Multiverse Computing raises $215M for tech that could radically lower AI costs
Spanish startup Multiversum Computing on Thursday said It has raised a huge series of B round of € 189 million (about $ 215 million) based on a technology that calls it ‘compactifai’.
Compactifai is a compression technology inspired by quantum computing that is able to reduce the size of LLMs by a maximum of 95% without influencing model performance, the company said.
In particular, Multiversum offers compressed versions of well-known, open source LLMs, Probably small models-such as Llama 4 Scout, Llama 3.3 70b, Llama 3.1 8B and Mistral Small 3.1. The company will soon be planning to release a version of Deepseek R1 and said it is working on more open source and reasoning models. Upbone models from OpenAi and others are not supported.
The “Slim” models, as the company mentions them, are available via Amazon Web Services or can have a permit for use on the on -premise. The company says that its models are 4x to 12x faster than comparable non-compressed versions, which translates into a reduction of inference costs from 50% to 80%. For example Multiversum out The Llama 4 Scout Slim costs 10 cents per million tokens on AWS compared to 14 cents from Lama 4 Scout.
Multiversum claims that some of his models can be made so small and energy efficient that they can be performed on PCs, telephones, cars, drones and even the favorite small PC of the DIY-Enthusiast, Raspberry Pi. (We suddenly imagine that those fantastic Raspberry Pi Christmas light houses have been upgraded with LLM-driven interactive talking Santas.)
Multiversum has a technical power behind it. It was co -founded by his CTO Román Orús, a professor at the Donostia International Physics Center in San Sebastián, Spain. Orús is known for his pioneering work on tensor networks (Not to be confused with all AI-related projects with the name Tensor at Google).
Tensor networks are computational tools that simulate quantum computers but are performed on normal computers. One of their primary applications nowadays is compression of deep learning models.
The other co-founder of Multiverse, CEO Enrique Lizaso Olmos, also has several mathematical degrees and has been a university professor. He spent most of his career in the banking system and is best known as the former deputy CEO of Unnim Banc.
The series B was led by Bullhound Capital, who supported companies such as Spotify, Revolut, Delivery Hero, Avito and Discord. HP Tech Ventures, Set, Forgepoint Capital International, CDP Venture Capital, Santander Climate VC, Toshiba and Capital Riesgo De Euskadi – Grupo Spr Spr took also participated in De Ronde.
Multiversum says it has 160 patents and 100 customers worldwide, including Iberdrola, Bosch and the Bank of Canada. With this financing it has collected around $ 250 million so far.
