Multiverse Computing and YellowDog.ai have formed a strategic agreement to address the growing need for secure and efficient artificial intelligence infrastructure within Europe. The collaboration will combine Multiverse Computing’s CompactifAI technology, using quantum-inspired tensor-network methods to compress large AI models, with YellowDog.ai’s dynamic compute management platform.

This pairing aims to accelerate AI development for enterprises, research organizations and public-sector institutions in sectors including financial services, life sciences and defence. “Sovereign AI will only succeed if it is economically and technically sustainable,” said Enrique Lizaso, co-founder and CEO of Multiverse Computing, adding that their compressed AI models are designed to run faster and use less energy.

CompactifAI Compression with Dynamic Compute Orchestration

CompactifAI compression uses quantum-inspired tensor-network methods to address the growing demand for efficient AI model deployment, particularly in light of current GPU shortages. This technology directly tackles the challenge of large model sizes by reducing the computational resources required for both training and inference, a critical step toward wider accessibility. Multiverse Computing’s approach focuses on minimizing the energy consumption and operational costs associated with running complex AI applications, a key consideration for European AI Gigafactories, the company says.

These facilities, described by the European Commission as large-scale infrastructure projects, aim to expand access to advanced computing capacity for developing and deploying AI models. YellowDog.ai’s dynamic compute management platform complements CompactifAI by intelligently orchestrating workloads across hybrid and multi-cloud environments. This orchestration is designed to aggregate available GPU capacity, a vital function given the ongoing supply constraints, and to ensure secure access to processing power.

The combined system allows customers to benchmark, optimize and deploy compressed AI models across diverse infrastructure, including cloud, on-premises, edge and sovereign data centers. This flexibility is particularly relevant for sectors like financial services, life sciences and defense, where data residency and security are paramount. “Europe has the talent, data and industrial base to lead in AI, but it needs infrastructure that is fast, efficient, secure and sovereign by design,” said Bruce Beckloff, CEO of YellowDog.ai.

The partnership specifically targets the requirements of European sovereign AI, emphasizing compliance with EU rules and values regarding data protection, safety and technological autonomy. The resulting infrastructure is intended to support AI experimentation, reduce operational costs through model compression and workload scheduling, and facilitate sovereign deployment patterns.

These AI Gigafactories are expected to combine AI-optimized compute with high-capacity storage, advanced networking and specialized AI support services. “By partnering with Multiverse Computing, we are creating practical capacity solutions where organisations can scale compute intelligently, reduce bottlenecks and retain control of critical workloads,” Beckloff added.

The focus on reducing energy use through model compression aligns with the broader goals of sustainable AI development and the creation of energy-efficient data centers. The collaboration aims to provide a powerful route to deploy efficient AI across a range of environments, addressing the need for scalable, secure and cost-effective AI infrastructure in Europe.