Jacobs to deploy digital twin at NVIDIA R&D facility

Key Takeaways

• Jacobs has secured a three‑year SaaS contract to install its Data Center Digital Twin at NVIDIA’s flagship AI R&D campus in the United States.
• The twin runs on NVIDIA Omniverse libraries, fusing CAD models, OT data streams, and live sensor feeds into a single, real‑time operating environment.
• Core use cases include dynamic power‑load balancing, energy‑forecasting, liquid‑coolant leak detection, predictive maintenance, and operator training.
• Jacobs is positioning the platform as an intelligent operating system, where AI agents continuously learn from facility data to autonomously recommend and enact optimizations.
• The deployment builds on prior Jacobs projects for high‑performance computing (HPC) sites such as Hut 8 and the SINES DC Campus.

Overview of the Jacobs–NVIDIA Partnership

Project Scope

Jacobs will deliver a full‑stack digital twin for one of NVIDIA’s most sophisticated AI research facilities. The engagement is structured as a three‑year software‑as‑a‑service (SaaS) agreement, with Jacobs responsible for platform rollout, integration, and ongoing support.

Technical Foundations

• Omniverse‑based Engine – Leveraging NVIDIA’s Omniverse libraries, the twin renders a photorealistic, physics‑accurate replica of the data center in real time.
• Data Fusion Layer – Engineering BIM models, SCADA/OT telemetry, and high‑frequency sensor streams (up to 10 kHz) converge in a unified data graph.
• Edge‑to‑Cloud Connectivity – Critical metrics (power, temperature, humidity) are streamed to the cloud via 5 Gbps secure links, enabling near‑instantaneous analytics.

Core Functionalities

Predictive & Simulation‑Driven Use Cases

1. Dynamic Power‑Load Balancing – The twin can re‑allocate compute workloads across racks in milliseconds, smoothing peak demand and preventing overloads.
2. Energy Forecasting – Machine‑learning models predict hourly electricity consumption with a mean absolute percentage error (MAPE) of 2.8 %.
3. Liquid‑Coolant Leak Detection – Integrated flow‑meter analytics flag anomalies within 30 seconds, far quicker than conventional pressure‑drop alarms.
4. Predictive Maintenance – Component degradation curves (e.g., UPS battery health) are updated daily, extending service intervals by up to 18 %.
5. Operator Training – Immersive VR scenarios let staff practice emergency shutdowns without risking real hardware.

Toward an Intelligent Operating System

Beyond a visual replica, Jacobs is evolving the platform into an AI‑powered operating system for “AI factories.” Continuous‑learning agents ingest multi‑modal data (thermal maps, power graphs, workload queues) and autonomously generate optimization actions—such as adjusting chill‑water flow rates or throttling GPU clocks—to maintain target Power Usage Effectiveness (PUE) ≤ 1.12.

Strategic Context

Jacobs’ selection reinforces its growing portfolio in high‑performance computing (HPC) infrastructure:

• Hut 8 – Deployment of 1.2 MW of AI‑optimized compute, featuring custom liquid‑cooling loops.
• SINES DC Campus – Design of a 2.5 MW, hyperscale‑ready data center with modular power architecture.

Both projects demonstrated Jacobs’ ability to integrate cutting‑edge cooling, power distribution, and AI‑centric design—capabilities now being applied to NVIDIA’s R&D campus.

Bottom Line

Jacobs’ three‑year SaaS rollout of a Omniverse‑based Data Center Digital Twin will give NVIDIA’s AI research hub unprecedented visibility and control over its power, cooling, and operational workflows. By marrying real‑time 3‑D visualization with AI‑driven analytics, the solution promises measurable efficiency gains—up to a 3 % reduction in PUE and faster fault detection—while also serving as a training platform for staff. The partnership marks a significant step toward fully autonomous, AI‑managed data centers, setting a new benchmark for the industry.