Colocation Emerges as Key Infrastructure for AI Workloads

Colocation Emerges as Key Infrastructure for AI Workloads

Enterprises are fundamentally rethinking their infrastructure strategies as artificial intelligence workloads move from experimentation to production, with colocation facilities emerging as a critical component of hybrid deployments.

More than half of organizations have now implemented or are actively upgrading AI technologies, representing growth from the previous year, according to research from CoreSite. The shift reflects AI's maturation beyond pilot projects, with generative AI, chatbots, predictive analytics, and agentic AI now running in production environments at many organizations.

The infrastructure implications are substantial. AI systems demand levels of compute capacity, power, cooling, and low-latency connectivity that many organizations struggle to provision effectively in traditional environments.

Why it matters

This infrastructure shift has direct cost and performance implications for technology leaders. As AI token usage grows and cloud invoices climb, CIOs are discovering they lack accurate visibility into how widely these tools are deployed across their organizations. The move toward colocation represents a strategic response to managing both predictable performance requirements and escalating operational costs.

Hybrid becomes the default

Hybrid environments have become the preferred deployment model for AI and machine learning workloads, while interest in purely on-premises deployments continues declining. Organizations are using public cloud for rapid experimentation and initial deployment, then shifting production workloads requiring predictable performance or dedicated infrastructure to colocation facilities.

"The levels of compute that AI requires are something new that enterprises are grappling with to manage effectively," said Juan Font, President and CEO of CoreSite and SVP of American Tower. "While the AI tools are effective, CIOs may not currently have accurate reporting on how widespread the usage is within their organizations."

Font noted that when IT leaders see actual invoices for large language model usage and token consumption, they begin rationalizing and prioritizing projects based on return on investment.

Connectivity drives selection

Direct connectivity between enterprise infrastructure and major cloud providers has become a critical requirement. Seventy-nine percent of IT leaders identified native, direct cloud connections as very important capabilities when selecting colocation providers.

These connections deliver lower-latency access to cloud services, reduce dependence on the public internet, and simplify data movement between enterprise and cloud environments. Organizations are evaluating provider ecosystems that combine cloud platforms, network carriers, AI services, security products, and managed services to support workloads across multiple infrastructure types.

Beyond AI-specific workloads, enterprises are expanding colocation use for web applications, human resources systems, security workloads, and augmented AI applications. When selecting providers, organizations prioritize security, uptime, and predictable performance alongside scalable infrastructure that supports the higher-density power and cooling requirements of AI systems.

Some workloads previously deployed in public cloud are moving to colocation as organizations reassess placement based on performance, security, and infrastructure requirements.

These findings were first reported by Help Net Security based on CoreSite research.