The power landscape is growing increasingly vast and unwieldy. Operators are struggling to integrate new and diverse power sources to support ever-denser chips and server racks. As more of this equipment occupies valuable data center space that cannot easily be scaled, the need for a solution that condenses and simplifies power infrastructure becomes clear.

The challenge is to simplify the integration of diverse power sources and loads, removing power as a constraint on the growth of the data center sector.

In this DCD>Talks episode, Haroon Inam, CEO at DG Matrix, describes the company’s mission to transform power infrastructure for AI data centers and for the era of mass electrification.

Fleet to grid

DG Matrix develops multi-port solid-state transformers. Founded primarily as a power electronics business to solve behind-the-meter power aggregation challenges, its initial focus was on deep, high-quality feature integration.

At the time, however, the data center world was not quite ready for the level of density the company was delivering, prompting a pivot toward fleet electrification. With several members of the DG Matrix team bringing aerospace experience, its designs have supported applications ranging from aircraft and air-compression systems to jet engine starters.

Then came the AI data center boom. Suddenly, a highly aggregated solution for combining multiple AC and DC power sources was already in production – a cross-sector technology backed by major industry heavyweights. As Inam explains:

“As part of this journey, we aim to become a major reference architecture for data centers. Now, our job is to bring the global deployment of our solution.”

Interport over conventional

Conventional approaches to data center power management typically rely on UPS architectures, standalone battery systems, or additional behind-the-meter power aggregation capability.

The concept of stranded power in traditional data centers arises because the conservative nature of breakers and UPS systems requires a high peak-handling capacity. On the data center side, however, repeated surges in demand lead to unused capacity and unnecessary power loss. Stranded power at this scale ultimately contributes to lower revenue.

DG Matrix addresses this challenge through its interport device, as Inam explains:

“It’s not that we’re trying to disrupt anybody’s business, but we have a highly integrated solution that ticks multiple boxes at the same time in a footprint that’s half to two-thirds the size of legacy systems, with significantly lower Capex and far better operating efficiency. It is the solution for the new world of AI data center power.”

The multi-port solid-state transformer (SST) interport device can aggregate AC or DC sources within milliseconds through multiple ports, as illustrated in DG Matrix’s whitepaper. Portions of these sources can be dynamically added to grid power in either transient or steady-state form, enabling unprecedented control and delivery of power to AC loads, 800 VDC, or both.

In this way, the full power load is delivered cleanly while peak demands are managed across all sides of the system, helping recover stranded capacity and generate additional revenue.

“The time has come for SST. Designed specifically for AI data centers, the economics of a multi-port SST are compelling – not to mention the enhanced level of protection and resiliency it provides in the face of failure,” says Inam.

Pilot A

The first pilot, installed in April 2024, operated in an outdoor environment for nearly nine months before being decommissioned so the team could capture and incorporate key learnings. A second set of pilots, deployed in July 2025, has been running across multiple locations, from California to North Carolina.

These deployments are designed to demonstrate source aggregation, UPS capability, battery blending and charging, and the ability to deliver surge power to GenAI servers.

The application of these pilots is valuable for both new build and retrofit scenarios. For new builds, pre-engineered designs with shorter lead times are a clear advantage for operators competing in the race for speed to power. For retrofit projects, the condensation of data center infrastructure offers a significant benefit for sites at risk of outgrowing their existing footprint.

“If you are a modular data center, such as those required for inference engines deployed across corporate campuses or outdoor environments, then we really are the best solution because of the small footprint and the aggregation of all the sources in one place.”

Reclaiming space

Designing new facilities – particularly for the GenAI era – is a substantial undertaking. As modern compute requires ever greater levels of cooling and networking equipment to support denser servers, space within the data center comes at a premium.

When fewer power stages are needed to deliver energy from the source to the load, both losses and physical footprint are reduced.

“In some configurations, we are only ten to 15 percent of a legacy solution, and we demonstrate that in some of our deployments. In others, we could be as much as two-thirds of the footprint – but even that represents a substantial saving in the gray space and outside the data center.”

The complexity involved in aggregating multiple power electronic stages, rectifiers, transformers, and protection systems should not be underestimated.

To learn more about reducing the data center footprint to this significant degree, watch the full DCD>Talks episode here.