Mindbeam has achieved at least 40% GPU energy efficiency gains, according to an independent audit by Monash University commissioned by deployment partner Celero Infrastructure. The validation tested Mindbeam's Litespark technology across standard high-density GPU configurations in AWS environments, with some multi-node enterprise settings reportedly achieving energy efficiency exceeding 80%. The software optimises GPU efficiency and accelerates large language model pre-training without requiring code changes. By reducing energy consumption whilst maintaining compute output, Litespark enables data centres to scale AI operations within existing power infrastructure constraints. Celero Infrastructure plans to integrate the technology across its Digital Energy Hubs in Asia Pacific and facilitate broader commercial adoption by third-party data centre operators. The breakthrough addresses mounting concerns about AI's strain on global energy grids.