Quantum computing software startup Multiverse Computing S.L. said today it has raised €25M (USD $27.1 million) in a new early-stage funding round.

The funds from the oversubscribed Series A round will enable it use emerging quantum computers to try and tackle some of the world’s biggest challenges around artificial intelligence and environmental change.

Today’s round was led by Columbus Venture Partners and saw significant contributions from Quantonation Ventures and new investors, including the European Innovation Council Fund, Redstone QAI Quantum Fund and Indi Partners, among other prominent backers.

Multiverse doesn’t build quantum computers itself. Rather, it specializes in the development of what it calls “quantum-inspired algorithms,” which are advanced tensor networks based on the principles of quantum computing. They’re said to be the most efficient algorithms that can run in today’s existing classical computers, and they’re especially suited for industries such as AI, energy, manufacturing, finance, defense and healthcare.

Its main offerings include a product called Singularity, which helps users with no quantum background solve challenges around AI and optimization applications using quantum and quantum-inspired computers. The startup has also built a product called CompactifAI, which is designed to compress large language models using quantum-inspired tensor networks in order to reduce their size by up to 80% without affecting their performance. It’s a promising development because LLMs are what sit at the heart of powerful generative AI applications such as ChatGPT.

With respect to LLMs, the company says their growing power is going to be a big problem, since the computational costs of such workloads are exploding. It’s estimated that the costs of training the most powerful LLMs available today can exceed $100 million, yet we’re still only at the dawn of the nascent generative AI industry.

As a result, the company said, LLM training costs will likely double every 10 months. Such costs are unsustainable, and unless a new technological paradigm is introduced, the AI industry will consume intolerable amounts of energy and money.

This is where Multiverse’s CompactifAI can make a difference. It’s uniquely able to compact AI models using Tensorized networks that have been shown to train LLMs faster and at much lower costs, with early studies showing it has achieved accelerations of up to 1,000 times.

According to the startup, by tensorizing LLMs it can locate the most relevant and also the not-so-relevant parts of any model and compress them in order to significantly reduce its size, without compromising performance. This is key, not only from the perspective of saving costs, but also in terms of practicality. For instance, the enormous resources required by LLMs make them impossible to deploy at the edge, limiting their usefulness in autonomous vehicles, for example.

Multiverse’s quantum software can also accelerate other AI applications, such as computer vision in manufacturing, localized weather forecasting, anomaly detection in cybersecurity, AI-based trading and AI-based protein design.

The startup has also developed an intriguing LLM Lobotomizer, which is designed to help trained LLMs forget undesirable information they have learned. Believe it or not, this is presently very difficult to do, and it can cause problems such as toxic AI or AI hallucinations, which is when AI generates fake responses. The best method for helping LLMs forget something they’ve learned is to use post-training filters, but studies show that these aren’t always effective.

In addition, Multiverse said, its quantum and quantum-inspired software algorithms also have great potential in terms of environmental sustainability, helping to solve difficulties around electrical distribution in renewable energies, battery design, hydrogen production and more.

Multiverse’s advances in these areas are the reason why it was recently named as one of the world’s most promising 100 AI companies by CB Insights.

“Singularity has made the benefits of quantum and quantum-inspired computing more accessible to industry, and we will continue diversifying its applications,” said Roman Orus, chief scientific officer at Multiverse.

Multiverse explains quantum-inspired algorithms can outperform classical computers in many complex tasks, and for that reason they have the potential to solve many currently intractable challenges that will never be overcome using traditional methods. It’s for this reason that McKinsey & Co. estimates quantum technology could deliver up to $1.3 trillion in value over the next decade, as it matures.

Javier Garcia, a partner at Columbus Venture Partners, said Multiverse is poised to deliver groundbreaking solutions in areas such as AI, life sciences and biotechnology. “Columbus Venture Partners will help to identify unmet market needs and high-profile industrial partners,” he added.

The startup said the funding from today’s round will help it to accelerate the development of its quantum-inspired algorithms and support its global expansion, with a particular focus on the U.S.

Images: Multiverse Computing