cloud infrastructure beyond the GPU

The business case for artificial intelligence infrastructure is evolving rapidly. Organizations that have moved past experimentation and into deploying AI applications at scale are discovering that powerful accelerators alone cannot meet their operational needs. Agentic AI systems demand the ability to execute code, retrieve data, integrate with existing business applications and maintain security across complex, distributed environments—requirements that extend far beyond raw GPU performance.

This shift is reshaping what enterprises expect from cloud providers. Rather than simply provisioning specialized hardware, infrastructure vendors must now deliver integrated platforms that handle compute, networking, storage, security and the orchestration services required to manage applications across their entire lifecycle. The transition is opening doors for neocloud providers willing to expand beyond their original focus on accelerator capacity, while simultaneously raising the bar for what constitutes a competitive offering in the market.

Kevin Cochrane, chief marketing officer at Vultr, discussed these shifting requirements during a live interview at KubeCon + CloudNativeCon North America 2026 with theCUBE Research's Paul Nashawaty. Vultr, which entered the cloud market in 2014 as a developer-focused platform emphasizing performance, cost efficiency and accessibility, has since expanded into AI infrastructure while maintaining its broader cloud services portfolio.

Enterprise AI needs more than GPU infrastructure

Cochrane explained that Vultr adopted a deliberate strategy in the AI infrastructure market, emphasizing enterprise workload requirements and platform-engineering needs rather than pursuing GPU availability as its primary differentiator. "They need a full cloud platform," Cochrane said. "They need cloud compute, specialized AI compute, networking, storage and an open ecosystem."

The early wave of AI infrastructure demand came primarily from hyperscalers, research organizations and model-building startups that needed access to GPU clusters, frequently through bare-metal deployments. The current phase looks fundamentally different. Enterprises are deploying AI agents to optimize operations, enhance customer engagement and unlock new business models. These applications must coexist with legacy systems, satisfy regulatory compliance mandates and fit into established network and application designs.

Bare-metal GPU infrastructure alone cannot satisfy these requirements. Cochrane identified operational complexity and insufficient specialized expertise as significant barriers to enterprise AI adoption. Even when organizations possess access to powerful accelerators, they still require the supporting infrastructure and services to construct, launch and maintain applications in production environments.

Vultr's approach centers on delivering a more comprehensive environment that includes traditional cloud compute, GPU infrastructure, networking, storage and integration with cloud-native projects. The platform supports deployment of more than 60 Cloud Native Computing Foundation projects, allowing customers to access tested, production-ready components directly.

The objective is to minimize the engineering effort required to assemble a functional AI environment. Rather than forcing enterprises to integrate every component independently, providers can deliver preconfigured services and modular infrastructure that development teams can deploy according to their specific needs. This distinction matters increasingly as AI transitions from exploration to production deployment, where infrastructure must support operational applications rather than merely training and inference experiments.

Agentic AI is bringing the CPU back into the conversation

The emphasis on GPUs has sometimes overshadowed the continuing importance of traditional computing resources in AI systems. Agentic AI is expanding the diversity of workloads that infrastructure providers must accommodate.

An AI agent performs far more than submitting a prompt to a model and collecting a response. It may run scripts, interact with code repositories, execute build processes, query databases and invoke third-party tools. These activities demand computing resources that extend beyond GPU capabilities. "Not all workloads need to be on a GPU," Cochrane said.

CPUs remain fundamental for databases, application logic and data processing pipelines. They can also run smaller, optimized open-source models and inference tasks that do not require dedicated accelerators. Another emerging requirement involves the execution environment for agent activities. As agents gain capabilities to run code and interact with systems, enterprises require secure, isolated sandboxes that separate those operations from production systems and sensitive data.

Cochrane pointed to ongoing work in the cloud-native ecosystem focused on safe and secure agent sandboxes as evidence of how infrastructure requirements are shifting. This creates a more layered infrastructure model in which GPUs handle computationally intensive model reasoning and inference, while CPUs support the surrounding applications, data pipelines and execution environments.

For organizations, this means enterprise AI infrastructure decisions must encompass the entire agentic workflow. Purchasing GPU capacity without evaluating the compute, storage, networking and isolation requirements around it risks creating another isolated infrastructure silo. A comprehensive approach enables platform teams to allocate resources to individual workloads more flexibly while maintaining a unified operational environment.

Data sovereignty is becoming an infrastructure architecture decision

As enterprises expand AI deployments across multiple geographic regions, data sovereignty has evolved beyond a simple question of storage location. Organizations operating in regulated sectors or subject to national data protection requirements must understand where information physically resides, which services can access it and what operational consequences arise if infrastructure outside a specific region becomes unavailable.

Cochrane described Vultr's regional architecture as an isolation model in which each of its 34 global data center regions operates independently. Data within a region stays within that region and is not accessed by Vultr services operating outside that geography.

This distinction extends beyond data residency to encompass application operations themselves. Cochrane used France as an example: an application running in a French region should continue functioning even if Vultr's global control infrastructure and other regions become unavailable. "If the rest of the world went dark, your production application in France? It just runs," he said.

This model aims to reduce reliance on global services that could otherwise disrupt regional application availability. It also establishes clearer boundaries for managing sensitive workloads and ensuring operational resilience. For enterprises, this represents a critical architectural consideration. A cloud provider may offer regional data storage, but that does not guarantee that every service required to operate an application is isolated within the same geographic zone.

As AI applications become more tightly woven into business operations, organizations must evaluate sovereignty across the entire application stack, encompassing data, compute, management services and operational dependencies.

AI economics are shifting toward infrastructure ownership

The financial implications of running AI applications are emerging as a major concern as organizations transition from limited pilots to sustained production workloads. Inference expenses can escalate rapidly when agents repeatedly invoke models, gather information and execute complex, multi-step processes. As usage increases, enterprises must weigh model performance, throughput and infrastructure spending against each other.

Cochrane believes growing numbers of organizations will adopt open-source models that can be customized with their own data and operated on infrastructure under their direct control. Rather than depending exclusively on large commercial models for every task, enterprises can fine-tune specialized models using proprietary information and deploy them for operational workloads where this approach delivers value.

He cited an example involving a platform engineering team that achieved an 80% reduction in inference costs by fine-tuning an open-source model and running it on its own infrastructure. While this example demonstrates a potential economic advantage, actual results will vary based on the specific workload, model selection, infrastructure capabilities and operational requirements.

The underlying point is that organizations are beginning to examine AI infrastructure economics across the complete stack. Model selection matters, but so do infrastructure utilization, throughput, data movement and the ongoing expense of operating agents. For certain workloads, maintaining greater control over infrastructure and model operations may provide superior cost management compared to relying entirely on external inference services.

This does not imply that every enterprise should construct and operate its own models. It does suggest that organizations require greater flexibility to determine where models execute and how workloads are distributed across their infrastructure.

Platform engineering will shape the next cloud-native stack

The evolving infrastructure demands of agentic AI are also expanding what platform engineering teams must accomplish. Cochrane anticipates that platform engineers will become essential to enterprise AI adoption, collaborating more closely with business units to identify which agents and preconfigured application environments the organization requires.

This expands platform engineering beyond infrastructure provisioning and developer enablement. Teams will increasingly define the foundational building blocks that allow developers to deploy governed AI applications without assembling every component from scratch.

Cochrane also envisions an opportunity for the cloud-native community to evolve how it distributes infrastructure patterns. The CNCF ecosystem has enabled developers to adopt individual open-source projects across networking, storage, observability and other infrastructure domains. As architectures grow more complex, organizations need guidance on how these components integrate.

Cochrane predicted the emergence of marketplaces where developers can share opinionated, prebuilt infrastructure stacks designed for specific workload categories. Instead of selecting and configuring individual components, teams could begin with a tested architecture and modify it to fit their needs. This could reduce the complexity of developing agentic applications while promoting more standardized deployment practices.

Cochrane also expects consolidation in the neocloud sector. Providers focused primarily on bare-metal GPU capacity may encounter challenges as enterprises demand broader cloud capabilities. Those able to expand into comprehensive public cloud platforms could be better positioned to serve production workloads. Vultr's existing cloud platform provides a foundation for this transition, with investments spanning both traditional compute and specialized AI infrastructure.

The bottom line

Enterprise AI is redefining what cloud infrastructure means. GPUs remain significant, but agentic applications require a broader foundation encompassing CPUs, secure execution environments, networking, storage, data sovereignty and integrated cloud-native services. This creates opportunity for providers delivering more than specialized accelerator capacity. It also positions platform engineering teams in a more strategic role as enterprises work to standardize how AI applications are built, deployed and governed.

The next chapter of the neocloud market may be determined less by who can supply the most GPU capacity and more by who can provide the complete environment enterprises need to operate AI at production scale. The infrastructure challenge has shifted from simply making AI compute available to making the entire system around AI practical to operate at enterprise scale.