Up to $570m · Series C · Deeptech · Donostia, Spain
Multiverse Computing, a Donostia-based developer of efficient AI software, has announced an open Series C fundraising targeting up to $570m (€500m). Forgepoint Capital International, BNPP Solar Impulse Venture Fund and Bullhound Capital are co-leading, with commitments to date from strategic, financial and public investors including HP, Orange Ventures, Santander Alternative Investments, Tikehau Capital, Scania Invest, NAventures, Qatar Development Bank, SETT and the EIC Fund. The company plans to expand its model library, research, sovereign-AI infrastructure and presence in Asia, the Middle East and North America.
The announced amount is a target, not a confirmed closed total. Multiverse says the round may remain open to additional strategic investors and would take cumulative funding to $800m if fully subscribed.
Compression shifts the AI cost equation
Multiverse’s CompactifAI technology applies tensor-network methods from quantum physics to shrink AI models. The company says it can reduce model size by 80–95% with immaterial accuracy loss, then deploy the compressed models in cloud, on-premises and device environments. Its router assigns workloads locally or to cloud backends, while its broader platform adds GPU orchestration and controls for sovereign deployments. That gives enterprises a route to use existing hardware and keep sensitive workloads closer to their own infrastructure instead of sending every request to a hyperscaler.
The commercial mechanism is capacity as much as software efficiency. A smaller model can lower memory, compute and energy requirements; for a device maker, enterprise or data-centre operator, that can mean serving more inference from the same hardware or moving workloads onto constrained edge systems. The claim still depends on performance holding up in each customer’s environment, but it gives Multiverse a clear value metric: usable AI capacity per unit of infrastructure, rather than model access alone.
Large targets raise the execution bar
Forgepoint says Multiverse’s annualised revenue grew more than tenfold after its June 2025 Series B and Q1 2026 sales rose 96 times year over year. The company says its models run across millions of devices and systems and that it serves more than 100 customers, including Bosch, Iberdrola and the Bank of Canada. Those are company and investor claims, but they matter because a round of this scale requires the compression layer to work across different hardware, security requirements and procurement cycles. The next test is whether expanding the model library and sovereign-infrastructure stack turns those deployments into repeatable enterprise growth.
Explore this theme
AI hardware: Where the next opportunities are emerging — Synaptic, adapted by EUVC · 16 October 2026. Its analysis of inference, power and edge constraints explains the infrastructure bottlenecks that model compression is trying to loosen. Added 26 October 2026.
Sources checked:
Multiverse Computing company announcement · Forgepoint Capital investor announcement · Tech.eu Funding Explorer
Image: Multiverse Computing, via the official Series C announcement.
