Machine Learning pathway increased my salary by 40%
Topic 10 of 12
Advanced
8-10 hours
Digital Twins & Spatial Computing
AI Simulation: Surrogate Models, Neural Simulation and Generative Environments
Master AI-driven simulation, surrogate models, physics-informed ML, neural operators, generative simulation, and synthetic environments. Comprehensive guide covering AI-accelerated simulation for enterprise applications.
AI SimulationSurrogate ModelsPhysics-Informed Neural NetworksNeural OperatorsGenerative SimulationSynthetic EnvironmentsReduced-Order ModelsAI-Accelerated Simulation
Executive Summary
AI simulation represents the convergence of artificial intelligence with traditional simulation methods, creating new capabilities for speed, accuracy, and automation. AI-driven simulation uses surrogate models, physics-informed machine learning, neural operators, and generative techniques to accelerate simulation, enable new types of analysis, and generate synthetic environments for AI training. In 2026, AI simulation is transforming how enterprises design products, optimize operations, train autonomous systems, and generate training data. Traditional numerical simulation remains foundational, but AI-accelerated simulation is enabling new use cases that were previously computationally infeasible. The key is understanding when to use traditional simulation, when to use AI-accelerated simulation, and how to combine both for optimal results.
Definition
AI simulation is the use of artificial intelligence techniques to enhance, accelerate, or generate simulations. It includes AI surrogate models (ML models that approximate expensive simulations), physics-informed neural networks (PINNs that incorporate physical laws into ML training), neural operators (ML models that learn solution operators for PDEs), generative simulation (AI-generated simulation scenarios and environments), synthetic environments (AI-generated virtual worlds for training and testing), and simulation acceleration (using AI to speed up traditional simulation). AI simulation bridges the gap between the accuracy of traditional physics-based simulation and the speed of AI inference.
Why It Matters
Traditional physics-based simulation is accurate but computationally expensive. AI simulation can accelerate simulation by orders of magnitude while maintaining acceptable accuracy, enabling real-time simulation, design space exploration, and optimization that were previously infeasible. AI simulation also enables new capabilities: generative simulation creates novel scenarios, synthetic environments provide training data for AI, and physics-informed ML combines the best of physics and data-driven approaches. Simulation is becoming infrastructure for AI—providing training data, testing environments, and validation for autonomous systems. The convergence of AI and simulation is a key trend in 2026, with platforms like NVIDIA Omniverse, Isaac Sim, and PhysicsNeMo leading the way.
Evolution of the Technology
Traditional Numerical Simulation (1950s-2000s): FEA, CFD, and numerical methods. Accurate but computationally expensive. Reduced-Order Models (2000s-2010s): Simplified models that capture essential physics with less computation. Machine Learning Surrogates (2010s): ML models trained to approximate simulation results. Faster but may sacrifice accuracy. Physics-Informed Neural Networks (2010s-2020s): PINNs incorporate physical laws into neural network training. Combines physics and ML. Neural Operators (2020s): Learn solution operators for PDEs. Generalize across different input conditions. Generative Simulation (2025+): AI-generated simulation scenarios and environments. Automated scenario creation. Simulation as AI Infrastructure (2026): Simulation as infrastructure for AI training, testing, and validation.
Learning Objectives
- Understand AI simulation methods: surrogate models, PINNs, neural operators, generative simulation
- Master the trade-offs between traditional and AI-accelerated simulation
- Understand physics-informed neural networks and their applications
- Design AI simulation pipelines for enterprise applications
- Understand synthetic environment generation for AI training
- Evaluate accuracy, speed, generalization, and uncertainty in AI simulation
- Apply AI simulation across industries: manufacturing, automotive, aerospace, energy
- Integrate AI simulation with digital twins and autonomous systems
- Design validation frameworks for AI-accelerated simulation
- Understand the 2026 technology landscape and emerging trends in AI simulation
Prerequisites
- Strong understanding of simulation and modeling
- Familiarity with machine learning and deep learning
- Understanding of physics and engineering principles
- Basic knowledge of PDEs and numerical methods
- Familiarity with digital twin concepts
Traditional vs AI-Accelerated Simulation
Traditional numerical simulation uses mathematical models to solve physical equations directly—accurate but computationally expensive. AI-accelerated simulation uses machine learning to approximate or enhance simulation—faster but may sacrifice some accuracy. The key is understanding when to use each approach and how to combine them. Physics-based models are prerequisites for detailed representation, but computational complexity poses challenges. Surrogate/reduced-order models created with machine learning optimize speed while preserving fidelity.
Traditional vs AI-Accelerated Simulation
| Aspect | Traditional Simulation | AI-Accelerated Simulation |
|---|---|---|
| Method | Numerical solution of equations | ML approximation of solutions |
| Accuracy | High (physics-based) | Good (depends on training) |
| Speed | Slow (hours to days) | Fast (milliseconds to seconds) |
| Generalization | Limited to specific model | Can generalize across conditions |
| Data Requirements | Minimal (physics-based) | Large training datasets needed |
| Interpretability | High (physics-based) | Lower (black box) |
| Uncertainty | Well-characterized | Needs quantification |
| Best For | Detailed engineering analysis | Real-time, optimization, exploration |
Physics-Informed Neural Networks (PINNs)
Physics-Informed Neural Networks (PINNs) are neural networks that incorporate physical laws (expressed as PDEs) into their training process. Instead of learning purely from data, PINNs learn to satisfy both data and physical constraints. This combines the flexibility of neural networks with the reliability of physics-based modeling. PINNs are particularly useful for problems with limited data but known physics, inverse problems, and real-time simulation. They bridge the gap between purely data-driven ML and purely physics-based simulation.
AI Simulation Techniques
| Technique | How It Works | Strengths | Maturity |
|---|---|---|---|
| Surrogate Models | ML trained to approximate simulation | Fast inference, good for optimization | Enterprise Adoption |
| PINNs | Neural networks with physics constraints | Data-efficient, physics-consistent | Early Adoption |
| Neural Operators | Learn solution operators for PDEs | Generalize across conditions | Research |
| Generative Simulation | AI generates simulation scenarios | Automated scenario creation | Emerging |
| Reduced-Order Models | Simplified models capturing essential physics | Fast, physics-based | Enterprise Adoption |
| Hybrid Approaches | Combine physics + ML | Best of both worlds | Early Adoption |
Synthetic Environments and Generative Simulation
Synthetic environments are AI-generated virtual worlds used for training and testing AI systems, particularly for computer vision, robotics, and autonomous vehicles. Generative simulation uses AI to automatically create simulation scenarios, enabling rapid exploration of design spaces and edge cases. These technologies are transforming how AI systems are trained and tested. Instead of collecting massive real-world datasets (expensive, time-consuming, and limited), synthetic environments can generate unlimited diverse training data with perfect labels. This is particularly valuable for edge cases and rare events that are difficult to capture in the real world.
Synthetic Environment Applications
| Application | Description | Benefits | Maturity |
|---|---|---|---|
| Computer Vision Training | Synthetic images for CV model training | Unlimited labeled data, edge cases | Enterprise Adoption |
| Robotics Training | Virtual environments for robot learning | Safe, repeatable, diverse scenarios | Early Adoption |
| AV Testing | Synthetic driving scenarios | Edge cases, rare events, safety | Early Adoption |
| Industrial AI | Synthetic factory environments | Process optimization, safety testing | Emerging |
| Smart City Simulation | Synthetic urban environments | Traffic, planning, emergency scenarios | Emerging |
| AI Safety Testing | Synthetic adversarial scenarios | Safety validation before deployment | Research |
Architecture
The AI simulation architecture stack connects traditional simulation to AI acceleration and generative capabilities.
1
Traditional Simulation
FEA, CFD, multiphysics, discrete-event simulation. Physics-based numerical methods. Ground truth for AI training.
2
Data Collection
Collect simulation results, sensor data, and experimental data for AI model training. High-quality, diverse datasets.
3
AI Model Training
Train surrogate models, PINNs, neural operators, and generative models using simulation and real-world data.
4
AI Acceleration
Use trained AI models for fast inference, replacing or enhancing traditional simulation for real-time applications.
5
Generative Simulation
AI-generated scenarios, environments, and synthetic data for training, testing, and exploration.
6
Validation & Uncertainty
Validate AI simulation against traditional simulation and real-world data. Quantify uncertainty and confidence.
7
Integration & Deployment
Integrate AI simulation into digital twins, autonomous systems, and enterprise workflows.
AI Surrogate Model Pipeline
From traditional simulation to AI-accelerated inference
1
TRADITIONAL SIMULATION: Run high-fidelity simulation for training data
2
DATA COLLECTION: Collect input-output pairs
3
AI MODEL TRAINING: Train surrogate model (neural network)
4
VALIDATION: Validate against held-out simulation data
5
DEPLOYMENT: Deploy for fast inference
6
CONTINUOUS IMPROVEMENT: Update with new simulation data
Generative Simulation Pipeline
From scenario requirements to synthetic environments
1
REQUIREMENTS: Define scenario requirements and parameters
2
GENERATION: AI generates simulation scenarios
3
VARIATION: Create variations and edge cases
4
VALIDATION: Validate physical plausibility
5
TRAINING: Use for AI model training
6
TESTING: Test AI systems in synthetic environments
7
DEPLOYMENT: Deploy trained AI to real world
AI Simulation Technology Stack
| Technology | Provider | Capability | Maturity |
|---|---|---|---|
| NVIDIA Omniverse | NVIDIA | 3D simulation platform with AI integration | Enterprise Adoption |
| NVIDIA Isaac Sim | NVIDIA | Robotics simulation with AI | Enterprise Adoption |
| NVIDIA PhysicsNeMo | NVIDIA | Physics-informed neural networks | Early Adoption |
| Ansys + AI | Ansys | Traditional simulation with AI acceleration | Enterprise Adoption |
| Siemens Simcenter | Siemens | Simulation with AI integration | Enterprise Adoption |
| TensorFlow/PyTorch | Open Source | ML frameworks for surrogate models | Production Ready |
| NVIDIA Modulus | NVIDIA | Physics-informed ML framework | Early Adoption |
| DeepXDE | Open Source | PINN framework | Research |
AI Simulation Use Cases by Industry
| Industry | Use Case | AI Technique | Benefit |
|---|---|---|---|
| Manufacturing | Design optimization | Surrogate models | 100x faster design space exploration |
| Automotive | Crash simulation | Neural operators | Real-time crash prediction |
| Aerospace | Aerodynamics | PINNs | Fast aerodynamic analysis |
| Energy | Grid simulation | Surrogate models | Real-time grid optimization |
| Robotics | Robot training | Generative simulation | Unlimited training scenarios |
| AV | Scenario testing | Synthetic environments | Edge case coverage |
| Healthcare | Patient modeling | PINNs | Personalized prediction |
| Smart Cities | Traffic simulation | AI-accelerated simulation | Real-time optimization |
Enterprise Use Cases
Manufacturing
Enterprise Adoption
Design Optimization
AI surrogate models enable 100x faster design space exploration. Real-time optimization of product designs.
Automotive
Early Adoption
Crash Simulation
Neural operators for real-time crash prediction. Faster vehicle design iteration.
Aerospace
Early Adoption
Aerodynamics
PINNs for fast aerodynamic analysis. Rapid design exploration for aircraft and components.
Energy
Enterprise Adoption
Grid Simulation
Surrogate models for real-time grid optimization. Renewable energy integration analysis.
Robotics
Early Adoption
Robot Training
Generative simulation for unlimited training scenarios. Sim-to-real transfer for robot learning.
Automotive
Early Adoption
AV Scenario Testing
Synthetic environments for autonomous vehicle testing. Edge case generation and safety validation.
Healthcare
Research
Patient Modeling
PINNs for patient-specific disease modeling and treatment planning.
Smart Cities
Early Adoption
Traffic AI Simulation
AI-accelerated traffic simulation for real-time optimization and planning.
Case Studies
NVIDIA PhysicsNeMo for Physics-Informed ML
Problem: Need for physics-informed machine learning that combines physics accuracy with ML speed.
Architecture: NVIDIA PhysicsNeMo provides a framework for physics-informed neural networks that incorporate physical laws into ML training.
Outcome: Enables physics-consistent ML models that are faster than traditional simulation while maintaining physical accuracy. Used for semiconductor manufacturing, energy, and aerospace applications.
Lessons: Physics-informed ML bridges the gap between traditional simulation and pure ML. Combining physics and data-driven approaches provides best of both worlds.
AI-Accelerated Design Optimization
Problem: Traditional design optimization requires many expensive simulations, limiting exploration.
Architecture: AI surrogate models trained on simulation data replace expensive simulations for rapid design space exploration.
Outcome: 100x faster design space exploration enables more thorough optimization and better designs. Used in automotive, aerospace, and manufacturing.
Lessons: AI surrogates dramatically accelerate design optimization. Traditional simulation provides ground truth for training. Hybrid approaches are most effective.
Implementation Steps
1
Use Case Assessment
Identify simulation use cases where AI acceleration would provide value. Assess data availability, accuracy requirements, and speed needs.
2
Data Generation
Generate training data using traditional simulation. Ensure diversity and coverage of the design space. Collect real-world data for validation.
3
AI Model Development
Select appropriate AI technique (surrogate, PINN, neural operator). Train and validate AI model against traditional simulation.
4
Validation & Uncertainty
Validate AI simulation against traditional simulation and real-world data. Quantify uncertainty and establish confidence intervals.
5
Integration & Deployment
Integrate AI simulation into workflows, digital twins, and autonomous systems. Monitor performance and update models.
Enterprise SOPs
SOP: AI Surrogate Model Development
Purpose: Standardize development of AI surrogate models
Owner: Simulation Engineering Team
Steps: Use Case Definition → Data Generation → Model Training → Validation → Uncertainty Quantification → Deployment
Evidence: Model documentation, validation reports, uncertainty analysis
SOP: Physics-Informed ML Development
Purpose: Develop physics-informed neural networks
Owner: AI Research Team
Steps: Physics Definition → Network Architecture → Training with Physics Constraints → Validation → Deployment
Evidence: Physics documentation, training logs, validation results
SOP: Synthetic Environment Generation
Purpose: Generate synthetic environments for AI training
Owner: Simulation Team
Steps: Requirements → Environment Generation → Validation → Training Data Creation → Quality Assurance → Deployment
Evidence: Environment specifications, quality reports, training data documentation
Lab: Build an AI Simulation Pipeline
Problem: A manufacturing company needs to optimize a complex product design but traditional simulation is too slow for design space exploration.
Requirements:
- Identify the simulation use case
- Generate training data from traditional simulation
- Train an AI surrogate model
- Validate against traditional simulation
- Quantify uncertainty
- Deploy for fast design optimization
- Measure speed improvement
- Document the pipeline
Architecture: Traditional Simulation → Data Collection → AI Model Training → Validation → Fast Inference → Design Optimization
Outcome: AI-accelerated simulation pipeline with validated surrogate model, uncertainty quantification, and significant speed improvement.
GCC Applications
- AI simulation model development for global enterprises
- Surrogate model training and validation services
- Physics-informed ML research and development
- Synthetic environment generation for AI training
- AI-accelerated simulation center of excellence
- Generative simulation for design exploration
- Simulation-as-a-Service with AI acceleration
- Digital twin AI simulation integration
Key Metrics & KPIs
Speed Improvement
How much faster AI simulation is compared to traditional simulation
Accuracy
Accuracy of AI simulation compared to traditional simulation and real-world
Generalization
Ability to perform on inputs not seen during training
Uncertainty Quantification
Quality of uncertainty estimates from AI simulation
Training Data Efficiency
How much traditional simulation data is needed for training
Design Space Coverage
How thoroughly the AI model covers the design space
Real-Time Performance
Whether AI simulation can run in real-time
Physics Consistency
Degree to which AI simulation respects physical laws
Synthetic Data Quality
Quality and usefulness of generated synthetic data
ROI
Value delivered through faster simulation and better designs
Risks & Mitigation
Accuracy Loss
Mitigation: Rigorous validation, uncertainty quantification, hybrid approaches with traditional simulation
Generalization Failure
Mitigation: Diverse training data, domain adaptation, continuous learning, monitoring
Physics Inconsistency
Mitigation: Physics-informed training, validation against physical laws, expert review
Over-Reliance on AI
Mitigation: Human oversight, traditional simulation for critical decisions, safety constraints
Training Data Quality
Mitigation: High-quality simulation data, diverse coverage, validation datasets
Computational Cost of Training
Mitigation: Efficient training methods, transfer learning, cloud computing
Maturity Model
1
Traditional Simulation
Pure physics-based numerical simulation. No AI.
2
AI-Enhanced
AI used for post-processing or optimization of traditional simulation.
3
AI-Accelerated
AI surrogate models replace traditional simulation for fast inference.
4
Physics-Informed AI
PINNs and neural operators combine physics and ML.
5
Generative Simulation
AI generates simulation scenarios and environments automatically.
6
Autonomous Simulation
AI agents run simulations, analyze results, and optimize autonomously.
7
Simulation as AI Infrastructure
Simulation as foundational infrastructure for AI training, testing, and validation.
2026 Trends & Emerging Developments
Physics-Informed Neural Networks
Early Adoption
Neural Operators for PDEs
Research
Generative Simulation
Emerging
Synthetic Environment Generation
Enterprise Adoption
AI-Accelerated Design Optimization
Enterprise Adoption
Simulation as AI Infrastructure
Emerging
Hybrid Physics + ML Approaches
Early Adoption
Real-Time AI Simulation
Early Adoption
NVIDIA PhysicsNeMo/Modulus
Early Adoption
Autonomous Simulation Agents
Research
Career Applications
AI Simulation EngineerPhysics-Informed ML SpecialistSurrogate Model DeveloperNeural Operator ResearcherGenerative Simulation EngineerSynthetic Environment DeveloperAI-Accelerated Simulation ArchitectDigital Twin Simulation SpecialistAI Simulation ResearcherComputational AI EngineerSimulation AI ArchitectDesign Optimization EngineerAI Modeling ScientistSimulation Infrastructure EngineerAutonomous Simulation Researcher
Frequently Asked Questions
Q: What is AI simulation?
A: AI simulation is the use of artificial intelligence techniques to enhance, accelerate, or generate simulations, including surrogate models, PINNs, neural operators, and generative simulation.
Q: What is a surrogate model?
A: A surrogate model is a machine learning model trained to approximate the results of a more expensive traditional simulation, enabling fast inference.
Q: What are physics-informed neural networks (PINNs)?
A: PINNs are neural networks that incorporate physical laws (expressed as PDEs) into their training process, combining physics and data-driven learning.
Q: What are neural operators?
A: Neural operators are ML models that learn solution operators for partial differential equations, enabling generalization across different input conditions.
Q: What is generative simulation?
A: Generative simulation uses AI to automatically create simulation scenarios and environments, enabling rapid exploration and edge case generation.
Q: What are synthetic environments?
A: Synthetic environments are AI-generated virtual worlds used for training and testing AI systems, particularly for computer vision, robotics, and autonomous vehicles.
Q: How does AI simulation compare to traditional simulation?
A: AI simulation is faster but may sacrifice some accuracy. Traditional simulation is more accurate but slower. Hybrid approaches combine both for optimal results.
Q: What is reduced-order modeling?
A: Reduced-order modeling creates simplified models that capture essential physics with less computation, providing a middle ground between full simulation and AI surrogates.
Q: How do you validate AI simulation?
A: Validation includes comparison to traditional simulation, real-world data validation, uncertainty quantification, and physics consistency checks.
Q: What is NVIDIA PhysicsNeMo?
A: NVIDIA PhysicsNeMo is a framework for physics-informed neural networks that combine physics accuracy with ML speed.
Q: What is NVIDIA Modulus?
A: NVIDIA Modulus is a physics-informed ML framework for building neural network models that incorporate physical laws.
Q: How does AI simulation support design optimization?
A: AI surrogates enable 100x faster design space exploration, allowing more thorough optimization and better designs.
Q: What is the reality gap in AI simulation?
A: The reality gap is the difference between AI simulation predictions and real-world behavior, caused by training data limitations and unmodeled physics.
Q: How does AI simulation support autonomous vehicles?
A: AI simulation generates synthetic driving scenarios for AV training and testing, including edge cases and rare events.
Q: What is the role of AI simulation in digital twins?
A: AI simulation provides fast predictive capabilities for digital twins, enabling real-time what-if analysis and optimization.
Q: How do you handle uncertainty in AI simulation?
A: Uncertainty is handled through uncertainty quantification methods, ensemble models, Bayesian approaches, and confidence intervals.
Q: What is simulation as AI infrastructure?
A: Simulation as AI infrastructure means simulation is used as foundational infrastructure for AI training, testing, and validation at enterprise scale.
Q: How does AI simulation support robotics?
A: AI simulation generates training environments for robot learning, enables sim-to-real transfer, and provides safe testing for robot behaviors.
Q: What are the limitations of AI simulation?
A: Limitations include accuracy trade-offs, generalization challenges, training data requirements, physics consistency, and computational cost of training.
Q: How do you train a surrogate model?
A: Surrogate models are trained on input-output pairs from traditional simulation, using supervised learning to approximate the simulation function.
Q: What is the difference between PINNs and standard neural networks?
A: PINNs incorporate physical laws as constraints in training, while standard neural networks learn purely from data without physics constraints.
Q: How does AI simulation support manufacturing?
A: AI simulation enables fast design optimization, process simulation, quality prediction, and virtual commissioning for manufacturing.
Q: What is the future of AI simulation?
A: The future includes physics-informed AI, generative simulation, autonomous simulation agents, and simulation as AI infrastructure.
Q: How do you select between AI and traditional simulation?
A: Use traditional simulation for high-accuracy needs and AI simulation for speed, optimization, and real-time applications. Hybrid approaches often work best.
Q: What is the role of AI simulation in AI safety?
A: AI simulation provides safe environments for testing AI systems, generating adversarial scenarios, and validating safety before real-world deployment.
Q: How does AI simulation support aerospace?
A: AI simulation enables fast aerodynamic analysis, structural optimization, and flight dynamics simulation for aerospace design.
Q: What is the relationship between AI simulation and world models?
A: AI simulation can generate training data for world models. World models can serve as learned simulators. They are complementary technologies.
Q: How do you ensure physics consistency in AI simulation?
A: Physics consistency is ensured through physics-informed training, validation against physical laws, expert review, and hybrid physics-ML approaches.
Q: What is the computational cost of AI simulation?
A: Training AI simulation models can be expensive, but inference is typically much faster than traditional simulation, providing net computational savings.
Q: How does AI simulation support energy applications?
A: AI simulation enables real-time grid optimization, renewable energy forecasting, and rapid analysis of energy systems.
Q: What is the role of synthetic data in AI simulation?
A: Synthetic data generated from AI simulation provides unlimited labeled training data for AI models, particularly for edge cases and rare events.
Q: How do you measure AI simulation ROI?
A: ROI is measured through speed improvements, better designs, reduced computational costs, and faster time-to-market.
Q: What is the relationship between AI simulation and generative AI?
A: Generative AI techniques are used in AI simulation to generate scenarios, environments, and synthetic data. They are complementary technologies.
Q: How does AI simulation support healthcare?
A: AI simulation enables patient-specific modeling, disease progression prediction, and treatment planning with physics-informed ML.
Q: What is the role of cloud computing in AI simulation?
A: Cloud computing provides scalable infrastructure for training AI simulation models and running large-scale simulations.
Q: How do you handle distribution shift in AI simulation?
A: Distribution shift is handled through domain adaptation, continuous learning, monitoring, and retraining with new data.
Q: What is the enterprise potential of AI simulation?
A: Enterprise potential includes faster design cycles, real-time optimization, better products, AI training infrastructure, and competitive advantage.
Q: How does AI simulation relate to Industry 4.0?
A: AI simulation is a key enabler of Industry 4.0, providing fast simulation for smart manufacturing, digital twins, and autonomous operations.
Q: What is the role of AI simulation in sustainability?
A: AI simulation enables energy optimization, waste reduction, and better resource efficiency through fast analysis and optimization.
Q: How do you build an AI simulation team?
A: Build a team with simulation engineers, ML specialists, domain experts, and validation engineers with strong cross-disciplinary collaboration.
Q: What is the future of simulation technology?
A: The future includes AI-accelerated simulation, generative environments, physics-informed AI, autonomous simulation agents, and simulation as foundational AI infrastructure.
Research References
NVIDIA PhysicsNeMo (2026). "Physics-informed neural network framework." [Industry]
NVIDIA Modulus (2026). "Physics-informed ML framework for PDEs." [Industry]
Raissi, M., et al. (2019). "Physics-informed neural networks: A deep learning framework for solving forward and inverse PDE problems." [Research]
Li, Z., et al. (2020). "Neural Operator: Graph Kernel Network for PDEs (Fourier Neural Operator)." [Research]
NVIDIA Omniverse (2026). "3D simulation platform with AI integration." [Industry]
Ansys AI (2026). "AI-accelerated simulation capabilities." [Industry]