Stealthium has joined NVIDIA Inception, NVIDIA’s program for startups building on accelerated computing. Stealthium is the runtime observability and security company for GPUs. We make AI accelerated compute observable, secure, and controllable.

Joining Inception is the first step in a deliberate direction for us: working more closely with the full NVIDIA stack, and building security and observability made specifically for the runtime layers NVIDIA customers depend on to train, serve, and scale AI.

Runtime Is Where AI Happens

For most organizations running AI in production, the NVIDIA stack is the runtime. That means the driver, CUDA, the container runtime, GPU partitioning, and the software and frameworks above them that schedules and serves models. Models are trained there, weights sit in GPU memory there, and inference is served from there.

Security tooling does not currently reach these levels of the AI runtime stack. Instrumentation built for the CPU stops at the driver boundary. Whatever executes on the GPU never reaches it: the kernels, device memory, and the workload sharing the device next door. Stealthium exists to close that gap.

What Comes Next

Inception gives us access to NVIDIA’s developer resources, technical training through the Deep Learning Institute, and the wider NVIDIA ecosystem. We will use it to go deeper, building security and observability tailored to NVIDIA’s GPU and AI runtime offerings so the teams running them get:

• Observable: GPU-level telemetry correlated from process to device, and across nodes and clusters.
• Secure: Runtime detection for attacks that execute on the GPU (unseen by CPU-focused tooling).
• Controllable: Policy and response at the layer where the workload actually runs.

As the NVIDIA stack evolves, Stealthium will evolve with it.

“The AI Runtime stack is where inference is served, where training runs, and where model weights & IP live in memory. It’s also where security has been absent. For most of our customers, AI is NVIDIA. Joining NVIDIA Inception is the start of building Stealthium deeper into that stack, so the teams running it can see and trust what is executing on their GPUs.”

Ahmed Shosha, CEO & Founder, Stealthium

“From neoclouds to enterprises building AI factories, the people we work with standardize on NVIDIA. Inception brings us closer to that ecosystem and the teams building on it, and helps us bring runtime security and observability to every environment where NVIDIA GPUs do the work.”

Sherif El Nabawi, CBO & Co-Founder, Stealthium

See What’s Running On Your GPUs

Running on NVIDIA infrastructure? Talk to us to see how Stealthium makes your GPU runtime observable, secure, and controllable.