GreenLLMs

With their unique Quantum AI software approach (Tensor Networks) Multiverse Computing has developed a tool that optimises, makes more efficient and sustainable any AI model and/or LLM: CompactifAI. The Company is able to compress an LLM to a maximum of 99% of its original size, without almost losing any accuracy, which in return, reduces the electricity need by half. Such a small compressed model could then be deployed on edge or any device with no cloud connection, such as drones, phones or satellites. Many companies struggle to deploy Generative AI (GenAI) due to high costs, infrastructure limitations, and regulatory concerns. Training and running GenAI models require massive computing power, with costs reaching tens to hundreds of millions of euros, making large-scale adoption unsustainable. Additionally, most state-of-the-art GenAI models are too large to be efficiently deployed on-premise or at the edge, forcing enterprises to rely on expensive cloud infrastructure controlled by non-EU providers. This creates scalability and latency challenges while increasing dependency on foreign tech giants. Beyond cost and infrastructure, regulatory barriers prevent many companies from adopting GenAI. Businesses in finance, healthcare, and defence face strict data privacy regulations such as GDPR and the EU AI Act, making it impossible to use third-party cloud-hosted AI models. Companies need sovereign AI solutions that allow for on-premise deployment without compromising security. CompactifAI solves these challenges by compressing AI models by up to 93%, cutting operational costs by 50%, and enabling faster, cheaper, and more efficient deployment. By removing these barriers, CompactifAI bridges the gap between AI builders and enterprises, making GenAI truly scalable, affordable, sustainable and compliant within the EU.