DIMAAG-AI has introduced its ZettaWatt Power Platform that the California startup says is a first-of-its-kind, modular grid-to-800 VDC power conversion and stabilisation solution. The startup is led by Co-Founders Satish Padmanabhan, CEO, and Ian Wright, a former Tesla co-founder, as CTO.
Engineered for hyperscale AI data centres, this system integrates grid conversion, patented immersion-cooled battery technology, and intelligent energy management into a single deployable unit, the company said in a press release, yesterday.
Why is this significant? As AI training loads hit gigawatt scales, they can create large fluctuations that threaten grid stability. Utilities such as Texas’s ERCOT, for example, are beginning to enforce rules requiring Large Computational Loads (LCLs) to ride through specified voltage and frequency disturbances rather than disconnect during grid events.
In parallel, NVIDIA and hyperscalers have announced their transition from 50 VDC to an 800 VDC power architecture, reflecting the industry’s shift toward higher-voltage DC power distribution for next-generation AI servers and megawatt-class racks, DIMAAG notes in its press release.
Its platform holds cyclic load fluctuations to less than 1 percent as seen by the grid, DIMAAG says.
Beyond compliance, the platform bridges the gap between the utility grid and the 800 volt DC power standard for next-gen AI servers. By eliminating traditional double power conversion and bulky AC switchgear, it reduces energy losses, shrinks the physical footprint, and significantly lowers capital costs.
CTO Wright says this redefines hyperscale infrastructure economics, transforming highly dynamic AI data centres into predictable, grid-aware loads.
