Expanse, a startup that predicts computing requirements for artificial intelligence workloads before execution, has closed a $5.3 million seed round. The investment was led by Crane Venture Partners with participation from PXN Group, Y Combinator, Standard Capital, Scribble Ventures, AIP Seed, General Advance, Ritual Capital, Robinhood Ventures, and Twenty Two Ventures. Angel investors include former DeepMind researchers and leaders from AI infrastructure teams.
The Scheduling Challenge
Teams running large scale compute often follow a familiar ritual before a job runs. They are asked how much hardware they want and for how long, but nobody truly knows the answer. Asking for too little can cause a job to die hours in, while asking for too much leaves hardware reserved and idle while other work waits.
Measured Waste
Expanse measured the scale of this problem directly after examining real cluster usage across demanding computing environments. In one cluster, the company identified nearly $8 million of idle compute capacity in a single month. The finding highlights how padding estimates can fill clusters with reservations that nobody is actively using while other work remains queued.
How Expanse Works
Expanse predicts what every job needs before it runs, an approach the company calls compute certainty. The models read the job's code, the cluster's own record of similar runs, and the hardware available, then forecast resource needs before a single GPU is committed. This allows organizations to avoid both job failures and unused reservations.
Deployment and Integration
The platform operates on Slurm, Kubernetes, and major cloud providers, supporting on-premises, hybrid, and cloud environments. Expanse sits inside a customer's own environment, so code and data never leave that infrastructure. The company contrasts this with monitoring tools that chart usage after the fact, noting that Expanse works before a job starts.
Industry Context
The timing of Expanse's solution aligns with growing pressure on AI infrastructure. New GPUs take months to arrive, and in a growing number of regions the power needed for new data centers is capped or refused. For many teams, the fastest path to new capacity is to improve utilization of hardware already installed.
Funding and Expansion Plans
The seed capital will be used to hire engineers and bring Expanse to more clusters across AI labs, quantitative finance, life sciences, and research computing. The company also plans to deepen its product capabilities for existing customers. By improving utilization, Expanse says teams can get as much as twice the work from the same hardware.
Founding Team and Investor Support
Expanse was founded by Ismaeel Bashir, Nikodem B., Yafet, and Eren Mendi, drawing on research at the Edinburgh Parallel Computing Centre and engineering leadership in quantitative finance. The round received support from Crane Venture Partners, PXN Group, Y Combinator, and Standard Capital, among other investors. The team also acknowledged early customers who allowed Expanse to run on their clusters before the company was widely known.
Expanse's seed round arrives at a moment when enterprises and research institutions face both surging AI demand and constrained hardware availability. The company's focus on pre-run prediction offers a practical route to recovering unused capacity without expanding physical infrastructure. Its ability to integrate within existing environments while keeping data in place could prove valuable across AI labs, finance, life sciences, and research computing.