Berliner Tageszeitung - Klique Announces Expanded Integrations with NVIDIA to Accelerate Enterprise AI Workflows and Optimize GPU Utilization

SAN FRANCISCO, CA / ACCESS Newswire / October 21, 2026 / Klique, an enterprise control plane for AI resource management, governance, and cost visibility, today announced a suite of new integrations with NVIDIA designed to optimize GPU performance and deliver enhanced capabilities for enterprises.

Building on a strong foundation of its collaboration with NVIDIA, Klique helping enterprises pair NVIDIA's cutting-edge infrastructure with secure enterprise execution across three key operational areas:

Optimizing Compute Cost and Data Sovereignty
The NVIDIA Nemotron family of open models provides open, high-efficiency, low-latency performance designed for high-volume agentic workflows on a single GPU. By running Nemotron 3.5 Super VL on Klique, enterprise users can deploy NVIDIA's model directly on their own hardware, ensuring no data leaves the tenant while eliminating third-party per-token billing. Complementing this, Klique customers can now leverage NVIDIA NeMo Switchyard's intelligent model routing algorithms alongside Klique's LLM gateway, giving organizations the flexibility to optimize spend on frontier models or ensure strict data sovereignty.

Running Agents at Production Scale
A single agent task can trigger many model calls as the agent plans, uses tools, and reasons through each step, so inference demand climbs fast once agents reach production. NVIDIA Dynamo is built for that load: it distributes inference across GPU workers and sends repeated requests to the workers that already hold the relevant cached context. Klique customers can deploy NVIDIA Dynamo directly from the Klique platform and run agents at production scale on their own GPU infrastructure or in the cloud, with Klique adding both the security access rights and observability as well as LLM routing making sure large scale agentic workflows are diverted into cost effective AI resources. Teams can also use Klique to retrain agents on the results of their past work, and the platform records the data, model version, and compute cost behind every update so each change can be reviewed and traced. With Dynamo serving inference across their GPUs and Klique tracking every agent's usage, cost, and changes, enterprises can scale agents to production demand on hardware they already own.

Securing Autonomous Agent Workflows
NVIDIA OpenShell establishes a secure, monitored runtime environment designed to protect internal files, networks, and credentials during agent execution. By integrating OpenShell into the platform, Klique allows enterprises to deploy autonomous agents like those built with NemoClaw open blueprints, Claude Code, Pi, Hermes, and Codex with total confidence, providing the governance layer, enterprise policy compliance, and audit trails required for production agentic workflows.

"Scaling production-grade agentic workflows shouldn't force enterprises to choose between spiraling cloud costs and data privacy," said Moses Guttmann, CEO and CO-founder, Klique. "Our expanded collaboration with NVIDIA pairs low-latency models like Nemotron 3.5 Super VL and smart routing algorithms via NeMo Switchyard with Klique's cost control and governance framework. This allows our customers to maximize GPU utilization, maintain strict data sovereignty, and run frontier-level AI workloads directly on their infrastructure without third-party per-token inflation."

"Every enterprise is working out how to run agents efficiently and under its own policies," said John Fanelli, vice president, Enterprise Software at NVIDIA. "By building on NVIDIA Nemotron, Dynamo and OpenShell, Klique helps customers put agents to work within their unique environments."

To learn more about Klique's integrations with NVIDIA, visit https://klique.ai.

About Klique

Klique (formerly ClearML) is the control plane for enterprise AI, where the people building AI, the people using it, and the IT teams running the infrastructure all work under the same rules. It routes every AI request to the right commercial or self-hosted model based on cost, complexity, and data sensitivity; enforces budgets, quotas, and access policies once, everywhere; and manages GPUs and compute on any silicon, from Kubernetes and bare metal to fully air-gapped environments. Every token, GPU cycle, and dollar is traced back to the project that spent it, so AI spend grows only as fast as policy allows. Headquartered in Berkeley, California, Klique is trusted by more than 2,100 organizations and 300,000 AI builders worldwide. Learn more at klique.ai.

Contact Information

Rence Holt
Phone: +14048196709

SOURCE: Klique

View the original press release on ACCESS Newswire

D. O'Sullivan--BTZ