Multiverse Computing has opened a funding round led by Forgepoint Capital International, BNP Paribas SIVF, and Bullhound Capital, with proceeds aimed at expanding its model library, boosting R&D, and investing in sovereign AI gigafactories to expand in Asia, the Middle East, and North America.

The company is targeting a Series C of up to $570 million (€500 million), potentially bringing total funding to about $800 million if fully raised.

Investors cited include HP Inc., Orange Ventures, Scania Group Invest, NAventures, Qatar Development Bank, Tikehau Capital, and European funds tied to the Basque Government and the EIC Fund, signaling broad cross-industry interest.

CEO says AI can run at full performance on smartphones and sovereign data centers without cloud connections, underscoring a vision to reduce cloud dependence and energy use.

The platform features the CompactifAI Router, which dynamically routes workloads to local devices or the cloud, along with an end-to-end software layer unifying model compression, GPU orchestration, AI services, and sovereign controls.

Multiverse is building a platform to optimize model selection and deployment, integrating model compression, GPU management, AI services, and control systems.

CompactifAI enables open-source models like Llama and Mistral to run 4–12 times faster at 50–80% lower cost and can operate on phones, drones, or factory floors without cloud connection; the router decides on-device versus cloud execution.

Investors are eyeing inference efficiency as a dominant cost for buyers, making smaller, more efficient models financially attractive despite not growing model size.

Independent evaluation by Sopra Steria of CompatifAI on Llama 3.1 8B reportedly shows significant resource reductions while maintaining accuracy.

The company has released optimized and uncensored versions of open-source models GLM 5.1 and Qwen 3.6 27B (Alibaba) to broaden research and sensitive-context applicability.

Forgepoint Capital International’s Damien Henault describes Multiverse as evolving from LLM compression to a full AI foundry and operating system, highlighting strong technical foundations and commercial traction.

The broader European push in quantum and AI hardware/software emphasizes reducing energy and cost footprint of AI inference rather than expanding model size.