Multiverse Computing unveils CompactifAI, a quantum-inspired technology that compresses large language models by up to 95% while preserving performance. Backed by $215 million in new funding, this breakthrough promises faster, cheaper, and more energy-efficient AI, enabling powerful models to run on edge devices.
The relentless march of artificial intelligence, particularly the powerful Large Language Models (LLMs) that have captured the world’s imagination, has always come with a formidable price tag.
These colossal digital brains, capable of everything from crafting poetry to coding software, demand vast, specialized infrastructure, chewing through data center budgets and energy grids at an alarming rate.
The prevailing wisdom has been that to shrink these models was to inevitably diminish their brilliance, a trade-off accepted as the cost of doing business in the AI age.
But a quiet revolution, born in the Basque Country of Spain, is now challenging that very premise, promising to reshape the economic and environmental landscape of AI deployment.
Multiverse Computing, a company that has been diligently toiling at the intersection of quantum science and artificial intelligence, has unveiled CompactifAI.
This isn’t just another incremental improvement; it’s a seismic shift.
The technology boasts an astonishing ability to compress LLMs by up to 95% while miraculously preserving their performance.
Imagine taking a sprawling metropolis of data and condensing it into a bustling village, yet retaining every essential function, every nuance of its original complexity.
That’s the audacious claim, and with a fresh infusion of €189 million ($215 million) in Series B funding, the world is clearly taking notice.
This substantial investment, led by Bullhound Capital and bolstered by strategic giants like HP Tech Ventures, Toshiba, and numerous other international and climate-focused investors, is a powerful endorsement of Multiverse’s vision.
It signals a collective belief that the future of AI isn’t just about bigger models, but smarter, more accessible ones.
The implications are profound, poised to disrupt the $106 billion AI inference market by dismantling the economic barriers that have largely confined cutting-edge LLMs to the cloud-dwelling titans of tech.
Traditionally, attempts to slim down LLMs through techniques like quantization and pruning have resulted in a frustrating compromise: smaller models, yes, but at the expense of accuracy and overall capability.
Multiverse, however, has charted a different course.
Their quantum-inspired approach, rooted in the intricate mathematics of Tensor Networks pioneered by co-founder and Chief Scientific Officer Román Orús, allows for an unprecedented level of optimization.
“For the first time in history, we are able to profile the inner workings of a neural network to eliminate billions of spurious correlations to truly optimize all sorts of AI models,” Orús explained, highlighting the scientific rigor behind the breakthrough.
The results are compelling: CompactifAI models aren’t just smaller, they’re faster – 4 to 12 times faster, to be precise.
And critically, they slash inference costs by a remarkable 50% to 80%.
This isn’t merely a boon for corporate bottom lines; it’s a game-changer for the entire AI ecosystem.
Compressed models become affordable, energy-efficient, and incredibly versatile.
They can still hum along in the cloud or within private data centers, but crucially, the ultra-compressed versions can run directly on edge devices – think PCs, smartphones, cars, drones, and even humble Raspberry Pi computers.
This capability unlocks a universe of new applications, bringing powerful AI capabilities directly to the point of need, enhancing privacy, personalization, and responsiveness in ways previously unimaginable.
Enrique Lizaso Olmos, Founder and CEO of Multiverse Computing, articulated the company’s disruptive philosophy: “The prevailing wisdom is that shrinking LLMs comes at a cost. Multiverse is changing that.”
He noted how quickly their breakthrough proved transformative, earning rapid adoption by drastically reducing hardware requirements.
The broad syndicate of investors, he added, will allow them to “further advance our laser-focused delivery of compressed AI models that offer outstanding performance with minimal infrastructure.”
The enthusiasm from investors is palpable.
Per Roman, Co-founder & Managing Partner at Bullhound Capital, lauded Multiverse’s ingenuity for addressing the global need for greater efficiency in AI, even accelerating European sovereignty in the competitive AI race.
Tuan Tran, President of Technology and Innovation at HP Inc., emphasized how making AI applications more accessible at the edge will bring enhanced performance, personalization, privacy, and cost efficiency to businesses of all sizes.
Damien Henault of Forgepoint Capital International went further, positioning Multiverse as a potential “foundational layer of the AI infrastructure stack,” enabling “smarter, cheaper and greener AI.”
Having spent 2024 developing the technology and rolling it out to initial customers, Multiverse Computing isn’t just promising a future; they’re delivering it.
Compressed versions of popular open-source models like Llama, DeepSeek, and Mistral are already available, with more on the horizon.
With over 160 patents, more than 100 global customers including industrial giants like Iberdrola and Bosch, and accolades like DigitalEurope’s 2024 Future Unicorn award, Multiverse Computing is no longer a nascent startup.
It is a proven innovator, now fully capitalized to lead the charge in democratizing AI, chipping away at its prohibitive costs, and ushering in an era where powerful intelligence is not just for the few, but for everyone, everywhere.
A single malware infection on a senior officer’s computer exposed the sensitive interrogation records of a Syrian National Army unit, revealing systemic security failures.
Ten years after its release, the zombie thriller remains a case study in how local social anxieties propelled Korean film to international prominence.
The 2026 UPSC Mains GS4 paper challenges candidates to reconcile modern technological efficiency with the enduring requirements of administrative integrity and human rights.
