Transforming AI Model Efficiency Through Quantum-Inspired compression
Multiverse Computing Secures Notable Funding to Enhance LLM compression
Spanish tech innovator Multiverse Computing has successfully completed a ample Series B funding round, amassing €189 million (around $215 million) to accelerate the advancement of its pioneering technology, CompactifAI. This cutting-edge solution aims to revolutionize large language models (LLMs) by dramatically minimizing their size while preserving their high-level performance.
The Breakthrough of compactifai: Maximizing Compression Without Losing Precision
CompactifAI employs concepts inspired by quantum computing principles to compress LLMs by as much as 95%, all while maintaining their original accuracy and functionality. This innovation facilitates more resource-efficient AI deployments, substantially reducing both computational load and operational expenses.
Wide-Ranging Model Compatibility and future Launches
The startup currently provides compressed versions of several popular open-source LLMs, focusing primarily on smaller-scale architectures such as Llama 4 Scout, Llama 3.3 70B, Llama 3.1 8B, and Mistral small 3.1. Upcoming releases will include a compressed edition of DeepSeek R1 along with other open-source reasoning models. At present, proprietary platforms like OpenAI’s offerings are not supported.
Tangible Performance Improvements and Cost Savings in Practical Use Cases
The streamlined models created through CompactifAI are accessible via Amazon Web Services or can be licensed for deployment on-premises.These optimized versions deliver inference speeds that are four to twelve times faster than the original models, leading to cost reductions between 50% and 80%. For instance, running the slimmed-down Llama 4 Scout on AWS costs roughly $0.10 per million tokens compared with $0.14 for its standard counterpart.
Broadening AI Usability Across Devices and Sectors
A standout feature of these compressed models is their ability to operate efficiently on low-power devices such as laptops, smartphones, autonomous drones, electric vehicles, or compact systems like Raspberry Pi units-enabling innovative edge computing applications. Picture smart agricultural sensors powered by conversational AI functioning locally without relying on cloud connectivity.
The Scientific Roots Behind Multiverse’s Technological edge
This breakthrough is grounded in co-founder Román Orús’ expertise in tensor networks-a sophisticated mathematical framework originally developed within quantum physics that simulates quantum computations using classical hardware architectures. Tensor networks have proven essential for compressing deep learning structures while retaining critical facts.
“Tensor networks allow us to emulate certain facets of quantum computation efficiently on conventional machines,” highlights Orús’ foundational research that drives CompactifAI’s methodology.
A Leadership Team Merging Academic excellence with Industry Insight
Alongside Orús is CEO Enrique Lizaso Olmos who brings a unique blend of strong academic background in mathematics combined with extensive leadership experience from the banking sector-including his tenure as deputy CEO at Unnim Bank-providing strategic guidance vital for scaling advanced technologies into commercial successes.
Diverse Investor Support Demonstrates Confidence in Long-Term Potential
- Bullhound Capital leads this investment round; known for backing companies like Spotify and Revolut;
- HP Tech Ventures contributes strategic support;
- SETT participates actively;
- Forgepoint Capital International adds financial strength;
- Certain venture arms including CDP Venture Capital and Santander Climate VC provide specialized funding;
- Toshiba joins alongside regional investor Capital Riesgo de Euskadi – grupo SPR.
An Expanding Global Presence Backed by Prestigious Clients
This recent capital injection brings Multiverse Computing’s total funds raised close to $250 million while safeguarding over 160 patents worldwide.
Its clientele spans major players such as Iberdrola (energy sector), Bosch (engineering), plus financial institutions including the Bank of Canada-reflecting widespread trust across industries requiring scalable AI solutions optimized for efficiency and performance at scale.
