Thunder Compute today announced it has raised $13 million in early funding to help graphics processing unit cloud providers squeeze out the last erg of compute capacity that sits idle and wasted at the long end of workload cycles. The company uses proprietary software to treat GPUs as network resources, making them accessible to workloads across the data center. This makes them more flexible and efficient, allowing capacity to be allocated much like storage and central processing unit resources, exactly when it is needed. The virtualization is invisible to the developer, and really the goal is for them to not care.
Virtualizing the hardware layer
Because GPUs are traditionally allocated as bare-metal resources, dedicated to individual workloads, these expensive chips can spend much of their reserved time sitting idle, waiting for work. Thunder Compute created software to treat GPUs as network resources, making them accessible to workloads across the data center. The firm sits between the developer and the cloud provider, abstracting away the GPU as a physical chip so it can drop directly into the existing workflow. The virtualization is invisible to the developer, and really the goal is for them to not care.
Related: Farmworkers endure hidden health and labor dangers
According to the CastAI 2026 State of Kubernetes Optimization Report, enterprise GPUs sit idle, averaging around 5% to 20% utilization. This waste leaves a tremendous amount of capacity at the long end, which the company says amounts to almost $200 billion left on the table waiting to be used. Much of that underutilization comes from how GPUs are reserved: They are allocated continuously to workloads regardless of whether they are actually being used. Thunder’s software separates a workload’s access to a GPU from the specific hardware serving it, allowing the underlying fleet to be scheduled much more efficiently.
The economic benefit goes primarily to the cloud provider or enterprise firm that bought the GPUs in the first place. Developers should barely notice that the startup is there; they simply ask for a GPU and get one. The payoff happens behind the scenes: Higher utilization lets operators squeeze more work out of the hardware they already paid for and potentially pass some of those savings back through lower cloud prices. To date, Thunder Compute has supplied compute to more than 10,000 users on its own cloud of virtualized GPUs.
Related: Ukrainian drones destroy US armor in simulation
Scaling the technology
The Series A marks a shift from proving the technology on its own cloud toward putting it into the hands of cloud providers and enterprises that already operate GPU infrastructure at scale. The funding will also help the startup build out the organization needed to support that shift. Thunder plans to hire systems researchers to continue pushing its virtualization technology, engineers to support enterprise GPU deployments and a sales team as it expands its go-to-market effort.
Until now, the company has effectively been selling to itself, running its own cloud service as a testbed for GPU use cases and becoming its own GPU cloud provider before scaling the software outward. The fastest way to do that was to launch our own cloud. The firm sits between the developer and the cloud provider, abstracting away the GPU as a physical chip so it can drop directly into the existing workflow. The virtualization is invisible to the developer, and really the goal is for them to not care.