Multiverse Computing targets €500M for efficient AI

AI models are getting larger, but many customers increasingly need them to become smaller, cheaper and easier to run. Multiverse Computing is positioning model compression as a strategic infrastructure layer for that shift.

What happened

The Spanish company announced a Series C targeting up to €500M at a €1.5B pre-money valuation. The financing is being co-led by Forgepoint Capital International, BNPP Solar Impulse Venture Fund and Bullhound Capital, with commitments from strategic and institutional investors including HP, Orange Ventures, Scania Invest and Santander Alternative Investments.

Multiverse applies tensor-network methods derived from quantum physics to reduce the size and computing requirements of AI models. Its CompactifAI platform is designed to make models easier to deploy on devices, industrial systems and sovereign infrastructure instead of relying exclusively on large cloud environments.

The company is targeting up to the stated amount, so the financing should not yet be treated as a fully closed €500M round.

Why it matters

Model compression is becoming more important as enterprises confront inference costs, power constraints, latency requirements and data-sovereignty rules. A smaller model can be cheaper to operate and easier to deploy close to where data is generated.

The commercial test is whether compressed models can preserve enough performance to justify replacing larger alternatives in demanding production environments.

The bigger picture

The AI market is beginning to value efficiency alongside raw capability. That creates room for infrastructure companies focused on compression, optimisation and deployment rather than training ever-larger foundation models.