The Pulse
Expanse Raises $5.3M to Predict AI Workload Needs
Expanse has raised a $5.3 million seed round led by Crane Venture Partners for software that predicts GPU, CPU, memory and runtime requirements before AI workloads run. The company will use the funding to expand engineering, develop its pla
AI.info Team ·
“Today, every AI workload begins with an educated guess. Engineers shouldn’t have to predict exactly how much compute their code will need before they press run. The machine should carry the uncertainty, not the person.”
Ismaeel Bashir, co-founder and CEO of Expanse
Expanse targets the guesswork before a job starts
Expanse, an AI infrastructure company with operations in San Francisco and London, announced a $5.3 million seed round on September 16, 2026. Crane Venture Partners led the financing, with participation from PXN Ventures and angel investors that include former DeepMind researchers and leaders from AI infrastructure teams.
The company develops software that predicts the resources an AI workload will require before execution. Its recommendations cover GPU and CPU allocation, memory and expected runtime, allowing infrastructure teams to configure jobs before they consume capacity or enter a scheduler queue.
Expanse says its system is designed for cloud and on-premises environments. The models run inside a customer’s infrastructure, keeping code and telemetry within that environment while examining the likely needs of each workload.
Why GPU allocation has become an operations problem
AI engineers commonly estimate resource requirements before launching training, inference or simulation jobs. An estimate that is too high leaves capacity unused; an estimate that is too low can cause a failed job, wasted queue time or an interruption that forces the workload to run again.
The company’s funding announcement cites industry estimates that about 30% of cloud spending is lost through over-allocation. It also points to Microsoft Research findings that GPU utilisation sits at around 50% across many internal deep-learning workloads. Those figures frame the commercial problem Expanse is pursuing: organisations may have substantial computing capacity available while still struggling to assign it efficiently.
Expanse distinguishes its product from monitoring and observability tools that explain a workload after it has completed. Its software operates before execution, recommending a configuration intended to reduce failed jobs and allow existing infrastructure to process more work.
A reported $8 million capacity gap
Expanse says one production deployment identified nearly $8 million in idle compute capacity during a single month. The company does not name the customer or provide a breakdown of how the figure was calculated in the funding announcement, so the result is best understood as a customer-reported outcome rather than an independently verified industry measure.
The startup was founded by four University of Edinburgh graduate engineers: Bashir, Nikodem Bieniek, Eren Mendi and Yafet Melake. The founders previously worked with large-scale computing infrastructure in quantitative finance and national supercomputing environments.
While at the Edinburgh Parallel Computing Centre, Bashir developed a multimodal high-performance-computing resource prediction system. The founders say their experience showed that even sophisticated compute environments still depended heavily on manual estimates when allocating resources.
Crane and PXN back a broader infrastructure push
Scott Sage, co-founder and partner at Crane Venture Partners, said the company’s software gives organisations a way to recover capacity from infrastructure they already own rather than relying only on new hardware purchases.
“The future of AI won't be defined only by who builds the biggest clusters, but by who uses them most intelligently,” Sage said. “Expanse gives organisations something they don't have today: certainty.”
Andy Barrow, managing director of PXN Ventures, said Expanse is addressing the cost and scaling problems created by the growing demand for AI infrastructure. Greg Steinbrecher also described efficient allocation of GPU hours across shared systems as one of the hardest problems in AI infrastructure, according to the company’s announcement.
Funding will support engineering and new deployments
Expanse plans to use the new capital to expand its engineering team, accelerate product development and bring its platform to more organisations. The company lists AI infrastructure, quantitative finance, life sciences, research and high-performance computing among its target markets.
Its product page describes support for SLURM, Kubernetes and Nomad clusters, with Expanse operating alongside existing schedulers rather than replacing them. The company says its system can predict GPU allocation, memory use, runtime and completion probability before a workload runs.
The immediate business case is straightforward: reduce over-allocation, prevent avoidable failures and make more of a fixed pool of computing hardware available for production work. Expanse’s $5.3 million round gives the four-person founding team capital to turn that operating thesis into a wider commercial product.
Read the company’s funding announcement at GlobeNewswire.