Quali unveils Torque AI control plane for GPU clouds

Quali has expanded its Torque platform by introducing an Agentic Control Plane designed to govern autonomous AI workloads and GPU-intensive infrastructure in real time. The company noted that organisations are currently navigating a fundamental shift in how AI systems consume and modify infrastructure, moving away from static, trigger-based automation towards systems capable of continuous reasoning.

This new approach allows infrastructure to be altered dynamically as conditions change, ensuring that the high costs and complex requirements of GPU-heavy environments are managed with greater precision. By providing this real-time oversight, Quali aims to help enterprises maintain governance and cost control as AI agents increasingly take the lead in making infrastructure decisions.

Control layer

Quali positioned the new functionality as a governance layer that sits across existing tooling. It said the Agentic Control Plane interprets workload intent and enforces policy during runtime. It also said it integrates with existing cloud services, infrastructure-as-code tools, and hybrid automation environments.

Torque's existing focus has been platform engineering and infrastructure automation. Quali said the expanded scope targets environments where autonomous AI assistants plan actions and reconfigure infrastructure without waiting for human instruction.

"AI workloads today behave less like automated jobs and more like autonomous assistants operating at machine speed. Traditional automation was never designed to govern systems that continuously adapt and make decisions on their own. Torque provides an intelligent control plane that interprets intent, enforces guardrails in real time, and governs change as it happens, while building on the automation enterprises already have. As infrastructure becomes increasingly autonomous, governance must be continuous, not static," said Lior Koriat, Chief Executive Officer, Quali.

Where it runs

Quali said the Agentic Control-Plane features operate across infrastructure-as-code frameworks, Kubernetes environments, GPU clusters, public cloud platforms, and on-prem systems. It described the control plane as a layer that coordinates human-defined automation and machine-initiated actions.

The company also set out a range of functions it said the platform can run through "modular agents". These include blueprint creation, cost modelling, drift remediation, lifecycle management, and compliance checks.

Quali said these agents act based on workload intent rather than static rules. The company drew a distinction between predefined workflows, which execute when a trigger fires, and systems that take initiative and modify infrastructure in response to changing conditions.

Runtime governance

Quali said the control plane applies security, compliance, and financial controls throughout the lifecycle of an environment. It said these controls adapt as AI systems change behaviour, scale, or topology.

The company also said the platform does not replace existing tooling. It said Torque "enhances" a customer's cloud and automation stack and keeps prior investments operational.

The announcement comes as enterprises expand their use of GPUs for AI training and inference. Many organisations now run a mix of cloud and on-prem infrastructure. They also manage multiple orchestration layers. Quali framed the new release as a response to this operational model and the risk of unmanaged change when autonomous systems make frequent infrastructure decisions.

Quali said the new capabilities govern GPU-dense and hybrid-cloud environments in a way that adapts as agentic AI pipelines evolve. It said the aim is to manage infrastructure changes across clouds, environments, and workloads as those systems operate.

Quali said the Agentic Control-Plane capabilities are available immediately as part of the Torque platform.