London-based Callosum raises €85.4 million in Atomico-led Seed round to unify AI models and chips

London-based Callosum, a startup developing infrastructure to unify diverse AI models and chips, has secured €85.4 million ($100 million) in a Seed round led by Atomico. The funding round saw participation from Atomico, Plural, DCVC, the UK Sovereign AI Fund, and several other investors and angels. This capital injection will support Callosum in "redefining how humanity computes for AI's next chapter," according to the company.

With AI transitioning from training to inference, the focus has shifted from just building the best models to efficiently delivering intelligence across a diverse landscape of models and chips. Niklas Zennström, Atomico's CEO, highlighted this shift, emphasizing Callosum's role in building the software layer that intelligently routes AI workloads to the most suitable compute, enhancing speed, efficiency, and accessibility without added complexity.

The Seed round follows a pre-Seed raise of €8.76 million ($10.25 million) in February 2026, led by Plural and supported by 22 investors, including the UK’s Advanced Research + Invention Agency (ARIA). Founded in 2025 by Jascha Achterberg and Danyal Akarca, Callosum's mission is rooted in the belief that AI's future lies not in the biggest models or the best infrastructure, but in orchestrating various models across different types of silicon.

The funding announcement coincides with the UK Sovereign AI Fund's first-ever investment in Callosum and its inclusion in the UK government’s £1.1 billion AI hardware plan. Kanishka Narayan, the UK Minister for Artificial Intelligence, emphasized the critical role of chips in AI development, noting that the demand for these chips is set to grow exponentially.

Callosum's approach revolves around "heterogeneous intelligence," aiming to harness the diversity of AI models and specialized chips to unlock significant gains in AI capability and cost-efficiency. The company revealed three key developments as part of its latest funding:

1. A series of global compute partnerships, including a flagship deal with Cerebras Systems to deliver low-latency, multi-agent AI at scale, and collaborations with chipmaker Rebellions and other infrastructure providers and Original Equipment Manufacturers (OEMs) worldwide.

2. Advances in "programmable heterogeneity," enabling AI workloads to be broken into smaller tasks and routed to the most efficient model and chip based on constraints such as cost, energy, and speed.

3. The rollout of "Tailored Inference" in production, a suite of APIs delivering performance and cost improvements across sectors like cybersecurity and finance.

Andrew Feldman, CEO of Cerebras Systems, highlighted the integration of Cerebras into Callosum's platform, enabling ultra-low-latency inference where it matters most, thereby enabling the creation of AI systems previously impractical.

With a founding team comprising experts in neuroscience and hardware-software co-design, and members from prestigious institutions such as Cambridge, Oxford, MIT, Imperial College London, Microsoft, ETH Zurich, and Intel, Callosum has assembled a team capable of driving innovation in the AI landscape. The company's growth has been marked by the addition of founding engineers, technical staff, and operational leaders based in the UK and the US.

Callosum's strategy addresses the traditional bottleneck of AI infrastructure, where the move to bigger models on the same chips is no longer viable due to the inherently heterogeneous nature of real-world AI problems. By offering a library of APIs that route AI workloads across a global network of silicon partners, continuously optimizing combinations of models and hardware to stay on the "Pareto frontier" of cost, speed, and performance, Callosum positions its "Tailored Inference" product as the solution to these challenges.