UK Startup CallosumAI Raises $10M to Challenge Nvidia

CallosumAI was founded by Cambridge scientists Danyal Akarca and Jascha Achterberg, who met during their PhDs. They are building "systems-level software" to counter the prevailing "AI monoculture," where single AI models are scaled on identical, expensive chips from a handful of dominant companies. Their goal is to create infrastructure where different AI models and diverse chip architectures can work together efficiently. The company's core thesis is that chip diversity, often seen as a disadvantage, can be exploited for better performance. Instead of just optimizing a single algorithm, CallosumAI's software is designed to control and extract benefits from the entire system, from the application down to the silicon. This approach targets complex, real-world problems that require a variety of capabilities, which they term "heterogeneous" tasks. This strategy directly challenges Nvidia's market position, which holds over 80% of the market for AI training and deployment GPUs. The high cost of Nvidia's top-tier chips, like the H100 which can cost up to $40,000, and the newer Blackwell chips priced between $30,000 and $40,000, creates a significant barrier to entry and a strong demand for more accessible alternatives. The pre-seed funding round was led by Plural, a venture capital firm founded in 2021 by former startup founders. Plural focuses on early-stage European companies with the potential for major impact, with typical investments ranging from €1-15 million. Their portfolio includes notable tech companies like Robin AI and Proxima Fusion. Entrepreneur First, the other backer, is a talent investor that supports individuals from the "pre-team, pre-idea" stage. Founded in 2011 by Matt Clifford and Alice Bentinck, EF provides a stipend and a structured program to help talented individuals find co-founders and develop business ideas, having helped create over 600 companies with a combined value exceeding $11 billion.