QuantumScape, the solid-state battery developer, used the run-up to the Open Compute Project's Global Summit to introduce a product aimed squarely at the AI data-centre power problem. The QS PowerBlock is an in-rack energy-storage reference design built around the company's solid-state lithium-metal cells.
The claims are bold. In a demonstration using an expected data-centre training workload, QuantumScape says the PowerBlock delivered four times the power density and five times the runtime of the industry baseline — the OCP's Open Rack V3 specification. The company also emphasises safety: its cells have shown thermal stability well above 300 °C in third-party and customer testing, compared with conventional lithium-ion cells, which typically enter catastrophic thermal runaway at around 180 °C.
The design targets 800 VDC power architectures and the 1 MW racks that next-generation AI facilities are moving toward. QuantumScape describes it as a modular reference design that can be adapted to various module, shelf and rack formats and deployed in standard power rack configurations. It will be shown at the OCP Global Summit in San Jose from 12 to 15 October, at a booth hosted by Megmeet, a power-component partner in Nvidia's MGX ecosystem.
QuantumScape also said it has joined the OCP, the consortium that defines open hardware standards for data centres, framing the move as part of building the standards infrastructure providers will rely on as AI workloads push densities higher.
The usual caveats apply and the company states them: the performance figures come from prototype testing and modelled system configurations under controlled conditions, benchmarked against ORV3, and final performance will depend on product design, configuration and real-world conditions. Solid-state cells have long promised better energy density and safety than lithium-ion, but have struggled to reach mass production; QuantumScape's move into data centres is a bet that the first large-scale market for its cells may be the rack rather than the road.
