Gen AI Demands on Data Center Capacity: Navigating a New Frontier

The Impact of Changing Requirements

Increased rack density is the primary metric used to gauge the impact of gen AI. According to JLL’s Data Centers 2024 Global Outlook report, average rack density in hyperscale data centers is expected to reach 48.7kW in 2027, up from 36.1kW in 2024. That’s a 7.8 percent compound annual growth rate.¹

But raw power demands don’t tell the whole story. The three stages of gen AI — model creation, training, and inference — have very different power demands. Thus, the power requirements of gen AI workloads tend to fluctuate more than those of traditional workloads. Additionally, inference requires not only high speed but also low latency transmission performance, creating the need for a densely clustered infrastructure. Data centers capable of supporting AI don’t just require more power, they look different.

As a result, operational practices are changing. Data center operators must increase compute capacity within the same physical space while improving efficiency and decreasing costs. Forward-thinking operators are using intelligent tools for monitoring, management, automation, and orchestration to increase energy efficiency and reduce staffing requirements.

¹ Data Centers 2024 Global Outlook, 2024

The Sustainability Conundrum

Sustainability remains a major driver of data center design and operations. Developers and operators remain under intense pressure to curb energy usage and reduce greenhouse gas emissions. However, the transition to clean energy is hindered by constraints on the electric grid. Some data centers are unable to rely on electric utilities for a consistent source of power. They must depend on backup power to offset load shedding and guard against total grid collapse. Traditional backup power sources, such as diesel generators, increase costs and emissions.

The Inflation Reduction Act offers incentives for data centers to invest in solar farms and provides tax credits for batteries that can store renewable energy. Some data centers are experimenting with alternative power sources such as hydrogen fuel cells, gas turbines, geothermal energy, and small modular reactors (SMRs). Others are partnering with independent power producers and utilities to bring transmission lines to the data center.

How Data Centers Should Respond

Data centers must move quickly if they hope to capitalize on the gen AI boom. However, supply chain challenges and power constraints are adding years to the time required to build a new facility. Many existing Tier III and Tier II data centers are unable to support next-generation AI technologies, and modernization takes time due to the same issues.

To overcome these hurdles, data center developers and operators need to invent new ways of doing things. They need trusted partners who are creative and can source all the solutions needed to build out an AI-ready infrastructure. This requires a collaborative, partnership-focused approach. Traditional relationships with siloed suppliers will not deliver the speed and ingenuity needed to take the data center into the future.

Wesco is uniquely positioned to help data centers respond to this transformative opportunity. We have subject-matter expertise in a wide range of disciplines, including energy, cooling, infrastructure and more. We also have finely tuned supply chain and distribution processes that ensure the right products arrive at the right time and at the right price. Let us help you navigate the new frontier of gen AI and gain strategic advantages in today’s data center marketplace.