About Expanse
Expanse helps AI and HPC teams get more out of the GPUs they already have by predicting what each job needs before it runs. It offers resource prediction, optimization suggestions, and failure prediction for workloads on Slurm, Kubernetes, and major clouds. It is aimed at research labs, quant funds, and organizations running large GPU clusters.
What is Expanse?
What is Expanse?
Expanse targets a common inefficiency in GPU clusters: engineers have to guess how much GPU, memory, and time a job will need before submitting it. Over-provisioning wastes capacity, and under-provisioning leads to jobs that fail hours in. Expanse analyzes job code, historical cluster data, and available hardware to forecast resource needs before a job reaches the scheduler.
The product has three capabilities. Resource prediction right-sizes job submissions. Optimization suggestions give researchers code and configuration changes they can apply themselves. Failure prediction flags jobs likely to fail before they consume GPU time. Expanse works with Slurm, Kubernetes, and major cloud platforms, on-premise or hybrid, and processes data inside the customer's own environment. In its seed announcement, the company said one cluster it analyzed showed nearly $8 million of idle compute capacity in a single month.
Expanse is part of Y Combinator's Spring 2026 batch and is based in San Francisco. Its four founders, led by CEO Ismaeel Bashir, previously ran HPC and GPU training workloads at quant funds and national supercomputing centers. In September 2026 Expanse announced a $5.3 million seed round led by Crane Venture Partners.
Pros
- Addresses a measurable cost problem: idle or failed GPU jobs
- Supports Slurm, Kubernetes, and major cloud platforms
- Runs inside the customer's environment rather than sending data out
- Combines prediction with concrete code and config suggestions for researchers
Cons
- Aimed at organizations with sizable clusters; not relevant for individual users
- Onboarding is through a discovery call, with no self-serve pricing shown
- Early-stage company with a small founding team
How to Use Expanse
- 1Book a discovery call
Contact Expanse through its website to discuss your cluster and workloads.
- 2Deploy in your environment
Set Expanse up alongside your Slurm, Kubernetes, or cloud scheduler, on-premise or hybrid.
- 3Analyze jobs before they run
Expanse returns evidence-backed recommendations for GPUs, memory, runtime, and completion probability.
- 4Apply suggestions
Use the recommendations and optimization suggestions to right-size jobs and avoid failures.