Multiverse Computing, active in the compression of quantum-inspired artificial intelligence models, has developed CompactifAI, a compression technology capable of reducing the size of LLMs (Large Language Models) by up to 95 per cent, while preserving their performance. After devoting 2024 to the development of the technology and the roll-out phase with its first customers, the company today announced a EUR 189 million (USD 215 million) investment round.
The Series B round is led by Bullhound Capital with support from HP Tech Ventures, SETT, Forgepoint Capital International, CDP Venture Capital, Santander Climate VC, Quantonation, Toshiba and Capital Riesgo de Euskadi – Grupo SPRI. In this new impetus towards its expansion, the company has obtained extensive support from a number of international strategic investors, including CDP Venture Capital, whose investment is aimed at strengthening the development of Multiverse Computing in Italy. The investment will accelerate the adoption of the technology on a large scale, helping to address the huge costs that hinder the deployment of large language models (LLM), revolutionising the $106.03 billion AI inference market (source: Polaris market research).
LLMs are typically run on specialised, cloud-based infrastructures that drive up data centre costs. Traditional compression techniques such as quantisation and pruning aim to address these challenges, but the resulting models have significantly lower performance than the original LLMs. With the development of CompactifAI, Multiverse has discovered a new approach. CompactifAI models are highly compressed versions of the leading open source LLMs, but retain the original accuracy, are 4-12 times faster and allow a 50%-80% reduction in inference costs. These cheap and energy-efficient compressed models can be run in the cloud, in private data centres or, in the case of ultra-compressed LLMs, directly on devices such as PCs, phones, cars, drones and even Raspberry PIs.
“The prevailing view is that the reduction of LLM has a cost. Multiverse is changing that,’ Enrique Lizaso Olmos, founder and CEO of Multiverse Computing (pictured with the other co-founders), said in a note. ‘What started out as an innovation in model compression has quickly proved transformative, unlocking new efficiencies in AI implementation and gaining rapid adoption with the ability to radically reduce the hardware requirements for running artificial intelligence models. With the entry of a unique pool of experienced and strategic global investors, and with Bullhound’s Capital as the lead investor, we can now further advance our hyper-focused compressed AI model offering, which delivers exceptional performance with minimal infrastructure.”
CompactifAI was created using tensor networks, a quantum-inspired approach to simplify neural networks. Tensor networks are a specialised field of study pioneered by Román Orús, co-founder and chief scientific officer of Multiverse: ‘For the first time in history, we are able to profile the inner workings of a neural network in order to eliminate billions of spurious correlations and truly optimise all kinds of artificial intelligence models. Compressed versions of the main models Llama, DeepSeek and Mistral are available now, with more models on the way.
Per Roman, co-founder and managing partner of Bullhound Capital, adds: ‘Multiverse’s CompactifAI introduces substantial changes in artificial intelligence processing, responding to the global need for greater efficiency in the field of AI. Roman Orus convinced us that he and his team of engineers are developing world-class solutions in this highly complex and computationally intensive field. Enrique Lizaso is the perfect CEO to rapidly expand the company in the global race for AI dominance. I am also delighted to see that so many high-profile investors, such as HP and Forgepoint, have decided to contribute to the round. We welcome their participation.”
Tuan Tran, president of Technology and Innovation at HP says: “At HP, we are committed to driving the future of work by delivering solutions that drive business growth and enhance professional fulfilment. Our investment in Multiverse Computing supports this ambition. By making AI applications more accessible, Multiverse’s innovative approach has the potential to bring the benefits of AI in terms of performance enhancement, personalisation, privacy and cost efficiency to businesses of all sizes’.
Damien Henault, managing director of Forgepoint Capital International, adds: ‘The Multiverse team solved a deeply complex problem with far-reaching implications. The company is well positioned to be a fundamental layer of the AI infrastructure stack. Multiverse represents a quantum leap for the global deployment and application of AI models, enabling smarter, cheaper and greener AI. This is just the beginning of a huge market opportunity’.
ALL RIGHTS RESERVED ©