Addressing the pain points of large language models (LLMs) being bulky and costly to deploy, Spanish AI startup Multiverse Computing is breaking the deadlock with its unique "compression technology," aiming to compete with giants like OpenAI in the enterprise AI market.

Core Technology: A 60B model occupying only 32GB

Multiverse's core competitiveness comes from its compression technology CompactifAI, inspired by quantum computing. With this technology, the company successfully reduced the size of its HyperNova60B model, based on the OpenAI prototype (gpt-oss-120b), by half.

The latest released version HyperNova60B2602 is now available for free on Hugging Face. The model size is only 32GB, significantly reducing memory usage and latency while maintaining accuracy almost equal to the original version. In addition, the new version has been specifically optimized for scenarios with high inference costs, such as "tool calls" and "agent encoding." Multiverse claims that the model has surpassed Mistral AI's flagship model Mistral Large3, a French AI unicorn.

Chip AI drawing (1)

The Rise of European AI: Geopolitical and Business Landscape

Similar to Mistral, Multiverse is becoming a representative of Europe's independent AI technology. Currently, the company's business spans Europe and the Americas, and it has won the support of heavyweights such as Bosch, Iberdrola, and the Bank of Canada.

This "sovereign AI" positioning has earned government support. SETT, the Spanish Agency for Technological Transition, participated in its recent 215 million dollar Series B financing. With its ability to provide autonomous solutions across the entire artificial intelligence technology stack, Multiverse is rapidly filling the demand for alternatives to American technology in the market.

Valuation Surge: On the Way to a 1.5 Billion Euro Unicorn

Although Multiverse remains cautious about market rumors, according to insiders, the company is negotiating a new round of 500 million euros funding, with an estimated valuation exceeding 1.5 billion euros.

Although its annual recurring revenue (ARR) is reportedly around 100 million euros, it still has a huge gap compared to OpenAI's 20 billion dollars, but it is approaching Mistral's 400 million level. With more compressed models planned to be open-sourced in 2026, this startup from the Basque region is highly likely to become Spain's first AI unicorn.