U.S. and U.K.-Based Expanse Raises €4.5 Million to Make AI Infrastructure More Predictive

Expanse has raised €4.5 million ($5.3 million) in Seed funding to expand its engineering team, accelerate product development and bring its predictive AI infrastructure platform to more organisations. The round was led by Crane Venture Partners, with participation from PXN Ventures, AIP Seed and angel investors including former DeepMind researchers.

Founded in 2025, Expanse is building what it describes as an intelligence layer for AI and high-performance computing infrastructure. Its software predicts the resources an AI workload will require before it runs, helping engineering teams avoid over-allocation, failed jobs and wasted GPU capacity. The company's models can operate within customers' cloud or on-premises environments, keeping code and telemetry inside the organisation's infrastructure.

The company was founded by four University of Edinburgh engineers: Ismaeel Bashir, co-founder and CEO; Nikodem Bieniek, co-founder and CTO; Eren Mendi, co-founder and CPO; and Yafet Melake, co-founder and COO. The team previously worked with large-scale compute infrastructure across quantitative finance and national supercomputing environments.

Expanse is focused on a fundamental challenge in AI infrastructure: engineers often have to estimate how much GPU, CPU, memory and runtime a workload will need before submitting it. Its platform replaces those estimates with predictions based on evidence from the underlying infrastructure. In one production deployment, the company identified nearly $8 million (€6.9 million) of idle compute capacity in a single month.

The startup plans to use the new capital to strengthen engineering, accelerate its product roadmap and expand adoption across AI infrastructure, quantitative finance, life sciences, research and high-performance computing. Its broader goal is to help organisations extract more useful work from the compute infrastructure they already have, rather than relying solely on additional hardware.