Gerchamp Launches Modular AI Data Center Architecture
Factory-assembled 500 kW to 3.5 MW containers integrate direct-to-chip liquid cooling and Nickel-Zinc battery backup for off-grid AI compute.
Modular AI infrastructure developer Gerchamp launched a factory-assembled data center portfolio designed to deploy high-density compute directly to sites with available power. The system scales from 500 kW single-container units to 3.5 MW multi-container deployments, enabling operators to bypass multi-year utility transmission queue delays. By integrating power distribution, direct-to-chip liquid cooling, and Nickel-Zinc battery backup into modular blocks, the platform allows high-density AI clusters to go live in weeks rather than years.
Key details
The newly launched portfolio addresses the critical speed-to-power bottleneck facing high-density artificial intelligence deployments:
- Scalable modular architecture: The entry-level 500 kW all-in-one single container supports IT space and power distribution for five Open Compute Project (OCP) racks or four NVIDIA NVL72 racks.
- Multi-megawatt configurations: For larger compute needs, Gerchamp links multiple factory-tested containers into a coordinated 3.5 MW system comprising two IT containers, two cooling containers, and one power container.
- Extreme rack density: Multi-container configurations support up to 24 OCP racks or 20 NVIDIA NVL72 racks, supporting rack power densities of up to 175 kW per rack.
- Vertically integrated subsystems: Each container combines proprietary direct-to-chip liquid cooling, uninterruptible power supplies (UPS), Nickel-Zinc battery energy storage, Building Management Systems (BMS), and Data Center Infrastructure Management (DCIM).
- Rapid site deployment: Factory pre-assembly and pre-testing allow a 500 kW module to be brought online in weeks, while multi-megawatt systems can be delivered and powered in months, bypassing traditional 5-to-10-year grid interconnection delays.
Why this matters
The global AI buildout has run into a severe transmission queue bottleneck, with utility interconnections frequently requiring five to ten years to construct. High-density AI accelerators, such as NVIDIA NVL72 racks, require extreme power and liquid cooling that traditional air-cooled data centers cannot support without extensive retrofitting.
By delivering pre-integrated power, direct liquid cooling, and battery energy storage directly to sites where power capacity already exists—such as industrial plants, telecommunications facilities, or behind-the-meter generation sites—operators can deploy high-density AI workloads without waiting for centralized grid expansions.
Context
Gerchamp's modular launch reflects a broader industry transition toward decentralized, behind-the-meter, and liquid-cooled data center architectures. As electrical utilities struggle to keep pace with gigawatt-scale AI load requests, infrastructure providers like Vertiv, Flex, and ONE Nuclear are increasingly offering modular, microgrid, and on-site power solutions.
Furthermore, as direct-to-chip liquid cooling becomes mandatory for GPUs exceeding 1 kW per chip, factory-integrated cooling infrastructure reduces deployment risks and water consumption compared to traditional evaporative cooling towers.
Risks and open questions
Deploying containerized AI infrastructure at non-traditional sites introduces several operational considerations:
- Site-level power constraints: While modular units avoid new utility line builds, facilities must still secure adequate local transformer capacity and behind-the-meter power contracts.
- Environmental durability: Edge and industrial sites expose liquid-cooling loops and battery systems to ambient temperature swings and airborne contaminants, requiring rigorous long-term maintenance.
- Regulatory and zoning approvals: Municipalities increasingly scrutinize noise levels and battery storage safety for modular deployments, even on existing industrial land.
What happens next
Gerchamp will begin delivering factory-tested 500 kW and 3.5 MW modular units to enterprise, telecommunications, and industrial customers seeking rapid AI inference and training capacity.
As hyperscalers and neocloud providers face expanding interconnection queues across major data center hubs, industry analysts expect increased adoption of containerized liquid-cooled modules to bring high-density AI compute online wherever local power can be tapped.
Source: Data Center Dynamics Published on AI Usage Global, author: AUG Bot
