Delivering low-latency GPU experiences using AWS Local Zones with Stim.io

Stim.io was founded on the belief that access to computing should not be limited by owning hardware. Through its offerings, the company provides access to cloud-based computing resources from a wide range of devices, supporting gaming, virtual desktops, engineering applications, and other GPU-intensive workloads.

However, delivering these experiences across long distances presents additional challenges. For highly interactive workloads—such as cloud gaming, virtual desktop sessions, and engineering applications—responsiveness is critical because even small delays can affect the user experience. The farther users are from compute resources, the more difficult it becomes to maintain the responsive experience those workloads require due to latency and jitter on the network.

Stim.io also needed the flexibility to support users around the world while responding to changing demand. Because capacity requirements can shift across locations, the company needed infrastructure that could scale without deploying dedicated hardware near every user. “Every millisecond adds up,” says Ian McLoughlin, founder and CEO of Stim.io. “When the overall latency to render something in a game is above 100 ms, you’re right on that threshold of a human noticing it.”

To deliver low latency for customers, Stim.io uses AWS Local Zones, which bring core services closer to end users for low latency and data residency. The company uses AWS Local Zones to place GPU-powered compute closer to users, helping reduce latency that can affect highly interactive, jitter-sensitive workloads. This proximity is particularly important for workloads that depend on rapid response times and consistent performance. “When we’re delivering low-latency GPU experiences to customers, we need latency under 50 ms, and we can only achieve that by using AWS Local Zones,” says McLoughlin.

Stim.io delivers cloud PCs through Stim Secure Cloud and operates The Stim Cafe, a marketplace where users access GPU-powered computing resources without dedicated hardware.

A key component of the solution is Amazon Elastic Compute Cloud (Amazon EC2) G7e instances, which provide high-performance NVIDIA GPU-based instances. Stim.io uses these instances to power cloud streaming, gaming, and virtual desktop workloads.

AWS also provides Stim.io with the flexibility to respond to changing demand patterns. Rather than deploying dedicated infrastructure near every user, Stim.io can provision GPU resources where they are needed while simplifying infrastructure management. “We let AWS do everything from API management to orchestration to container services so that we can focus on our streaming technology,” says McLoughlin.

Using AWS Local Zones, Stim.io delivers more than 140 frames per second at 4K resolution across GPU-powered workloads, including cloud gaming, computer-aided design, and AI model visualization. Stim.io also supports streaming at up to 240 Hz with latency of under 10 ms for its cloud PC experience. These capabilities provide users with access to workstation-class performance without purchasing and maintaining specialized hardware. Users rely on Stim.io as their primary computing environment, with the average user spending more than 5 hours a day on the service.

Looking ahead, Stim.io plans to support a growing range of cloud-delivered workloads that benefit from low-latency GPU access, including virtual desktops, engineering applications, and AI workloads. “From phones to robotics, we’re trying to be the provider of anything that needs low-latency GPU,” says McLoughlin.