Thunder Compute raises $13M to squeeze more work out of idle GPUs

Thunder Compute has secured $13 million in early funding with the aim of maximizing utilization of idle GPUs in cloud computing. GPUs are typically allocated as dedicated resources for individual workloads, which often leads to extended periods of underutilization. According to CastAI 2026 State of Kubernetes Optimization Report, enterprise GPUs average between 5% and 20% utilization, leaving a substantial amount of untapped computing power.

By virtualizing GPUs, Thunder Compute seeks to create a platform akin to VMware for GPU resources. The software abstracts away the physical GPU, allowing for more efficient scheduling and allocation of computing capacity, much like storage or central processing units. This approach addresses the historical lack of GPU virtualization that exists in other types of computer hardware.

Thunder Compute's proprietary software separates a workload's GPU access from the specific hardware serving it, enabling the fleet to be managed more efficiently. By treating GPUs as network resources, the company allows workloads to access GPU capacity on-demand, optimizing utilization and potentially reducing cloud costs for enterprises.

The company is positioned between developers and cloud providers, making GPUs accessible to workloads directly within existing workflows without requiring developers to concern themselves with the underlying hardware. This enables higher utilization of existing GPU investments, with the potential to pass savings onto cloud providers or enterprises.

Thunder Compute currently serves over 10,000 users on its own cloud of virtualized GPUs, with plans to expand this capability to larger GPU fleets. The recent Series A funding will facilitate the company's transition from a proof-of-concept to a service offered to other cloud providers and enterprises with existing GPU infrastructure.