Callosum Raises $100M Seed to Route AI Workloads Across Chips and Models

Founded by Cambridge neuroscientists, Callosum promotes an approach it terms heterogeneous intelligence. According to Tech.eu, the company challenges the view that greater AI capability will come from scaling a single model on identical chips, a process it describes as energy- and capital-intensive. Callosum argues that real-world problems are complex and varied and require different capabilities, and says its approach makes solving them faster and cheaper at scale.

The startup has announced a partnership with US chip company Cerebras. Its CEO, Andrew Feldman, said that by integrating Cerebras into Callosum's platform, the companies are making ultra-low-latency inference available where it creates value, according to Tech.eu.

Atomico CEO Niklas Zennström said that as AI moves from training to inference, the question is no longer only who builds the best models but how intelligence is delivered efficiently across a diverse landscape of models and chips, EU-Startups reported. He described Callosum as building a software layer that routes AI workloads to the right compute.

Callosum said the funding would help it redefine how humanity computes for AI's next chapter, according to EU-Startups.