Businesses moving artificial intelligence (AI) from testing into daily operations need to check whether their facilities have enough power, cooling, and connectivity to support it, according to research behind a new Vertiv enterprise AI infrastructure guide.
Vertiv Holdings Co., a critical digital infrastructure company, and Deloitte Center for Integrated Research introduced a collaboration addressing where businesses should run AI and whether their infrastructure can handle it. The Vertiv Designing Infrastructure for Enterprise AI Guide was developed by Vertiv and STL Partners, a digital infrastructure research firm.
“The next phase of enterprise AI will be shaped not only by models and computing, but also by whether organizations can provide the power, cooling, connectivity, and operating environment each workload requires,” said Giordano Albertazzi, CEO of Vertiv. “By starting with the workload, enterprises can make infrastructure decisions earlier, identify constraints sooner, and build a more credible path from AI ambition to production.”
The guide gives business, IT, and facilities teams four steps: identify the business goal and AI task, assess its computing and operating needs, choose where it should run, and match it with an infrastructure plan.
These decisions depend on what the AI system needs to do. Early experiments may stay in the cloud. Systems used regularly or for essential business tasks may need company-managed infrastructure or a mix of cloud and company resources.
AI that needs faster responses or must process data close to its source may need nearby computing equipment, known as edge infrastructure.
“The economics and operating requirements of AI change as workloads move into production,” said Nicholas Merizzi, principal, Silicon2Service US Lead, Deloitte Consulting LLP. “Organizations need to consider business value, governance, operating ownership, and infrastructure together. Bringing those questions into the planning process earlier gives leaders a stronger basis for deciding where AI workloads should run and how they should be supported over time.”
The research groups AI tasks into six types, covering model development, retraining, and four categories of production use. These help teams compare tasks with similar needs.
Each task can be matched with one of four physical infrastructure models: Edge AI Cluster, Small AI Cluster, Large AI Cluster, and High-Density AI Cluster.
The models provide starting points for planning power, cooling, connectivity, operational support, and expansion. Designs must then be adapted to the site and technical requirements.
Checking facilities early can reveal limits before they delay deployment, including insufficient electricity, cooling capacity, or space for growth.