Over the past four years in the Middle East, industry leaders have witnessed a rapid shift in regional focus towards hyperscale and AI data centers. The sense of transformation has been immense, with critical power architectures evolving to cater to a new type of AI-enabled facility that demands speed and energy.

Such changes have also led to organizations like ABB having to adapt quickly to meet market requirements – expanding manufacturing capacity, aligning its sales focus, restructuring operations and technical R&D teams – and executing changes quickly.

Mehdi Rajehi, regional sales, service and project execution manager for the Middle East and Africa at ABB, speaks in a recent DCD>Talks episode about how next-generation power architectures can enable scalable AI data centers.

Having worked in power conversion across EMEA for nearly 15 years, he is no stranger to data centers. Today, he leads power conversion and the Critical Power product group in the Middle East, Africa, and Türkiye, working with end users, EPCs, consultants, and system integrators across critical segments like the data center industry.

“Data center power demand is growing from kilowatts and megawatts to gigawatts,” he explains. “AI is a new type of electrical load, requiring concentrated power, but also dynamic demand, with rapid peaks driven by GPU utilization.”

Overcoming power infrastructure challenges

AI ultimately continues to change power infrastructure requirements for data centers. The challenge for data center power infrastructure is not only to supply more power, but also to accommodate load variation while maintaining four criteria. The architecture needs to meet this criteria and provide bidirectional protection for both GPUs and the grid.

Rajehi unpacks these in the conversation:

“The first is uptime and availability, through high reliability and resiliency: high mean time between failures and low mean time to repair. The second is scalability in megawatts, using a modular architecture that can accommodate future growth – for example, when a campus grows from 200MW to GW,” he explains.

The third criteria is optimized total cost of ownership. As an investor, Rajehi’s goal is to manage both capital and operating costs by maintaining high system efficiency, sustainability, and ease of service. The fourth is AI integration.

On this, he adds: “We need to protect the grid from peaks, minimize the impact of sudden power fluctuations, and reduce the resulting stress. This requires a different architecture, bringing together components such as the UPS, energy storage system, distribution, cooling, transformers, and generators.”

At the same time, there are also growing discussions around medium-voltage UPS data center architectures. These are becoming increasingly important for AI facilities, as they can use energy storage systems to smooth AI peaks, reduce stress on the grid, improve power quality, and support peak shaving.

For Rajehi, it also addresses all four criteria, using fewer components without compromising the uptime and availability of the full system.

“It also has a smaller footprint, which helps maximize the whitespace available for computing,” he says. “Studies show it can improve efficiency by up to three percent.”

Catering to future demand

To help scale future AI data centers, ABB recently launched its Infinitus Direct Current (DC) Portfolio, an industry-first integrated source-to-rack DC power infrastructure solution.

It’s designed to move from medium voltage to DC with minimal conversion stages, coupling the medium-voltage UPS with a DC rectifier or DC solid-state transformer, along with DC distribution. As a result, it can save roughly five percent in losses by reducing power conversion, using less copper, and keeping infrastructure simple.

As Rajehi explains: “What makes this solution exciting is the combination of ABB’s medium-voltage solution with Infinitus DC distribution. Together, they create a highly optimized architecture to enable faster deployment, higher efficiency, and maximum use of power and space for the next generation of AI data centers.”

ABB works across several aspects of the data center system, from grid to chip. Rajehi explains the organization remains invested in advanced DC technologies, energy storage, and energy management systems.

“My final message to EPCs, consultants, and users is: please feel free to get in touch with our experts,” he says. “We’d be happy to support you with your design, explain the technologies, or help with project execution.”

Watch the DCD>Talks episode with Mehdi Rajehi here.