Multiverse Computing has closed a $570 million funding round (around €500 million), led by Forgepoint Capital International, BNPP SIVF, and Bullhound Capital. As a result, the San Sebastián-based company has boosted its valuation to €1.5 billion, becoming a new unicorn and joining a select club of Spanish companies that had previously reached that milestone, such as Factorial, Cabify, and Jobandtalent.
The funds will be used to expand its library of efficient models, strengthen research and development, make strategic investments in sovereign AI gigafactories, and consolidate its presence in markets across Asia, the Middle East, and North America. The company will also continue scaling CompactifAI, its technology for reducing the size of AI models by between 80% and 95%.
Multiverse Computing is a deep tech company founded in 2019 in San Sebastián by Enrique Lizaso (CEO), Román Orús, Sam Mugel, and Alfonso Rubio. From the outset, the startup has specialized in developing quantum, hybrid, and quantum-inspired computing solutions for the enterprise space. It currently also has offices in the United States, Canada, and several European countries, where it works with more than 100 global clients, including Iberdrola, Bosch, and the Bank of Canada.
Since its Series B round, closed in June 2025, the company has multiplied its annualized revenue by more than ten and says that its sales in the first quarter of 2026 were 96 times higher than those recorded a year earlier.
The company was born specializing in quantum software and developed Singularity, a platform for applying quantum and quantum-inspired algorithms to business problems. That experience ultimately gave rise to CompactifAI, which now sits at the heart of its strategy.
CompactifAI is an artificial intelligence model compression technology. Using tensor networks, a mathematical concept inspired by quantum physics, it reduces the number of parameters as well as the memory and computing requirements of language models.
As Enrique Lizaso, co-founder and CEO of Multiverse Computing, explains, “For years, the AI industry has accepted a false constraint as fact: that powerful models require costly infrastructure. That limitation no longer exists. We have shown that AI can run at full performance on a smartphone, inside a sovereign data center, or in an industrial plant without a cloud connection. This round marks the moment to scale that capability across every sector that needs it.”
The company says this approach makes it possible to reduce the size of large language models by between 80% and 95% with minimal loss of accuracy. “The result is faster AI with lower energy consumption, which can be deployed in environments where sending data to a hyperscaler is too expensive, too slow, or simply not allowed,” they explain.
The company says it has achieved size reductions of up to 93%–95% in certain models, although the degree of compression and the loss of accuracy vary in each case. This greater efficiency is what stands out, for example, to Per Roman, co-founder and Managing Partner of Bullhound Capital: “For years, Europe has feared it would not be able to compete at the cutting edge of artificial intelligence. Multiverse is the answer: a team from San Sebastián whose technology allows anyone to run powerful models on their own terms, in their own territory, and without depending on third parties. Backing Román, Enrique, and this team has been one of the most important decisions we have made.”
The latest major launch announced by Multiverse came just a few days ago, with the introduction of optimized and “uncensored” versions of two open-source models: GLM 5.1 and Qwen 3.6 27B (developed by Alibaba). According to the company, those restrictions reduced their usefulness for research, journalism, and other applications that require complete answers on politically sensitive issues
The truth is that the number of launches and new developments unveiled by the startup in recent months is hard to keep up with. For example, among the other solutions developed by Multiverse is CompactifAI Router, which decides in real time whether a workload should run locally or be sent to the cloud. In addition, the Basque startup is also implementing a comprehensive software layer for AI factories and next-generation data centers, “bringing together model compression, GPU orchestration, AI services, and sovereign-level controls on a single platform,” they explain.
Image: ChatGPT
Your email address will not be published. Required fields are marked *
Δ