Spanish AI Startup Multiverse Computing Secures $215M to Revolutionize Smartphone AI

Multiverse Computing, a San Sebastian-based AI startup, has closed a $215 million Series B funding round led by Bullhound Capital, HP Tech Ventures, and Toshiba. The company''s breakthrough CompactifAI technology leverages quantum physics principles to compress large language models by up to 95% while maintaining 97-98% accuracy.

The compression method reportedly reduces parameter counts by 70% and model memory by 93%, with compressed models like Llama-2 7B achieving 25% faster inference speeds. This innovation addresses the critical challenge of AI model size, potentially enabling complex language models to run on smartphones rather than requiring data center infrastructure.

"We''re profiling neural networks to eliminate billions of spurious correlations," said Chief Scientific Officer Román Orús. The tensor network-based approach combines compression with rapid retraining, claiming 50% performance improvements at inference while preserving nearly all original accuracy.