As the artificial intelligence industry pours billions of dollars into ever larger models and increasingly powerful chips, UK startup Callosum is taking a different route. The Cambridge founded company believes the next phase of AI will depend less on building one dominant model and more on creating infrastructure that can combine different models and computing systems to solve complex problems efficiently.
$100 Million Seed Round
Callosum has raised $100 million in a seed funding round led by European venture capital firm Atomico. The round also attracted significant participation from Plural, DCVC, Sovereign AI and a group of angel investors.
The investment comes just months after Callosum emerged from stealth with $10.25 million in funding in February. The latest financing makes the company one of the most highly valued early stage AI infrastructure startups to emerge from Europe.
The new capital will support the development of Callosum’s technology and help the company expand its infrastructure for running advanced AI workloads across different computing architectures and models.
Challenging the AI Monoculture
Callosum was founded by Cambridge neuroscientists who question the increasingly dominant assumption that artificial general intelligence or superintelligence will emerge by scaling a single enormous AI model and running it on vast fleets of similar chips.
The company describes this approach as an AI “monoculture”.
According to Callosum, real world problems are rarely uniform. They can require different forms of reasoning, specialised knowledge and varying levels of computational power. Instead of forcing every task through one model and one type of processor, the startup is developing infrastructure that can select and combine different AI models and chips according to the requirements of the task.
The company believes this approach could make AI systems faster and less expensive while allowing organisations to tackle complex workloads at much larger scale.
Making AI Infrastructure More Flexible
At the centre of Callosum’s proposition is software designed to allow AI workloads to operate across different chips and models.
This chip agnostic approach is significant because the AI computing market has become heavily concentrated around a small number of hardware and model providers. Training and operating increasingly sophisticated AI systems requires enormous amounts of computing power, energy and capital.
Callosum argues that relying on a single model architecture and hardware platform can increase these costs while limiting flexibility.
Its infrastructure is instead designed to allow businesses to use different combinations of AI models and computing resources, potentially matching particular workloads with the technology best suited to them.
Cerebras Partnership
Callosum has also announced a partnership with US AI chipmaker Cerebras, bringing its high performance computing technology into the startup’s platform.
Cerebras CEO Andrew Feldman said integrating Cerebras into Callosum’s infrastructure would make ultra low latency inference available where it can have the greatest impact. He said the combination could allow customers to develop AI systems that were previously impractical.
The partnership gives Callosum access to another important computing architecture as it develops its broader multi chip and multi model strategy.
A Different Path for European AI
Callosum’s approach has attracted attention from both investors and policymakers. The company has received support from the UK AI minister and is positioning itself within a European AI ecosystem seeking greater technological independence and infrastructure flexibility.
The startup’s argument is that the future of AI does not necessarily have to revolve around one model, one company or one type of chip.
Instead, Callosum wants to build the infrastructure that allows different AI models and computing systems to work together, giving businesses greater choice while potentially reducing the cost and energy required to deploy advanced AI.
With its new $100 million war chest, the company now has the resources to turn that alternative vision into a commercial AI infrastructure platform.
