From WhatsApp friends to a $500 million–plus valuation: These founders argue their tiny AI models are better for customers and the planet

Multiverse Computing is emerging as a key player in the 'frugal AI' movement, developing small language models (SLMs) by compressing large language models (LLMs) to run efficiently on standard CPUs, thereby significantly reducing energy consumption and operational costs with minimal accuracy loss. The company, which has raised $290 million, is valued over $500 million, and projects $25 million in sales this year, addresses the escalating financial and environmental impact of traditional LLMs. Its quantum-inspired technology is being adopted by diverse clients including defense contractors, financial institutions like Moody's and Bank of Canada, and public service providers, positioning Multiverse as a potential disruptor offering more accessible and sustainable AI solutions across various industries, despite some expert skepticism regarding long-term accuracy and energy savings with widespread adoption.

Analysis

Multiverse Computing is positioning itself as a key innovator in "frugal AI," developing small language models (SLMs) that significantly compress large language models (LLMs) to run efficiently on standard CPUs. This proprietary "quantum-inspired" technology reportedly reduces energy consumption by 84% and compute costs with only a 2-3% accuracy loss, addressing the high financial and environmental burdens of traditional LLMs. The company, valued over $500 million with $290 million raised, projects $25 million in sales this year, indicating early commercial traction.

The firm's approach directly counters the escalating costs associated with GPU-intensive LLM deployment, a concern echoed by industry leaders like Sam Altman and Jensen Huang. Multiverse has secured diverse clients, including defense contractors, financial institutions like Moody's (MCO) and Bank of Canada, and public service providers collaborating with Intel (INTC) and Deloitte. Its successful redesign of Telefónica's (TEF) customer service system, drastically cutting LLM costs, exemplifies its value proposition.

While some experts question the long-term performance and accuracy of highly compressed models, particularly for complex or multilingual applications, Multiverse's ability to shrink open-source models like Meta's (META) Llama series and Mistral AI demonstrates significant technological capability. The company's strategic move to remove censorship from models like DeepSeek also highlights a unique market differentiation.

Key Decisions for Investors

• Investors should monitor Multiverse Computing's commercial scaling and the independent validation of its claimed 84% energy efficiency and minimal accuracy loss across diverse enterprise applications.
• Evaluate the potential long-term impact of 'frugal AI' solutions on companies heavily invested in GPU-centric AI infrastructure (e.g., NVDA) and large-scale data center operations (e.g., GOOGL), considering potential shifts in demand.
• Assess the competitive landscape and how established AI players might respond to Multiverse's disruptive, cost-effective SLM technology, particularly in industries with high energy costs or regulatory pressures for sustainability.