Gartner IT Symposium: Key Takeaways on the Future of Enterprise AI

Enterprise AI is moving beyond experimentation, but turning early progress into production-scale business impact requires more than access to GPUs.

That was a central theme at Gartner IT Symposium, where Vultr joined conversations with industry analysts and enterprise technology leaders about the evolving AI infrastructure market, the growing role of CPUs, and what organizations need to deploy AI with trust, control, and measurable results.

Across these discussions, one message stood out: Enterprises need infrastructure built around the applications they want to run, not just the accelerators they want to buy.

Here are three key takeaways from this year’s Symposium.

1. Enterprise AI needs more than GPU capacity

The rise of neocloud providers and the growing investment in AI infrastructure were major topics at Symposium. As organizations commit more resources to AI, questions about infrastructure economics, long-term value, and the ability to translate investment into business outcomes are becoming increasingly important.

Gartner’s October 2026 announcement that more than $3.5 trillion in AI infrastructure spending could be impaired by the end of 2030 added urgency to these conversations. The announcement underscored the need for organizations to evaluate AI infrastructure decisions carefully as the market evolves.

For Vultr, the opportunity extends beyond providing GPU as a service (GPUaaS). Enterprises need a complete cloud infrastructure foundation that supports the full lifecycle of AI applications, including compute, accelerated infrastructure, storage, networking, and security.

During Symposium, Vultr emphasized the importance of bringing together the strengths of traditional cloud computing and accelerated computing to help organizations move from individual AI experiments to real applications and production workloads.

That broader approach also means giving enterprises flexibility to choose infrastructure based on their technical and business requirements, rather than assuming every AI workload needs the same architecture.

2. CPUs remain critical as agentic AI evolves

While GPUs continue to attract significant attention, the role of CPUs was another important theme throughout Symposium.

In his discussions about the changing demands of AI infrastructure, Vultr CMO Kevin Cochrane reinforced a simple point: The CPU never went away.

As enterprises deploy AI agents, the work extends beyond model training and inference. Agents need to coordinate tasks, call tools, retrieve information, interact with existing systems, and execute multi-step workflows. These activities place additional demands on CPU resources alongside GPUs.

As Cochrane discussed in a recent interview with Fierce Network, Vultr is seeing demand for CPU resources increase as agentic AI workloads become more prevalent. The infrastructure requirements for enterprise applications differ from those of environments focused primarily on supplying large quantities of GPU capacity.

For enterprises, the objective is to run complete applications that solve real problems. That requires balancing accelerated computing with the CPU performance and supporting infrastructure needed to coordinate and execute workloads.

Vultr’s AMD EPYC-powered VX1 Cloud Compute is part of this broader approach, providing dedicated CPU resources for the application logic and orchestration that support modern AI workloads.

The takeaway from Symposium was clear: As AI moves into production, organizations must account for the entire workload rather than focusing exclusively on GPUs.

3. Start with the workload and measure the impact

Vultr’s Piyush Shukla joined a customer panel moderated by HPE, alongside Sudheer Matta, SVP of HPE Networking, and Laura Lincoln of Grant Thornton. The discussion explored how organizations can scale AI with trust, control, and measurable impact.

One of the central takeaways was to start with the workload and work backward.

Before choosing a model or investing in infrastructure, organizations need to understand the business problem they are trying to solve. That means identifying the desired outcome, determining the application’s requirements, and selecting the resources and architecture best suited to deliver the result.

This approach helps enterprises make more informed infrastructure decisions. Some workloads may require substantial GPU capacity, while others depend more heavily on CPU performance, data access, storage, or networking. Many production applications require a combination of these capabilities.

The panel also highlighted the importance of metrics. Deploying an AI application is only the beginning. Organizations need to establish how they will measure success, demonstrate business impact, and determine whether an initiative is delivering the expected results.

Whether the objective is to improve operational efficiency, accelerate a business process, or enable a new capability, measurable outcomes provide a foundation for making AI investments sustainable over time.

Trust and control are equally important. As AI agents gain the ability to interact with enterprise systems and take action, security must be part of the infrastructure conversation from the beginning. Organizations need the appropriate controls to protect sensitive workloads while enabling teams to move forward with AI adoption.

From Symposium conversations to enterprise AI infrastructure

These discussions reflect the direction of Vultr’s ongoing investments in enterprise AI.

At Symposium, Vultr’s broader enterprise AI developments and its Chennai expansion provided additional context on how the company is building infrastructure to support organizations as they move AI into production.

The Chennai cloud data center region expands Vultr’s global footprint and access to infrastructure for AI and enterprise workloads. Meanwhile, Vultr’s enterprise AI platform developments focus on giving organizations the flexibility to combine compute, accelerated infrastructure, and supporting technologies according to their needs.

Together, these efforts reflect a broader view of what enterprises require from their infrastructure providers: global access, workload flexibility, performance, and the security and control necessary to operate AI applications in real-world environments.

As enterprise AI adoption advances, the buying conversation is expanding as well. Infrastructure decisions increasingly involve security leaders alongside IT, engineering, and business stakeholders. That makes a comprehensive approach to architecture and governance essential for organizations preparing to scale AI.

The next phase of enterprise AI

Kevin Cochrane used an analogy from his experience running ultramarathons to describe the current stage of AI adoption: Enterprises have reached the end of the first marathon, but the journey is far from over.

Organizations are making meaningful progress, moving beyond experimentation and beginning to put AI to work. The next challenge is building the infrastructure and operational foundations needed to scale those initiatives and deliver lasting business value.

The conversations at Gartner IT Symposium reinforced several priorities for that next phase: Look beyond GPU capacity, account for the growing role of CPUs, start infrastructure decisions with the workload, and measure success by business outcomes.

For Vultr, these priorities inform an approach to cloud infrastructure that combines accelerated computing with the broader capabilities enterprises need to build, deploy, and scale AI applications.

The first phase of enterprise AI has demonstrated what is possible. The next will be defined by how effectively organizations implement it.