Hit enter to search or ESC to close
Trending Now
As more businesses move artificial intelligence (AI) from experimentation into everyday operations, deciding where AI workloads should run is becoming an increasingly important part of technology planning.
To help enterprises navigate these decisions, Vertiv Holdings Co. (NYSE: VRT) and the Deloitte Center for Integrated Research have introduced a collaboration focused on enterprise AI infrastructure planning, alongside a new deployment guide developed by Vertiv and digital infrastructure research firm STL Partners.
The Vertiv Designing Infrastructure for Enterprise AI Guide provides business leaders, IT teams and facilities managers with a practical framework for evaluating AI workloads, choosing suitable deployment environments and determining whether their existing physical infrastructure can support them.
To help enterprises navigate these decisions, Vertiv Holdings Co. (NYSE: VRT) and the Deloitte Center for Integrated Research have introduced a collaboration focused on enterprise AI infrastructure planning, alongside a new deployment guide developed by Vertiv and digital infrastructure research firm STL Partners.
The Vertiv Designing Infrastructure for Enterprise AI Guide provides business leaders, IT teams and facilities managers with a practical framework for evaluating AI workloads, choosing suitable deployment environments and determining whether their existing physical infrastructure can support them.
Why Infrastructure Readiness Matters for Enterprise AI
While organisations often focus on AI models, computing capabilities and business applications, the physical infrastructure required to support these technologies can be just as important.
As AI workloads move into sustained production, businesses need to consider whether their facilities have sufficient power, cooling, connectivity and operational capacity to support the systems they intend to deploy.
These requirements can influence where an AI workload should operate, how quickly it can be deployed and whether an existing facility needs to be upgraded.
For example, an organisation may have the computing resources needed for an AI application but still face limitations involving electrical capacity, cooling systems or network connectivity.
The new guide aims to help businesses identify these challenges earlier, before making significant infrastructure investments.
While organisations often focus on AI models, computing capabilities and business applications, the physical infrastructure required to support these technologies can be just as important.
As AI workloads move into sustained production, businesses need to consider whether their facilities have sufficient power, cooling, connectivity and operational capacity to support the systems they intend to deploy.
These requirements can influence where an AI workload should operate, how quickly it can be deployed and whether an existing facility needs to be upgraded.
For example, an organisation may have the computing resources needed for an AI application but still face limitations involving electrical capacity, cooling systems or network connectivity.
The new guide aims to help businesses identify these challenges earlier, before making significant infrastructure investments.
A Four-Step Approach to Planning AI Infrastructure
Rather than recommending a single infrastructure solution for every organisation, the guide introduces a workload-based approach that connects business objectives with technical and operational requirements.
It outlines four steps to help enterprises make more informed AI deployment decisions:
1. Define the business outcome and AI workload
Identify what the organisation wants to achieve through AI and the type of workload required to deliver that outcome.
2. Evaluate workload requirements
Assess the workload's scale, performance expectations and operational needs, including the computing and infrastructure resources required.
3. Determine the deployment location
Consider where the workload should run based on factors such as performance, data proximity, control and available infrastructure.
4. Match the workload to an infrastructure model
Select a suitable physical infrastructure model and reference design that can be further adapted through technical assessment and engineering.
By following this process, organisations can make infrastructure decisions based on the requirements of individual AI applications rather than applying the same approach to every workload.
Rather than recommending a single infrastructure solution for every organisation, the guide introduces a workload-based approach that connects business objectives with technical and operational requirements.
It outlines four steps to help enterprises make more informed AI deployment decisions:
1. Define the business outcome and AI workload
Identify what the organisation wants to achieve through AI and the type of workload required to deliver that outcome.
2. Evaluate workload requirements
Assess the workload's scale, performance expectations and operational needs, including the computing and infrastructure resources required.
3. Determine the deployment location
Consider where the workload should run based on factors such as performance, data proximity, control and available infrastructure.
4. Match the workload to an infrastructure model
Select a suitable physical infrastructure model and reference design that can be further adapted through technical assessment and engineering.
By following this process, organisations can make infrastructure decisions based on the requirements of individual AI applications rather than applying the same approach to every workload.
Cloud, Hybrid or Edge AI: Choosing the Right Deployment Environment
One of the guide's central messages is that enterprise AI should be viewed as a collection of different workloads, each with its own infrastructure requirements.
For instance, businesses experimenting with AI applications may find public cloud environments suitable because of their flexibility.
However, workloads that operate continuously or support business-critical functions may benefit from hybrid infrastructure or computing capacity managed directly by the enterprise.
Meanwhile, AI applications requiring faster response times, greater operational control or closer proximity to data sources may need to operate at the edge or within another controlled environment.
The guide considers several deployment options, including public cloud, colocation facilities, enterprise-owned infrastructure, hybrid environments and edge computing.
Instead of positioning one option as universally better, it provides a framework for evaluating which environment is most appropriate for each workload.
One of the guide's central messages is that enterprise AI should be viewed as a collection of different workloads, each with its own infrastructure requirements.
For instance, businesses experimenting with AI applications may find public cloud environments suitable because of their flexibility.
However, workloads that operate continuously or support business-critical functions may benefit from hybrid infrastructure or computing capacity managed directly by the enterprise.
Meanwhile, AI applications requiring faster response times, greater operational control or closer proximity to data sources may need to operate at the edge or within another controlled environment.
The guide considers several deployment options, including public cloud, colocation facilities, enterprise-owned infrastructure, hybrid environments and edge computing.
Instead of positioning one option as universally better, it provides a framework for evaluating which environment is most appropriate for each workload.
Six AI Workload Archetypes and Four Infrastructure Models
Research supporting the guide identifies six enterprise AI workload archetypes, covering model development, retraining and four categories of production use.
These archetypes group workloads with similar performance, operational and infrastructure requirements, helping businesses develop a more consistent approach to AI planning.
Once a workload has been assessed, it can be matched with one of four physical infrastructure models:
Edge AI Cluster
An infrastructure model intended for AI workloads that need to operate closer to users, devices or data sources.
Small AI Cluster
A model for smaller-scale AI deployments with more limited infrastructure requirements.
Large AI Cluster
An infrastructure model for larger AI environments requiring greater computing and supporting capacity.
High-Density AI Cluster
A model for demanding AI deployments that require high-density computing infrastructure and appropriate supporting systems.
These models provide a starting point for planning power distribution, cooling, connectivity, controls, resilience, operational support and future expansion.
The corresponding reference designs can then be adjusted according to site conditions, workload requirements and detailed engineering assessments.
Research supporting the guide identifies six enterprise AI workload archetypes, covering model development, retraining and four categories of production use.
These archetypes group workloads with similar performance, operational and infrastructure requirements, helping businesses develop a more consistent approach to AI planning.
Once a workload has been assessed, it can be matched with one of four physical infrastructure models:
Edge AI Cluster
An infrastructure model intended for AI workloads that need to operate closer to users, devices or data sources.
Small AI Cluster
A model for smaller-scale AI deployments with more limited infrastructure requirements.
Large AI Cluster
An infrastructure model for larger AI environments requiring greater computing and supporting capacity.
High-Density AI Cluster
A model for demanding AI deployments that require high-density computing infrastructure and appropriate supporting systems.
These models provide a starting point for planning power distribution, cooling, connectivity, controls, resilience, operational support and future expansion.
The corresponding reference designs can then be adjusted according to site conditions, workload requirements and detailed engineering assessments.
Vertiv Highlights the Importance of Early Infrastructure Planning
According to Giordano Albertazzi, CEO of Vertiv, the next phase of enterprise AI will depend not only on the development of AI models and computing technologies but also on organisations' ability to provide the physical infrastructure needed to support them.
He emphasised that starting with the requirements of each workload allows businesses to make infrastructure decisions earlier, identify potential constraints and establish a more practical route from AI experimentation to production.
This approach is particularly relevant as organisations begin integrating AI into essential business processes, where infrastructure limitations can affect operational performance and deployment timelines.
According to Giordano Albertazzi, CEO of Vertiv, the next phase of enterprise AI will depend not only on the development of AI models and computing technologies but also on organisations' ability to provide the physical infrastructure needed to support them.
He emphasised that starting with the requirements of each workload allows businesses to make infrastructure decisions earlier, identify potential constraints and establish a more practical route from AI experimentation to production.
This approach is particularly relevant as organisations begin integrating AI into essential business processes, where infrastructure limitations can affect operational performance and deployment timelines.
Deloitte Calls for Business Value and Infrastructure to Be Planned Together
Nicholas Merizzi, Principal and Silicon2Service U.S. Lead at Deloitte Consulting LLP, also highlighted how the economics and operating requirements of AI evolve as workloads move into production.
He noted that organisations need to consider business value, governance, operational ownership and infrastructure as part of the same planning process.
Addressing these considerations early can give business leaders a stronger foundation for deciding where AI workloads should run and how they should be supported over the long term.
The collaboration reflects a broader need for closer coordination between business decision-makers, IT specialists and facilities teams when planning enterprise AI deployments.
Nicholas Merizzi, Principal and Silicon2Service U.S. Lead at Deloitte Consulting LLP, also highlighted how the economics and operating requirements of AI evolve as workloads move into production.
He noted that organisations need to consider business value, governance, operational ownership and infrastructure as part of the same planning process.
Addressing these considerations early can give business leaders a stronger foundation for deciding where AI workloads should run and how they should be supported over the long term.
The collaboration reflects a broader need for closer coordination between business decision-makers, IT specialists and facilities teams when planning enterprise AI deployments.
Assessing Facility Readiness Before Deploying AI
Another important consideration highlighted by the guide is whether an organisation's intended deployment location is physically ready to accommodate its AI infrastructure.
Before proceeding, businesses are encouraged to assess several essential requirements, including available electrical power, cooling capacity, network connectivity, operational resources and opportunities for future expansion.
These assessments can help determine whether an existing facility is suitable, whether upgrades are necessary or whether a different deployment environment would be more practical.
Identifying potential limitations during the planning stage can reduce the risk of discovering major infrastructure challenges after investments have already been made.
Another important consideration highlighted by the guide is whether an organisation's intended deployment location is physically ready to accommodate its AI infrastructure.
Before proceeding, businesses are encouraged to assess several essential requirements, including available electrical power, cooling capacity, network connectivity, operational resources and opportunities for future expansion.
These assessments can help determine whether an existing facility is suitable, whether upgrades are necessary or whether a different deployment environment would be more practical.
Identifying potential limitations during the planning stage can reduce the risk of discovering major infrastructure challenges after investments have already been made.
Helping Businesses Move AI from Experimentation to Production
As AI becomes more deeply integrated into business operations, infrastructure planning is likely to play an increasingly important role in successful deployment.
The Vertiv Designing Infrastructure for Enterprise AI Guide offers organisations a structured way to connect their AI ambitions with the physical resources required to support them.
By evaluating workloads individually, selecting appropriate deployment environments and assessing facility readiness early, businesses can make more informed decisions about their AI infrastructure investments.
Ultimately, the approach aims to help enterprises move beyond AI experimentation and towards more reliable, scalable and sustainable production environments.
As AI becomes more deeply integrated into business operations, infrastructure planning is likely to play an increasingly important role in successful deployment.
The Vertiv Designing Infrastructure for Enterprise AI Guide offers organisations a structured way to connect their AI ambitions with the physical resources required to support them.
By evaluating workloads individually, selecting appropriate deployment environments and assessing facility readiness early, businesses can make more informed decisions about their AI infrastructure investments.
Ultimately, the approach aims to help enterprises move beyond AI experimentation and towards more reliable, scalable and sustainable production environments.
