Multiverse Computing’s $215M Bet Could Shrink AI Models by 95%
Quantum-Inspired Compression at the Forefront of AI Innovation
Spanish deep tech startup Multiverse Computing has made headlines with the announcement of a €189 million ($215 million) Series B funding round, which will accelerate its development and commercialization of a powerful new AI compression technology: CompactifAI.
Designed to dramatically reduce the size of large language models (LLMs) by up to 95% without compromising accuracy or performance, CompactifAI could be a game changer in the race to make AI more efficient, sustainable, and accessible.
Shrinking Models, Slashing Costs
CompactifAI works by creating compressed versions—dubbed “slim models”—of popular open-source LLMs such as Llama 4 Scout, Llama 3.3 70B, Llama 3.1 8B, and Mistral Small 3.1. Multiverse says these slim models are between 4x and 12x faster, cutting inference costs by 50% to 80%.
For instance, their Llama 4 Scout Slim costs 10 cents per million tokens on AWS, down from the original model’s 14 cents. That makes them not just faster and lighter, but also significantly more affordable at scale.
AI for the Edge—Literally
One of the standout benefits of CompactifAI is its ability to compress models so efficiently that they can run on low-resource devices—including PCs, smartphones, cars, drones, and even a Raspberry Pi.
This opens the door to powerful edge applications, where AI systems can function without relying on constant internet access or cloud infrastructure. Imagine voice assistants, smart homes, autonomous vehicles, or even wearables powered by lightweight LLMs capable of running locally.
Related: Meet Mustafa Suleyman: The Man Behind Ethical AI and Tech’s Most Powerful Moves
Deep Roots in Quantum Physics and Finance
Multiverse was founded by CTO Román Orús, a quantum physicist at the Donostia International Physics Center, and CEO Enrique Lizaso Olmos, a mathematician and former deputy CEO of Unnim Bank. The company uses tensor networks, a computational structure that mimics quantum computing behavior, to achieve these impressive compression results on classical hardware.
Román Orús
Tensor networks, once considered niche tools in physics, are now proving to be powerful in optimizing deep learning models for commercial use.
Related: The Business Secrets Behind Today’s Most Powerful Nonprofits
Backed by Global Tech and Finance Giants
This latest funding round was led by Bullhound Capital, known for backing tech unicorns like Spotify, Revolut, and Discord. Other investors include HP Tech Ventures, SETT, Forgepoint Capital International, Toshiba, and Santander Climate VC.
To date, Multiverse Computing has raised about $250 million and holds 160 patents. It serves over 100 enterprise clients, including names like Iberdrola, Bosch, and the Bank of Canada.
Related: Mark Carney: From Central Banker to Canada's Prime Minister
A Leaner, Smarter Future for AI
As AI systems continue to demand more computational power and energy, solutions like CompactifAI represent a pivotal shift toward sustainable and scalable machine learning. Whether it's reducing cloud bills or enabling real-time inference on edge devices, Multiverse's slim models could be a major milestone in AI’s next chapter.
And with $215 million in fresh capital and a head start in quantum-inspired model compression, Multiverse Computing looks poised to lead the charge.
