Multiverse Computing and CompactifAI: How quantum-inspired compression shrinks AI models by 95%

What problem does it solve? Large language models are extremely expensive to train and run, requiring massive clusters of high-end GPUs. Their size and energy consumption create a cost and infrastructure barrier for most enterprises — inference costs are high, and deployment is limited to large data centers . CompactifAI overcomes that barrier.

Reported results:

• 80%–95% reduction in hardware footprint (model size and memory)
• Only 2–3% precision loss on most benchmarks
• 4x–12x faster processing and 50%–80% lower inference costs
• Meta's LlaMA-2 7B was compressed to just 3% of its original size in one test
• Llama 3.3 70B was compressed to run on Intel Xeon 6 CPUs without dedicated GPUs
• A compressed OpenAI model (gpt-oss-120B) had its memory needs cut from 61GB to 32GB — a 50% reduction — while retaining near-parity performance

Series C Funding Round

In July 2026, Multiverse Computing announced a Series C round targeting up to $570 million (€500 million) at a $1.7 billion pre-money valuation (approximately €1.5 billion pre-money, with a post-money valuation of about €2 billion) . The round remains open for additional commitments . Total funding to date is approximately $800 million .

Co-Lead Investors

• Forgepoint Capital International (US-based venture capital)
• BNPP Solar Impulse Venture Fund (BNP Paribas' climate-focused venture fund)
• Bullhound Capital (UK-based investment bank and VC)

Other participating investors include Santander Alternative Investments, Tikehau Capital, HP Inc., Orange Ventures, Scania Invest, NAventures, Qatar Development Bank, Zouk Capital, SETT, the EIC Fund, and the Basque Government's Hazten Scale-Up Fund .

Where the Technology is Deployed

CompactifAI is deployed across a broad range of hardware, from edge devices (drones, cameras, satellites, vehicles, smartphones, Raspberry Pi) to cloud servers and sovereign AI infrastructure . The company has an API portal and a showcase app for developers, and compressed models from OpenAI, Meta, DeepSeek, and Mistral AI are available .

Customers and Partners

Named customers include Iberdrola (energy), Bosch (industrial/manufacturing), and more than 100 enterprises globally . Other named enterprise customers include Allianz, Bank of Canada, Indra, and Telefónica .

Key partnerships:

• PwC (November 2025) — simplifying enterprise AI adoption and lowering compute costs
• Axelera AI (March 2026) — integrating CompactifAI-compressed models on Dutch AI hardware
• Qualcomm (August 2026) — deploying compressed models on Qualcomm's data center accelerators
• Intel (July 2026) — CompactifAI-compressed Llama 3.3 70B runs on Intel Xeon 6 processors
• Tech Mahindra (2023) — delivering quantum computing solutions to global enterprise customers
• HP Inc. — participated as an investor in the Series C

Key Takeaway

Multiverse Computing represents a significant shift in the AI industry: rather than making models bigger, it makes them smaller, faster, and cheaper to run. Its quantum-inspired approach to model compression has attracted major strategic investors and partners, and its technology is already in use across multiple industries. For enterprises looking to deploy AI at scale without the massive GPU infrastructure that large models typically require, CompactifAI offers a concrete path forward.