London-based artificial intelligence startup Callosum has raised $100 million [1] in seed and early-stage financing to reduce the cost of AI tasks.
The funding targets the high cost of AI operations by allowing enterprises to move workloads across different hardware providers. This approach aims to break the current industry reliance on a single chip vendor.
Callosum is developing software designed to match specific AI tasks with the most suitable models and chips [1]. The technology enables companies to route workloads across various hardware options, including Nvidia, AMD, and Cerebras, without needing to rewrite their existing code [2]. By optimizing which chip handles which task, the startup intends to lower the overall financial burden of running large-scale AI systems [3].
The $100 million [1] round follows a previous funding event in February, during which the company raised $10.25 million [4]. The latest investment was backed by several firms, including Atomico, Plural, and DCVC, as well as the UK Sovereign AI Fund.
By creating a layer of orchestration between the AI model and the hardware, Callosum seeks to bypass the "monoculture" of current chip dominance [2]. This flexibility allows businesses to switch hardware providers based on price or performance availability without facing significant technical hurdles.
The startup's headquarters are located in London, where it continues to develop the routing software [5]. The company has not disclosed specific details regarding the timeline for a full commercial rollout of the software.
“Callosum has raised $100 million in seed and early-stage financing to reduce the cost of AI tasks.”
This investment signals a growing corporate push toward hardware agnosticism in the AI sector. By decoupling software from specific chip architectures, enterprises can mitigate supply chain risks and reduce operational costs, potentially challenging the market dominance of leading GPU manufacturers.
