Spain's Multiverse raises $217 million for compressing AI models

PARIS (Reuters) -Spanish AI firm Multiverse Computing said on Thursday it has raised 189 million euros ($217 million) from investment firm Bullhound Capital, HP Inc, Forgepoint Capital and Toshiba, to compress AI language models.

The company said it has developed a compression technology capable of reducing the size of large language models (LLMs) by up to 95% without hurting performance and reducing costs by up to 80%.

It combines ideas from quantum physics and machine learning in ways that mimic quantum systems but doesn't need a quantum computer.

The latest funding round makes Multiverse the largest Spanish AI startup, joining the list of top European AI startups such as Mistral, Aleph Alpha, Synthesia, Poolside and Owkin.

Multiverse has launched compressed versions of LLMs such as Meta's Llama, China's DeepSeek and France's Mistral, with additional models coming soon, the company said.

"We are focused just on compressing the most used open-source LLMs, the ones that the companies are already using," Chief Executive Officer Enrique Lizaso Olmos said.

"When you go to a corporation, most of them are using the Llama family of models."

The tool is also available on Amazon Web Services AI marketplace.

($1 = 0.8709 euros)

(Reporting by Supantha Mukherjee in Paris; Editing by Lisa Shumaker)