Energy and utilities
How utilities, power producers and energy companies apply NVIDIA simulation, robotics, edge computing and AI factory software to grid operations, plant construction and data center power, with safety, critical infrastructure security and customer data rules treated as design constraints.
The problem
Electricity demand is rising faster than grids can connect new supply. NVIDIA's energy page names power availability, interconnection delays and grid capacity limits as the brakes on growth, and calls energy the key constraint on scaling AI infrastructure itself; an NVIDIA podcast with EPRI on the same page looks at how AI inference is reshaping electricity demand.
Building new generation takes labor and time. Solar construction depends on crews placing and fastening heavy modules, and NVIDIA links field robotics to the gap between demand and construction capacity. On the distribution side, distributed energy resources add activity at the edge of the network, and Utilidata pitches analytics inside meters and grid devices to manage them.
Exploration and surface operations carry their own cost and safety pressures. Any AI that touches the grid or a plant is part of critical infrastructure: the EU AI Act lists AI used as a safety component in critical infrastructure among its high-risk categories.1234
The approach
Four patterns appear in NVIDIA's energy work. First, digital twins: Omniverse libraries let software makers add physically based simulation of energy assets to their own tools, and NVIDIA says twins of surface operations halve research-to-operation timelines. Second, robotics trained in simulation: AES's Maximo developed its solar installation robots with Omniverse libraries and Isaac Sim and reports 100 MW installed at the Bellefield complex. Third, computing at the grid edge: Utilidata's Karman platform runs on a custom module built on Jetson, and Aclara, a Hubbell division, announced plans to embed it in smart meters. Fourth, the data center as a grid participant: NVIDIA's DSX platform helps facilities sync with the grid, and NVIDIA states it can run up to 40 percent more GPUs within a fixed power budget.
NVIDIA also positions AI Enterprise for deploying open models and tools across energy companies, and says producers including AES, Constellation, NextEra Energy and Vistra plan to work with it and Emerald AI on generation strategies for AI factories.
Many problems need less. Established power flow and planning software, SCADA and outage management analytics, and classical optimization solvers already handle much grid work, and a utility's existing ADMS or meter vendor may offer the analytics it needs. Dedicated GPUs make sense when simulation scale, high-resolution sensing or fleets of robots exceed what those tools do.1235
Conceptual architecture
Diagram as a list
Applications & solutions
- Construction and inspection robots trained in simulation (Isaac Sim)Install modules or inspect assets alongside crewsConnects to Field assets and sensors (meters, feeders, substations, plants, solar sites)
- Utility operations systems (SCADA, ADMS, outage management) with AI modelsOperators review forecasts and recommendations before acting on the grid
Models & frameworks
- Asset and site digital twins (Omniverse libraries in engineering tools)Simulate plants, sites and equipment before changes in the fieldConnects to Construction and inspection robots trained in simulation (Isaac Sim), Utility operations systems (SCADA, ADMS, outage management) with AI models
Operations & orchestration
- Field assets and sensors (meters, feeders, substations, plants, solar sites)Produce measurements, video and status signals from the physical systemConnects to Grid-edge computers inside meters and devices (Jetson-based modules)
Accelerated computing
- Grid-edge computers inside meters and devices (Jetson-based modules)Analyze readings locally and send events rather than raw waveformsConnects to Utility operations systems (SCADA, ADMS, outage management) with AI models
- Power-flexible AI factory (DSX platform)Adjusts data center load in response to grid conditionsConnects to Utility operations systems (SCADA, ADMS, outage management) with AI models
Technologies and their roles
NVIDIA Omniverse2
Digital twins of energy assets and robot development
NVIDIA says Omniverse libraries let software makers simulate energy infrastructure, and Maximo used Omniverse libraries to develop and test its solar robots.
NVIDIA Isaac5
Robot simulation for solar construction
Maximo says it used the Isaac Sim robotics simulation framework to test and refine robot capabilities before field updates at AES's Bellefield complex.
NVIDIA Jetson3
Computing inside meters and grid devices
Utilidata says its Karman grid-edge platform is built on a custom module using the Jetson platform, and Aclara announced it would embed Karman in a smart meter.
NVIDIA DSX1
Grid-aware AI data centers
NVIDIA says DSX helps data centers sync with the grid programmatically and states it can run up to 40 percent more GPUs within a fixed power budget.
NVIDIA AI Enterprise1
Software platform for energy AI workloads
NVIDIA's energy page presents AI Enterprise for deploying open source tools and AI models and running AI workloads at scale.
What you need first
- A security architecture for operational technology that defines which networks AI systems may read from and which they may never write to
- Time-synchronized, well-labeled sensor and asset data with owners in operations, not only in IT
- Engineering acceptance criteria for any model output that informs switching, dispatch or protection settings
- Site safety procedures and union or workforce agreements before robots work beside crews
- Accurate asset geometry and CAD or BIM data if digital twins are planned
- Regulatory review of cost recovery for new technology in rate cases, where applicable
- A clear owner for each pilot with a measured baseline, such as installation rate or outage minutes
Risks and how to reduce them
- Unsafe automated actions on the grid or in a plant
- Keep AI in an advisory role for switching and protection until validated, require operator approval, and test changes in simulation or a lab first.
- Cyber attacks through new edge devices and AI services
- Treat grid-edge computers as critical assets: signed firmware, network segmentation, supply chain review and monitoring under the utility's existing cybersecurity program.
- Exposure of household consumption data
- Process meter data locally where possible, share only aggregates, and follow state and national privacy rules for customer energy data.
- Worker injury around robots on construction sites
- Define exclusion zones and stop procedures, train crews, and log every incident and near miss during rollout.
- Investment based on vendor performance claims
- Ask for baselines and conditions behind figures such as GPUs per power budget or installation rates, and verify them in a pilot.
Documented examples
The AES Corporation, through Maximo (its incubated solar robotics company) · Energy and utilities (utility-scale solar construction)
AES and Maximo: solar installation robots developed in Omniverse and Isaac Sim
Maximo, the solar robotics company AES incubated, reported 100 MW of panels installed by a fleet of four robots at AES's Bellefield complex in March 2026. It credits NVIDIA Omniverse libraries and Isaac Sim with letting it test robot updates in simulation before sending them to the field.
In production
Related
Sources
- Global Energy Solutions for the Energy Industry With AI (opens in a new tab)
- NVIDIA, Energy Leaders Accelerating Power-Flexible AI Factories to Fortify the Grid (opens in a new tab)
- Utilidata Partners with Aclara to Bring Distributed AI to the Grid Edge (opens in a new tab)
- AI Act: regulatory framework for AI (opens in a new tab)
- Maximo Completes 100 MW of Robotic Solar Installation (opens in a new tab)
Thank you. Your correction was sent.
The editors check it against the sources. If you left an email address, they may reply about it.