VRFB and Solid-State Transformer Combo Set to Run Live AI Compute in US

TerraFlow Energy and DG Matrix announced a commercial agreement on September 28, 2026, to deploy what the companies describe as the first integrated vanadium redox flow battery (VRFB) and solid-state transformer (SST) architecture in the United States to directly power a live high-performance computing cluster.

The release, datelined Houston, states that the initial deployment will integrate TerraFlow’s LDUPS long-duration uninterruptible power system with DG Matrix’s Interport solid-state transformer technology and Dell PowerEdge servers, putting a new data center power architecture into operation against real compute loads.

The companies said the significance goes beyond combining two technologies. AI data centers are introducing enormous new loads to the power system, and according to the release the challenge is not only how much electricity those facilities consume. High-performance compute can change power demand rapidly, pushing that volatility back through electrical infrastructure built for more predictable loads.

How the Integrated Architecture Manages Power

The companies describe an integrated architecture designed to manage power between the grid, on-site generation, energy storage, and compute in real time. TerraFlow says its LDUPS combines the millisecond response required for uninterruptible-power-supply functionality with hours of energy capacity, allowing it to continuously buffer changes in compute demand. DG Matrix’s Interport provides programmable solid-state power conversion, routing, and control between power sources and the compute load. Together, the technologies are designed to absorb rapid changes in demand before they reach the grid and other upstream electrical infrastructure.

The release also frames the architecture against an issue it says confronts utilities and data center developers: how to connect gigawatts of new compute demand without requiring the power system to respond instantaneously to every change occurring inside a data center. By controlling those changes behind the meter, the companies state, integrated storage and power distribution can help make large loads more predictable, more flexible, and easier for the grid to serve.

Executive Statements and the Commercial Framework

Jon Parrella, chief executive officer of TerraFlow Energy, said AI is changing the physics of the power system and that simply adding more generation does not solve that problem. He said the companies have to change how these enormous loads connect to and interact with the grid, and that placing long-duration storage and intelligent power distribution between the grid and the compute can give data centers the power performance they need without asking the grid to absorb every movement in demand, an approach he described as starting to turn data centers from grid liabilities into grid assets.

“We aren’t building another layer of backup power,” Parrella said. “We’re building the infrastructure that sits between compute and the grid and allows both of them to perform better.”

Haroon Inam, chief executive officer of DG Matrix, said AI infrastructure requires rethinking power distribution from the grid all the way to the compute. Solid-state transformers, he said, allow power to be controlled dynamically rather than simply moved from one voltage to another, and combining that capability with TerraFlow’s LDUPS creates an architecture that can respond to the compute load in real time while controlling how that load is presented to the grid.

Beyond the initial deployment, the commercial agreement establishes a framework for the two companies to bring the integrated LDUPS and Interport architecture into additional data center projects. The companies said they are targeting applications where AI and high-performance computing require greater control over how power is delivered, managed, and exchanged with the grid, and they describe the initial deployment as the starting point for a commercial architecture designed to scale with the power requirements of AI infrastructure.

The Companies Behind the Agreement

TerraFlow Energy describes itself as built on long-duration, non-flammable vanadium flow battery technology, with a battery-in-building architecture that combines uninterrupted power, long-duration energy storage, and controllable load management for AI data centers and other energy-intensive facilities. On its official site, the company states that its flow-based, in-series, fire-safe systems scale from 10 MWh to 5,000 MWh and are built to smooth volatile loads, filter harmonic distortion, and operate in series with generation assets. TerraFlow’s newsroom records an August 3, 2026, strategic memorandum of understanding with Shoals Technologies Group, an agreement the company’s headline describes as a multi-gigawatt partnership to advance energy storage.

DG Matrix, which calls itself the global leader in solid-state transformer solutions, says its AI-enabled Interport platform is built on what it describes as the world’s first commercially available, programmable multi-port solid-state transformer. According to the company, the platform converts, routes, and controls power between the grid, on-site generation, storage, and compute loads through a single software-defined stage, with native 800 VDC readiness and protection from GPU pulse loads.

DG Matrix’s press page records two earlier steps in that power-architecture buildout during 2026. On June 17, 2026, the company announced it had joined the NVIDIA MGX ecosystem as a solid-state transformer supplier for NVIDIA’s 800 VDC power architecture, citing more than two years of direct technical collaboration with NVIDIA. On July 24, 2026, the company announced a strategic technical and commercial collaboration with Skeleton Technologies that integrates Skeleton’s fast-response energy storage systems across the Interport platform to support the data center industry’s transition to 800 VDC power infrastructure.