As enterprise AI moves beyond the experimental stage into real work environments, competition in the cloud infrastructure market is changing. The contest, once focused on securing GPUs, is expanding to integrated platform capabilities spanning computing, networking, storage and security.
On Nov. 19, IT outlet SiliconANGLE reported that Vultr Chief Marketing Officer Kevin Cochrane (케빈 코크런) stressed at the recent KubeCon + CloudNativeCon North America 2026 that enterprise AI operations require comprehensive cloud infrastructure that goes beyond GPUs.
At the center of the shift is agentic AI. AI agents go beyond answering questions to perform complex tasks such as executing code, accessing databases and calling external applications. That means companies must consider not only model compute performance but also data processing, system integration and security.
The importance of CPUs, long overshadowed by GPUs, is also resurfacing. While GPUs handle high-performance AI inference, CPUs support databases and application execution as well as data pipelines. Some small open-source AI models can run on CPUs alone without dedicated accelerators.
Security and data sovereignty have also emerged as key competitive factors. Because AI agents access internal corporate systems, they require isolated environments that can protect sensitive information.
Vultr operates 34 data center regions worldwide independently. It said the design allows applications in other regions to keep running normally even if a system failure occurs in a specific region. Vultr is also expanding its global AI infrastructure, including unveiling an AMD GPU-based supercluster in Chennai, India, on Nov. 9.
Cutting costs is also an important task. As AI agents carry out multistep work, model-calling and inference costs rise. Cochrane introduced a case in which a platform engineering organization reduced inference costs by 80 percent after further training an open-source model and running it on its own infrastructure. He added that this was an individual case and the same effect is not guaranteed for all companies.
Cochrane forecast that companies offering integrated cloud services will gain competitiveness over neocloud providers that supply only GPU resources. He also expected the role of platform engineering organizations in charge of building and operating corporate infrastructure to expand as AI adoption accelerates.