Multiverse Computing

Multiverse Computing is a privately held software company headquartered in San Sebastián, Spain. Founded in 2019, it began with a focus on applying quantum computing and quantum-inspired techniques to practical problems in financial services and other industrial sectors. The company was co-founded by Enrique Lizaso, Román Orús, Alfonso Rubio-Manzanares and Sam Mugel. Its founding concept emerged from discussions between Lizaso and Orús about creating a company capable of translating quantum-computing research into commercial applications, initially with finance as a principal target market. The company develops software rather than quantum-computing hardware. Its technology portfolio includes quantum algorithms, quantum machine-learning methods, optimization tools and tensor-network techniques. These methods can be used with quantum processors, classical computers or hybrid computing environments. Multiverse has applied or explored its technology in areas including foreign-exchange trading, financial modeling, energy, logistics, manufacturing, chemistry, aerospace, cybersecurity and complex-network analysis. Its business expanded beyond quantum software with the development of CompactifAI, a model-compression platform introduced in 2023. CompactifAI uses tensor-network methods to reduce the parameter count, storage requirements and memory footprint of large AI models while seeking to preserve their capabilities. The resulting models are intended to be faster and less expensive to operate than their uncompressed counterparts, and can be deployed in public clouds, private data centers or edge devices. Potential edge applications include vehicles, industrial equipment, aircraft, satellites, phones, personal computers and other hardware with limited computational resources. Multiverse has used CompactifAI to produce compressed versions of models associated with providers and open-source communities including Meta, Microsoft, Qwen, DeepSeek, Mistral and OpenAI. The company has also developed its own model families, including compressed versions of Llama models and the nano-model concepts SuperFly and ChickenBrain. It has made selected models available through channels such as Hugging Face and has offered access to compressed-model inference through an API distributed via the AWS Marketplace. The company's commercial and research relationships have included the Bank of Canada, Bosch, BASF, BBVA, Crédit Agricole CIB and Repsol. In 2022, it worked with the Bank of Canada on the use of quantum computing to model cryptocurrency adoption and partnered with Bosch on quantum algorithms for digital-twin simulations and defect detection. BASF used Multiverse's Singularity software in work related to foreign-exchange trading optimization. In 2024, Multiverse also secured a contract from the German Aerospace Center concerning single-photon detectors for potential quantum-computing, deep-space communications and bio-imaging applications. Funding and institutional support have been important to the company's growth. It announced €12.5 million from the European Innovation Council Accelerator in 2021 and a further €25 million funding round in 2024, reportedly giving it a valuation of €100 million at that time. In June 2025, it announced a $215 million Series B round led by Bullhound Capital, HP Tech Ventures, SETT, Forgepoint Capital International, CDP Venture Capital, Santander Climate VC, Toshiba and Capital Riesgo de Euskadi, with Grupo SPRI also participating. The company has described the financing as supporting the scaling of compressed AI models and their deployment across cloud, on-premises and edge environments. Multiverse operates from Spain and maintains offices in several European and North American locations, including Madrid, Barcelona, Zaragoza, Paris, Munich, London, Milan, Toronto and San Francisco. Its positioning combines quantum-computing expertise with pragmatic software engineering: rather than presenting quantum hardware as a prerequisite, it emphasizes tools that can deliver value in hybrid, classical, quantum-inspired and quantum-computing settings. The company's current identity therefore spans two related areas: enterprise quantum software and efficient, highly compressed artificial-intelligence models.

Multiverse Computing was established in 2019 in Spain by Enrique Lizaso, Román Orús, Alfonso Rubio-Manzanares and Sam Mugel. The founding idea was to commercialize quantum-computing techniques for practical business problems, particularly in finance. Lizaso brought an entrepreneurial and industry-oriented perspective, while Orús contributed expertise in tensor networks and quantum physics. The company subsequently developed a broader platform intended to make quantum and quantum-inspired methods usable by organizations without requiring them to build quantum hardware themselves. During its early development, Multiverse focused on optimization, quantum machine learning and financial applications. Its software was designed to work across emerging quantum processors, classical systems and hybrid architectures. The company explored applications in trading, risk and financial modeling, while also extending its work into energy, manufacturing, logistics, chemistry, aerospace and cybersecurity. Singularity became associated with this phase of the business and was used in projects involving financial optimization and industrial simulation. In 2021, Multiverse announced €12.5 million from the European Innovation Council Accelerator. The funding supported the company's effort to turn quantum-computing research into deployable software and commercial projects. In 2022, the company developed relationships with several large organizations. Its collaboration with the Bank of Canada examined whether quantum computing could be used to simulate cryptocurrency adoption and complex networks. A partnership with Bosch investigated quantum algorithms in digital-twin workflows, including applications to simulation scale and defect detection. BASF later used Singularity in work on foreign-exchange trading optimization between the United States and European markets. BBVA, Crédit Agricole CIB and Repsol were also identified as organizations exploring the company's software. Multiverse's strategy widened in 2023 with the launch of CompactifAI. The platform applies tensor-network methods to AI models, reducing the number of parameters and lowering memory and storage demands. This approach addressed a major constraint in generative AI: large models can be expensive to train and serve, and their size limits their use on phones, vehicles, industrial equipment and other edge devices. CompactifAI enabled Multiverse to offer compressed versions of established models while positioning the company around energy efficiency, portability and lower inference costs rather than only quantum hardware access. In 2024, Multiverse announced a €25 million funding round at a reported valuation of €100 million and was selected for the European Innovation Council's Scaling Club. It was also chosen for AWS's 2024 Generative AI Accelerator, giving it access to mentoring and AWS infrastructure. The company expanded its research activity through a $1.4 million German Aerospace Center contract concerning single-photon detectors, with possible uses in quantum computing, deep-space communications and bio-imaging. CompactifAI models were also made available through an API on the AWS Marketplace, allowing organizations to access compressed-model inference through cloud infrastructure. The company's 2025 activity emphasized model compression as a core commercial growth engine. It launched compressed versions of Llama 3.1 8B and Llama 3.3 70B and described them as offering substantial reductions in parameters, energy use, latency and inference cost while retaining the source models' capabilities. These models were designed for cloud, on-premises and edge deployment. In June 2025, Multiverse announced a $215 million Series B investment led by a consortium including Bullhound Capital, HP Tech Ventures, SETT, Forgepoint Capital International, CDP Venture Capital, Santander Climate VC, Toshiba and Capital Riesgo de Euskadi, with Grupo SPRI also participating. Later in 2025, Multiverse presented SuperFly and ChickenBrain, two very small compressed models intended to demonstrate how advanced AI functions could run on ordinary or highly constrained devices. The company also received several forms of industry recognition, including a DIGITALEUROPE Future Unicorn award and appearances in technology and AI rankings. Its subsequent model work included HyperNova 60B and related releases described as open or openly distributed compressed models, as well as tools such as CompactifAI App that combine local, offline inference with a chat interface. Today, Multiverse Computing remains a private Spanish technology company with offices across Europe and North America. Its business spans quantum software, quantum-inspired algorithms, tensor-network research and AI model compression. The company presents these capabilities as complementary: quantum and quantum-inspired methods provide techniques for difficult optimization and simulation problems, while tensor-network compression makes large AI systems smaller, cheaper and easier to deploy. This hybrid positioning distinguishes Multiverse from both quantum-hardware manufacturers and conventional AI model providers.

• 2025

  Series B financing

  [1]Multiverse Computingen.wikipedia.org

  The company announced a $215 million Series B investment focused on scaling compressed AI models.

• 2025

  Compressed Llama models released

  [1]Multiverse Computingen.wikipedia.org

  Multiverse released compressed versions of Llama 3.1 8B and Llama 3.3 70B for cloud, on-premises and edge use.

• 2024

  Expansion of funding and institutional programs

  [1]Multiverse Computingen.wikipedia.org

  The company announced a €25 million round, joined the EIC Scaling Club and participated in AWS's Generative AI Accelerator.

• 2024

  German Aerospace Center contract

  [1]Multiverse Computingen.wikipedia.org

  Multiverse secured a $1.4 million contract concerning single-photon detector development.

• 2023

  CompactifAI launched

  [1]Multiverse Computingen.wikipedia.org

  Multiverse introduced its tensor-network-based AI model-compression platform.

• 2022

  Financial and industrial partnerships

  [1]Multiverse Computingen.wikipedia.org

  Multiverse worked with the Bank of Canada, Bosch and BASF on cryptocurrency-adoption modeling, digital twins and foreign-exchange optimization.

• 2021

  European Innovation Council support

  [1]Multiverse Computingen.wikipedia.org

  The company announced €12.5 million in funding from the European Innovation Council Accelerator program.

• 2019

  Company founded

  [1]Multiverse Computingen.wikipedia.org

  Enrique Lizaso, Román Orús, Alfonso Rubio-Manzanares and Sam Mugel co-founded Multiverse Computing in Spain.

An enterprise technology company combining quantum software, quantum-inspired tensor networks and AI model compression to make advanced computation more efficient, portable and commercially deployable.

• Quantum software and optimization

  Singularity

  Singularity is Multiverse Computing's quantum and quantum-inspired software platform for optimization, machine learning and industrial problem solving. It has been explored in financial applications such as foreign-exchange trading and in industrial workflows such as digital-twin simulation. The platform is intended to support hybrid use of quantum processors and classical computing resources rather than requiring customers to operate dedicated quantum hardware.

• AI model compression · 2023

  CompactifAI

  CompactifAI uses tensor-network techniques to compress large AI models, reducing parameter counts, memory requirements, storage needs and inference costs. The platform is designed to produce smaller versions of established large language models while preserving useful accuracy and capabilities. Its intended deployment environments include public cloud infrastructure, private data centers and edge devices such as phones, vehicles, industrial systems, drones and other hardware with limited resources.

• AI inference API

  CompactifAI API

  CompactifAI API provides access to compressed large language models through a serverless inference layer available on the AWS Marketplace. It is designed to use AWS infrastructure, including SageMaker HyperPod, to serve optimized models across GPU clusters. The offering gives developers a way to integrate compressed models into applications without managing the full underlying model-serving stack.

• Compressed large language models · 2025

  Compressed Llama models

  Multiverse has released compressed versions of Meta's Llama 3.1 8B and Llama 3.3 70B. The company describes these models as having materially fewer parameters and lower energy, latency and inference-cost requirements than their source models. They are intended for use in cloud, on-premises and edge settings, including devices such as laptops, phones, cars and industrial equipment.

• Nano AI model · 2025

  SuperFly

  SuperFly is a highly compressed model derived from SmolLM2 135M. Multiverse describes it as a roughly 94-million-parameter model small enough to operate locally on constrained devices. Its purpose is to demonstrate that conversational AI can run without a network connection on hardware such as phones and small computers, reducing dependence on cloud inference.

• Nano AI model · 2025

  ChickenBrain

  ChickenBrain is a compressed model derived from Meta's Llama 3.1. It is presented as a small reasoning-capable model suitable for constrained hardware, with demonstrations on a MacBook Pro and Raspberry Pi. The product illustrates Multiverse's effort to reduce the hardware footprint of capable AI systems while retaining useful benchmark performance.

• Open-source compressed AI model · 2026

  HyperNova 60B

  HyperNova 60B is an openly distributed compressed language model developed by Multiverse Computing. The company describes it as occupying about 32 GB and using less memory and providing lower latency than the larger model from which it was derived. Later iterations added improvements for tool calling and agentic coding, areas where inference efficiency can have a significant operational impact.

• AI application · 2026

  CompactifAI App

  CompactifAI App is an AI chat application built around locally executable models, including the compact Gilda model. Its distinguishing proposition is offline operation: users can interact with the application without sending every request to a remote service, provided the device has sufficient local resources.

• 2025

  Scale compressed AI modelsStrategy

  Background

  The company identified compressed models as a route to lower AI deployment costs and wider use on edge and private infrastructure.

  Action taken

  Multiverse raised a $215 million Series B and directed the financing toward scaling CompactifAI and its compressed-model portfolio.

  Aftermath

  The company expanded its compressed Llama offerings and promoted cloud, on-premises and device-level deployment.

  Series B investment. $215 million (June 2025)

  Source

• 2024

  Join the AWS Generative AI AcceleratorStrategy

  Background

  Multiverse sought broader access to cloud infrastructure, technical expertise and enterprise AI distribution channels.

  Action taken

  AWS selected Multiverse for its global 2024 Generative AI Accelerator program.

  Aftermath

  The company worked with AWS on optimized model access and later offered CompactifAI API through the AWS Marketplace.

  Source

• 2023

  Expand from quantum software into AI model compressionStrategy

  Background

  Large AI models were becoming increasingly expensive to store, serve and run on edge hardware, creating an adjacent commercial opportunity for Multiverse's tensor-network expertise.

  Action taken

  The company launched CompactifAI and began applying tensor-network methods to compress large language models and other AI systems.

  Aftermath

  AI model compression became a major part of Multiverse's product strategy, alongside quantum software and optimization.

  Source

• 2022.01.20

  Reported2022.01.20

  Multiverse Computing adds Fair Price feature to Singularity[3]Multiverse Computing为Singularity新增公平价格功能azoquantum.com

  Multiverse Computing announced a new Fair Price feature for its Singularity financial quantum computing solution, intended to provide more accurate fair-price calculations for financial institutions than current classical computing methods.

Read more

1. 1Multiverse Computingen.wikipedia.org
2. 2Multiverse Computing official websitemultiversecomputing.com
3. 3Multiverse Computing为Singularity新增公平价格功能azoquantum.com