Stealthium & Tenstorrent add AI runtime visibility

Stealthium and Tenstorrent have partnered to add runtime observability to AI workloads running on Tenstorrent systems, with a focus on visibility into the infrastructure layer that supports AI deployments.

Stealthium's platform will be offered as an integrated option for Tenstorrent deployments, giving customers a way to monitor accelerator activity, detect runtime anomalies and track workload execution on Tenstorrent hardware.

The announcement highlights a gap that security specialists increasingly identify in AI roll-outs: limited visibility into what happens when workloads run on specialised compute infrastructure. Many organisations have established tools for networks, cloud estates and applications, but the hardware and software layer handling AI processing often sits outside those controls.

That issue carries particular weight for businesses using AI in regulated or shared environments, where operators need to confirm workload isolation, understand how compute resources are being used and identify abnormal behaviour before it affects production systems.

Stealthium's software is designed to help operators see how AI infrastructure is being used and identify security or operational problems during runtime. Under the partnership, telemetry from Tenstorrent-based systems can also be integrated with customers' existing security and infrastructure monitoring platforms.

Tenstorrent, which develops AI processors and RISC-V-based CPUs, has positioned openness in its full-stack architecture as a way for partners to build additional software and monitoring layers around its systems. That approach is central to the agreement because Stealthium's offering depends on access across the AI software and hardware stack to provide direct runtime visibility into accelerators.

Security gap

The companies are targeting customers building more complex AI environments in sectors such as financial services, telecommunications and energy. Those industries often face stricter requirements around control, auditability and operational resilience, making detailed monitoring of AI infrastructure more important.

The partnership also comes as enterprises and cloud providers invest heavily in private and sovereign AI deployments. In those settings, organisations are under pressure not only to run models efficiently but also to show that workloads are isolated, systems are behaving as expected and incidents can be investigated when something goes wrong.

Ahmed Shosha outlined Stealthium's view of the market in a statement accompanying the announcement.

"AI is only as secure and trustworthy as the layer it runs on, yet for many that layer is currently invisible and indefensible. Stealthium exists to make AI accelerated compute observable, secure, and controlled. Tenstorrent is building exactly the kind of open, full-stack platform where that belongs from day one. Together we are offering customers an AI acceleration foundation they can see, verify, and trust as they scale," said Ahmed Shosha, Chief Executive Officer & Founder of Stealthium.

The partnership reflects a broader push in the AI infrastructure market to move beyond model governance and address runtime operations. Security teams have spent years building methods to inspect traffic, endpoints and cloud workloads, but AI accelerators and their surrounding software frameworks have introduced a layer that many existing tools were not designed to cover.

For operators, that creates practical as well as security concerns. Without detailed visibility into accelerator activity, they may struggle to explain failures, identify inefficient resource use or determine whether unexpected behaviour is the result of a software fault, misuse or malicious action.

Infrastructure focus

Tenstorrent has been expanding its profile as a supplier of AI compute systems and RISC-V intellectual property. The company builds processors and systems for developers, enterprises and sovereign infrastructure projects, and also licenses CPU and AI core technology to chip designers including Samsung and LG.

The business is led by Chief Executive Officer Jim Keller and has raised more than USD $1 billion from investors including Bezos Expeditions, Samsung, LG Electronics, Hyundai Motor Group and Fidelity, according to the company. Its operations span sites in North America, Europe and Asia.

Amr Elashmawi, Vice President of Strategy & Business Development at Tenstorrent, linked the agreement to demand for more controlled AI infrastructure.

"As enterprises and cloud providers move toward sovereign and private AI deployments, they need infrastructure that delivers high-performance compute and operational trust," said Amr Elashmawi, Vice President of Strategy & Business Development at Tenstorrent. "Tenstorrent's open, full-stack AI platform gives customers greater control over their deployments, while Stealthium extends visibility and security at the runtime layer. Together, we can help customers build enterprise-grade AI infrastructure that is more secure, more transparent, and ready for production at scale."