Fort Collins, Colorado, November 11, 2026 /PRNewswire/ — Ampt today announced a new battery converter designed to enable more compute power per megawatt of grid capacity during the transition of AI data centers to 800 VDC architecture.
The company states that this converter can deliver a regulated 800 VDC from a DC bus shared by battery storage and grid or on-site power sources, providing active power regulation with response times under one millisecond to GPU load changes. By using the battery to handle short-term peak loads, GPUs can operate without power throttling, while the grid connection capacity can be sized for average load rather than peak load.

Each 2U rack-mounted unit delivers 80 kilowatts (100 amps) of power, and multiple units can be paralleled to scale from a single rack to an entire row of cabinets or cover an entire GPU cluster. Ampt says this battery converter is currently undergoing pilot testing in data centers with partners and is expected to be fully available in Q3 2027.
Why is millisecond-level response important?
According to NVIDIA, data centers are transitioning to 800 VDC power architectures for rack-level AI systems, which are expected to support 1-megawatt IT racks and improve end-to-end power efficiency by up to 5% compared to 54-volt systems.
The article states that AI workloads can cause a single rack's power consumption to fluctuate between 30% and 100% within milliseconds. The grid requires stable, predictable loads, while GPUs demand highly dynamic power. According to NVIDIA’s white paper, energy storage is “needed to bridge the gap between these two conflicting demands.”
How does a battery converter work?
According to Ampt’s design, the battery, grid, and any on-site generation equipment are connected to a common DC bus. The battery converter steps down the bus voltage to a regulated 800 VDC to power the GPU racks.
Ampt stated that when GPU load changes, its converter can respond in less than one millisecond, enabling the battery to supply or absorb the difference, resulting in a smoother demand profile for the grid or on-site generation equipment. By smoothing these peaks, operators can size upstream power equipment closer to average demand, reducing equipment costs and enabling more stable load operation, which may also help sites gain faster grid access.
The grid connects to the DC bus via medium-voltage transformers and power conversion systems, supplying batteries rated for operation up to 1500 VDC. The company states that, when equipped with a grid-forming power conversion system, the batteries used to absorb GPU load fluctuations can also actively support grid voltage and frequency.
Ampt stated that these transformers, power conversion systems, and batteries are the same types of equipment used in current utility-scale energy storage, enabling data center operators to leverage established supply chains and Ampt’s nearly two decades of experience deploying DC power conversion in utility-scale solar and energy storage systems.
How does this design improve efficiency?
The article states that in AI factories with limited power, every kilowatt lost during conversion and distribution reduces the capacity available for GPU operation, and operators incur dual costs: one for electricity and another for cooling.
Ampt stated that its design reduces losses in converters, energy storage, and power distribution to preserve more electricity for computing power. The battery converter achieves a peak efficiency of up to 99.2%. The stored energy remains in DC form at all times, avoiding the costly DC-to-AC-to-DC conversion process found in AC-coupled energy storage systems.
Additionally, the DC bus voltage powering the converter varies between 1100 VDC and 1500 VDC depending on the battery’s state of charge, while the converter delivers a stable 800 VDC to the GPU racks. Higher voltage enables the same power to be transmitted at lower current, reducing conductor losses or allowing the use of smaller-gauge conductors.
Ampt said the battery converter is designed for first-generation 800 VDC racks, delivering the benefits of higher distribution voltages—voltages that currently appear only on future roadmaps.
How do I expand the converter?
Ampt stated that its battery converter can be paralleled to scale to any power level, so the same 2U, 80 kW (100 A) module can serve a single GPU rack, an entire row of cabinets, or an entire data center.
Adding one additional device provides N+1 redundancy, maintaining full power even if one device goes offline. The company also states that sites already using or planning to deploy solar power can connect their photovoltaic systems to the same battery DC bus via Ampt String Optimizers, without requiring additional inverters or transformers.
Ampt CEO Aaron Gomolak said: “In an AI factory, GPU loads fluctuate on a millisecond scale, so energy storage must operate continuously. With Ampt’s battery converters, the same battery system can provide stable power to the grid while also delivering the rapid-response power needed by GPUs. For operators, this means more reliable computing power and a faster path to bringing new AI capacity online. For surrounding communities, it means data centers that are more grid-friendly.”
Ampt says that for specifications or to arrange a pilot, visit the company website or contact [email protected].
About Ampt
Ampt states that it provides power conversion and communications technologies to reduce the cost of new photovoltaic systems, enhance the performance of existing systems, enable lower-cost DC-coupled energy storage, and deliver 800 VDC power to AI data centers. The company claims that, based on its global installation and service experience, Ampt is the leading DC optimizer company in the large-systems segment.
The company is headquartered in Fort Collins, Colorado, USA, with sales and support offices in North America, Europe, Japan, and Australia, and representatives in Asia and the Middle East. For more information, visit www.ampt.com or follow Ampt on LinkedIn.
Ampt also stated that NVIDIA is a trademark and/or registered trademark of NVIDIA Corporation in the United States and other countries.



