Callosum Raises $100M Seed Round Led by Atomico to Scale AI Infrastructure

Callosum Raises $100M Seed Round Led by Atomico to Scale AI Infrastructure

London-based AI infrastructure startup Callosum has raised $100 million in a seed funding round led by Atomico, making it one of the largest seed rounds in Europe and bringing the company's total funding to approximately $110 million since its foundation in 2025.

The round includes participation from Plural, DCVC, the UK's Sovereign AI Fund, other investors and angel backers. The latest financing comes just months after Callosum raised $10.25 million in pre-seed funding in February.

The company has not disclosed its valuation.

Founded by Danyal Akarca and Jascha Achterberg, Callosum was created by two founders who began working together during their PhDs at the University of Cambridge, where they studied neuroscience, computing and artificial intelligence. Their research has appeared in Nature journals, and both founders have also held positions at Intel and Google DeepMind.

Callosum is developing software designed to make AI inference more efficient by distributing different parts of an AI workload across different models and types of computing hardware.

The company's approach is based on the idea that AI workloads do not necessarily need to run entirely on identical chips. Instead, Callosum's software breaks an AI task into separate stages and assigns each stage to the hardware and model best suited to handle it.

For example, one stage of an autonomous computing workload may require rapid pattern matching, while another may require more complex reasoning. Callosum's system is designed to route those workloads to different computing resources rather than forcing a single type of hardware to handle the entire process.

The company claims its approach can deliver improvements of up to two times in accuracy, seven times in speed and four times in cost for complex workloads compared with uniform hardware. The detailed technical benchmark data supporting those claims has not been publicly disclosed.

Callosum's technology is targeting a major challenge in the AI industry as spending increasingly shifts from model training toward inference. As companies deploy AI systems at greater scale, the cost and efficiency of running those models can become a significant part of their infrastructure spending.

The company describes Nvidia's CUDA ecosystem as one of its main competitive challenges. Nvidia currently dominates the global GPU market, while a growing number of specialised chip companies are developing hardware designed to improve AI inference performance.

Callosum is taking a chip-agnostic approach rather than developing a single proprietary processor. The company aims to integrate multiple types of hardware into a unified infrastructure layer, allowing customers to select different computing resources for different parts of their AI workloads.

The startup has already begun establishing partnerships with semiconductor companies. Callosum is working with Cerebras on low-latency inference at scale and has also entered into a collaboration with Korean AI chipmaker Rebellions.

The Cerebras partnership is focused on integrating Cerebras' hardware into Callosum's platform, allowing customers to use its processing capabilities alongside other computing architectures. Cerebras CEO Andrew Feldman said the integration is intended to make ultra-low-latency inference available for workloads where it can have the greatest impact.

Callosum is also working with Rebellions to support a multi-chip approach rather than relying on a single hardware vendor. The company believes this could help customers avoid vendor lock-in and deploy different architectures according to workload requirements and geographic availability.

The new $100 million financing will be used to develop Callosum into what the company describes as a “global heterogeneous integrator” and to establish additional computing partnerships.

The funding represents a significant increase in the company's capital base just months after its pre-seed round. Callosum has now raised approximately $110 million since its foundation, giving the company substantial resources to expand its technology and commercial operations at an early stage.

The UK government is also positioning AI infrastructure and semiconductor access as strategic priorities. The country's Sovereign AI Fund participated in Callosum's latest round, reflecting growing interest in developing AI infrastructure capabilities and access to computing resources within Europe.

Callosum is entering a rapidly expanding AI infrastructure market in which companies are seeking ways to reduce inference costs, improve performance and make better use of increasingly diverse computing hardware.

For the startup, the challenge will be turning its heterogeneous computing approach into a widely adopted infrastructure layer while demonstrating that its claimed performance and cost advantages can translate into measurable benefits for enterprise AI customers.