Summary
- Rittal and Bosch have outlined a coordinated rack, liquid cooling and 800V DC architecture.
- The new design targets high rack densities within Open Compute Project infrastructure.
- An announced €100m European and US manufacturing investment does not establish deployment at a named customer site.
Rittal has introduced an integrated data centre infrastructure architecture combining high-density equipment racks, direct liquid cooling and 800V DC distribution, with power electronics supplied by Bosch. The German company's announcement accompanies the Open Compute Project Global Summit on 12 October. Its proposed system addresses the physical limits that emerge when computing hardware is concentrated into higher-powered racks.
The move towards higher voltage is driven by the electrical difficulty of distributing rising amounts of power at lower voltages. For a given delivered power, increasing voltage reduces current, which in turn can lower the quantity of conductor material and the resistive losses associated with high-current distribution. It also changes the protective switching, isolation and personnel-safety requirements. Any claimed system efficiency therefore has to be assessed against the complete conversion and protection chain, rather than treating a move to 800V as sufficient evidence of lower operating cost.
Direct liquid cooling serves a related but distinct function. High-density processors place more heat into concentrated packages than traditional rack airflow can practically remove. Liquid distribution brings heat transfer closer to the equipment but adds piping, manifolds, couplings, leak controls and thermal monitoring. Designing the rack and facility distribution as a coordinated system can help the manufacturer specify mechanical interfaces and service access in advance, rather than forcing builders to reconcile incompatible components during installation.
Rittal says its architecture builds on open OCP structures, including the Open Rack V3 ecosystem. Open interface specifications have the potential to limit unnecessary customisation, but interoperability depends on adopted versions, equipment tolerances and validation of complete assemblies. A compatible electrical format does not guarantee that two vendors' thermal components or control systems will integrate without further engineering. Claims of fast deployment should consequently be supported by project-level commissioning evidence.
The company is also investing €100 million in manufacturing capacity in Europe and the United States. That investment may strengthen supply for racks and associated physical infrastructure, although the announcement does not allocate the sum by plant or establish a new line's operating date. Manufacturing capability is one link in a longer supply chain: data centres still require switchgear, transformers, cooling plant, controls, construction contractors and sufficient utility power at the chosen location.
No named European customer deployment or independently measured efficiency comparison was disclosed alongside the system announcement. The products should be assessed as an integrated proposed solution, with performance dependent on configuration. Rack density, coolant temperature, power conversion efficiency, outage recovery requirements and maintenance access will vary between installations. A facility retrofitting an established power backbone would also face a different cost equation from a new building designed around 800V distribution.
A combined rack and power system also has to account for the difference between short transients and sustained demand. AI clusters can change load rapidly, and the rack may require local energy buffering or coordinated controls to prevent disturbances from propagating. High-voltage distribution can make the physical bus more manageable, but engineers still need to specify what happens when a power shelf or coolant branch develops a fault. Protection selectivity and maintainability matter as much as the maximum advertised rack capacity.
Liquid cooling choices must be matched to the facility. Warm-water operation can reduce chiller requirements in some climates, but achievable temperatures depend on the equipment and environment. Cooling distribution units may isolate the IT coolant from the building's water circuit, introducing pumps, filters and heat exchangers that need their own power and maintenance. No general energy saving can be established from the presence of liquid connections alone. Integrated design may reduce compatibility risks, but measured whole-site energy performance remains the appropriate comparison.
For European equipment buyers, the manufacturer's production expansion is a separate potential benefit. Standard modules supplied through a stable manufacturing programme can simplify spare-parts planning and repeat purchases. That advantage depends on final interface documentation, consistent component availability and support through several equipment generations. Operators will want evidence of how power electronics and liquid hardware can be replaced without lengthy rack outages, particularly where a single rack represents a large share of installed computing capacity.
Further detail on certification, volume availability and site-scale trials will determine how readily the architecture can be adopted. Rittal's announcement is a concrete engineering development, particularly in the integration of power and liquid cooling. Evidence of its operating performance at a commercial facility is still required before making stronger claims about savings or deployment speed.
